d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
6dc32306e9
freebsd-dev/sys/dev/cxgbe/t4_main.c

10903 lines
283 KiB
C
Raw Normal View History

cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
/*-
sys/dev: further adoption of SPDX licensing ID tags. Mainly focus on files that use BSD 2-Clause license, however the tool I was using misidentified many licenses so this was mostly a manual - error prone - task. The Software Package Data Exchange (SPDX) group provides a specification to make it easier for automated tools to detect and summarize well known opensource licenses. We are gradually adopting the specification, noting that the tags are considered only advisory and do not, in any way, superceed or replace the license texts.
2017-11-27 14:52:40 +00:00
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
* Copyright (c) 2011 Chelsio Communications, Inc.
* All rights reserved.
* Written by: Navdeep Parhar <np@FreeBSD.org>
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
Add a 'show t4 tcb <nexus> <tid>' command to dump a TCB from DDB. This allows the contents of a TCB to be extracted from a T4/T5 card in DDB after a panic.
2016-04-10 05:06:58 +00:00
#include "opt_ddb.h"
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
#include "opt_inet.h"
cxgbe(4): support for IPv6 TSO and LRO. Submitted by: bz (this is a modified version of that patch)
2012-06-29 19:51:06 +00:00
#include "opt_inet6.h"
cxgbe(4): Allocate offload Tx queues when a card has resources provisioned for NIC_ETHOFLD and the kernel has option RATELIMIT. It is possible to use the chip's offload queues for normal NIC Tx and not just TOE Tx. The difference is that these queues support out of order processing of work requests and have a per-"flowid" mechanism for tracking credits between the driver and hardware. This allows Tx for any number of flows bound to different rate limits to be submitted to a single Tx queue and the work requests for slow flows won't cause HOL blocking for the rest. Sponsored by: Chelsio Communications
2018-05-17 01:42:18 +00:00
#include "opt_ratelimit.h"
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
#include "opt_rss.h"
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
#include <sys/param.h>
#include <sys/conf.h>
#include <sys/priv.h>
#include <sys/kernel.h>
#include <sys/bus.h>
#include <sys/module.h>
Resume tx immediately in response to an SGE egress update from the hardware. MFC after: 1 week
2011-03-05 03:06:38 +00:00
#include <sys/malloc.h>
#include <sys/queue.h>
#include <sys/taskqueue.h>
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
#include <sys/pciio.h>
#include <dev/pci/pcireg.h>
#include <dev/pci/pcivar.h>
#include <dev/pci/pci_private.h>
#include <sys/firmware.h>
Display holdoff timers and packet counts as a list of numbers. MFC after: 1 week
2011-03-09 21:07:09 +00:00
#include <sys/sbuf.h>
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
#include <sys/smp.h>
#include <sys/socket.h>
#include <sys/sockio.h>
#include <sys/sysctl.h>
#include <net/ethernet.h>
#include <net/if.h>
#include <net/if_types.h>
#include <net/if_dl.h>
Add missing header. The test for VLAN_CAPABILITIES later in the file doesn't make sense without it. MFC after: 3 days
2011-05-17 00:40:11 +00:00
#include <net/if_vlan_var.h>
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
#ifdef RSS
#include <net/rss_config.h>
#endif
cxgbe(4): improve the accuracy of various TSO limits reported to the kernel. Sponsored by: Chelsio Communications
2018-10-22 23:57:59 +00:00
#include <netinet/in.h>
#include <netinet/ip.h>
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
#if defined(__i386__) || defined(__amd64__)
cxgbe(4): Add a knob to enable/disable PCIe relaxed ordering. Disable it by default when running on Intel CPUs. This is a crude fix for the performance issues alluded to in these Linux commits: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=87e09cdec4dae08acdb4aa49beb793c19d73e73e https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=a99b646afa8a02571ea298bedca6592d818229cd MFC after: 1 week Sponsored by: Chelsio Communications
2018-01-03 19:24:57 +00:00
#include <machine/md_var.h>
#include <machine/cputypes.h>
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
#include <vm/vm.h>
#include <vm/pmap.h>
#endif
Move #include for rijndael.h out of x86-specific region. The #include was added inside of the conditional by accident and the lack of it broke non-x86 builds. Reported by: lwhsu (jenkins), andrew
2018-02-27 17:51:58 +00:00
#include <crypto/rijndael/rijndael.h>
Add a 'show t4 tcb <nexus> <tid>' command to dump a TCB from DDB. This allows the contents of a TCB to be extracted from a T4/T5 card in DDB after a panic.
2016-04-10 05:06:58 +00:00
#ifdef DDB
#include <ddb/ddb.h>
#include <ddb/db_lex.h>
#endif
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
#include "common/common.h"
T4 packet filtering/steering. - Enable 5-tuple and every-packet lookup. - Setup the default filter mode to allow filtering/steering based on IP protocol, ingress port, inner VLAN ID, IP frag, FCoE, and MPS match type; all combined together. You can also filter based on MAC index, Ethernet type, IP TOS/IPv6 Traffic Class, and outer VLAN ID but you'll have to modify the default filter mode and exclude some of the match-fields in it. IPv4 and IPv6 SIP/DIP/SPORT/DPORT are always available in all filter rules. - Add driver ioctls to get/set the global filter mode. - Add driver ioctls to program and delete hardware filters. A couple of the "switch" actions that rewrite Ethernet and VLAN information and switch the packet out of another port may not work as the L2 code is not yet in place. Everything else, including all "drop" and "pass" rules with RSS or absolute qid, should work. Obtained from: Chelsio Communications
2011-05-05 02:04:56 +00:00
#include "common/t4_msg.h"
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
#include "common/t4_regs.h"
#include "common/t4_regs_values.h"
cxgbe(4): Initial import of the "collect" component of Chelsio unified debug (cudbg) code, hooked up to the main driver via an ioctl. The ioctl can be used to collect the chip's internal state in a compressed dump file. These dumps can be decoded with the "view" component of cudbg. Obtained from: Chelsio Communications MFC after: 2 months Sponsored by: Chelsio Communications
2017-08-03 14:43:30 +00:00
#include "cudbg/cudbg.h"
Move CLIP table handling out of TOM and into the base driver. - Store the clip table in 'struct adapter' instead of in the TOM softc. - Init the clip table during attach and teardown during detach. - While here, add a dev.<nexus>.<unit>.misc.clip sysctl to dump the CLIP table. This does mean that we update the clip table even if TOE is not enabled, but non-TOE things need the CLIP table anyway. Reviewed by: np, Krishnamraju Eraparaju @ Chelsio Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18010
2018-11-29 01:15:53 +00:00
#include "t4_clip.h"
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
#include "t4_ioctl.h"
L2 table code. This is enough to get the T4's switch + L2 rewrite filters working. (All other filters - switch without L2 info rewrite, steer, and drop - were already fully-functional). Some contrived examples of "switch" filters with L2 rewriting: # cxgbetool t4nex0 iport 0 dport 80 action switch vlan +9 eport 3 Intercept all packets received on physical port 0 with TCP port 80 as destination, insert a vlan tag with VID 9, and send them out of port 3. # cxgbetool t4nex0 sip 192.168.1.1/32 ivlan 5 action switch \ vlan =9 smac aa:bb:cc:dd:ee:ff eport 0 Intercept all packets (received on any port) with source IP address 192.168.1.1 and VLAN id 5, rewrite the VLAN id to 9, rewrite source mac to aa:bb:cc:dd:ee:ff, and send it out of port 0. MFC after: 1 week
2011-05-30 21:07:26 +00:00
#include "t4_l2t.h"
cxgbe(4): major tx rework. a) Front load as much work as possible in if_transmit, before any driver lock or software queue has to get involved. b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This is specifically for the tx multiqueue model where one of the if_transmit producer threads becomes the consumer and other producers carry on as usual. mp_ring is implemented as standalone code and it should be possible to use it in any driver with tx multiqueue. It also has: - the ability to enqueue/dequeue multiple items. This might become significant if packet batching is ever implemented. - an abdication mechanism to allow a thread to give up writing tx descriptors and have another if_transmit thread take over. A thread that's writing tx descriptors can end up doing so for an unbounded time period if a) there are other if_transmit threads continuously feeding the sofware queue, and b) the chip keeps up with whatever the thread is throwing at it. - accurate statistics about interesting events even when the stats come at the expense of additional branches/conditional code. The NIC txq lock is uncontested on the fast path at this point. I've left it there for synchronization with the control events (interface up/down, modload/unload). c) Add support for "type 1" coalescing work request in the normal NIC tx path. This work request is optimized for frames with a single item in the DMA gather list. These are very common when forwarding packets. Note that netmap tx in cxgbe already uses these "type 1" work requests. d) Do not request automatic cidx updates every 32 descriptors. Instead, request updates via bits in individual work requests (still every 32 descriptors approximately). Also, request an automatic final update when the queue idles after activity. This means NIC tx reclaim is still performed lazily but it will catch up quickly as soon as the queue idles. This seems to be the best middle ground and I'll probably do something similar for netmap tx as well. e) Implement a faster tx path for WRQs (used by TOE tx and control queues, _not_ by the normal NIC tx). Allow work requests to be written directly to the hardware descriptor ring if room is available. I will convert t4_tom and iw_cxgbe modules to this faster style gradually. MFC after: 2 months
2014-12-31 23:19:16 +00:00
#include "t4_mp_ring.h"
Add a driver to create VF devices on Chelsio T4/T5 NICs. Chelsio NICs are a bit unique compared to some other NICs in that they expose different functionality on different physical functions. In particular, PF4 is used to manage the NIC interfaces ('t4nex' and 't5nex'). However, PF4 is not able to create VF devices. Instead, VFs are only supported by physical functions 0 through 3. This commit adds 't4iov' and 't5iov' drivers that attach to PF0-3. One extra wrinkle is that the iov devices cannot enable SR-IOV until the firwmare has been initialized by the main PF4 driver. To handle this case, a new t4_if kobj interface has been added to permit cross-calls between the PF drivers. The PF4 driver notifies sibling drivers when it is fully attached. It also requests sibling drivers to detach before it detaches. Sibling drivers query the PF4 driver during their attach routine to see if it is attached. If not, the sibling drivers defer their attach actions until the PF4 driver informs them it is attached. VF devices are associated with a single port on the NIC. VF devices created from PF0 are associated with the first port on the NIC, VFs from PF1 are associated with the second port, etc. VF devices can only be created from a PF device that has an associated port. Thus, on a 2-port card, VFs are only supported on PF0 and PF1. Reviewed by: np (earlier versions) MFC after: 1 month Sponsored by: Chelsio Communications
2016-07-22 22:46:41 +00:00
#include "t4_if.h"
cxgbe(4): Add code to deal with the chip's source MAC table (aka SMT). Submitted by: Krishnamraju Eraparaju @ Chelsio Sponsored by: Chelsio Communications
2018-05-31 21:31:08 +00:00
#include "t4_smt.h"
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
/* T4 bus driver interface */
static int t4_probe(device_t);
static int t4_attach(device_t);
static int t4_detach(device_t);
Add support for port unit wiring to cxgbe(4). - Add a bus_child_location_str method to the nexus drivers that prints out 'port=N' as the location string exported via devinfo and the '%location' sysctl node. - We can't use a bus_hint_device_unit to wire the unit numbers of devices with a fixed devclass as the device gets assigned a unit in make_device() before the device creator can set softc, etc. Instead, when adding a child device, use a helper function much like a bus_hint_device_unit method to look for wiring hints or to return -1 to let the system choose a unit number. This function requires an "at" hint for the port pointing to the nexus device and a "port" hint listing the port number. For example: hint.cxl.4.at="t5nex0" hint.cxl.4.port="0" wires cxl4 to the first port on the t5nex0 adapter. Requested by: gallatin MFC after: 2 months
2018-11-01 21:46:37 +00:00
static int t4_child_location_str(device_t, device_t, char *, size_t);
Add a driver to create VF devices on Chelsio T4/T5 NICs. Chelsio NICs are a bit unique compared to some other NICs in that they expose different functionality on different physical functions. In particular, PF4 is used to manage the NIC interfaces ('t4nex' and 't5nex'). However, PF4 is not able to create VF devices. Instead, VFs are only supported by physical functions 0 through 3. This commit adds 't4iov' and 't5iov' drivers that attach to PF0-3. One extra wrinkle is that the iov devices cannot enable SR-IOV until the firwmare has been initialized by the main PF4 driver. To handle this case, a new t4_if kobj interface has been added to permit cross-calls between the PF drivers. The PF4 driver notifies sibling drivers when it is fully attached. It also requests sibling drivers to detach before it detaches. Sibling drivers query the PF4 driver during their attach routine to see if it is attached. If not, the sibling drivers defer their attach actions until the PF4 driver informs them it is attached. VF devices are associated with a single port on the NIC. VF devices created from PF0 are associated with the first port on the NIC, VFs from PF1 are associated with the second port, etc. VF devices can only be created from a PF device that has an associated port. Thus, on a 2-port card, VFs are only supported on PF0 and PF1. Reviewed by: np (earlier versions) MFC after: 1 month Sponsored by: Chelsio Communications
2016-07-22 22:46:41 +00:00
static int t4_ready(device_t);
Use the port device name for the iov device for Chelsio T4/T5 cards. Chelsio T4/T5 adapters are multifunction cards. The main driver uses physical function 4 (PF4). However, VF devices for SR-IOV are only supported on physical functions 0 through 3, where PF0 creates VFs tied to port 0, etc. The t4iov/t5iov driver was previously added to create VF devices for ports that are present on each adapter. This change uses the recently added pci_iov_attach_name() function to name the character device in /dev/iov after the associated port on the card (e.g. /dev/iov/cxl0 is used to create VFs that share the cxl0 port). With this in place, mark the t4iov/t5iov devices quiet to prevent them from cluttering dmesg. Reviewed by: rstone Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7402
2016-08-03 17:11:08 +00:00
static int t4_read_port_device(device_t, int, device_t *);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
static device_method_t t4_methods[] = {
DEVMETHOD(device_probe, t4_probe),
DEVMETHOD(device_attach, t4_attach),
DEVMETHOD(device_detach, t4_detach),
Add support for port unit wiring to cxgbe(4). - Add a bus_child_location_str method to the nexus drivers that prints out 'port=N' as the location string exported via devinfo and the '%location' sysctl node. - We can't use a bus_hint_device_unit to wire the unit numbers of devices with a fixed devclass as the device gets assigned a unit in make_device() before the device creator can set softc, etc. Instead, when adding a child device, use a helper function much like a bus_hint_device_unit method to look for wiring hints or to return -1 to let the system choose a unit number. This function requires an "at" hint for the port pointing to the nexus device and a "port" hint listing the port number. For example: hint.cxl.4.at="t5nex0" hint.cxl.4.port="0" wires cxl4 to the first port on the t5nex0 adapter. Requested by: gallatin MFC after: 2 months
2018-11-01 21:46:37 +00:00
DEVMETHOD(bus_child_location_str, t4_child_location_str),
Add a driver to create VF devices on Chelsio T4/T5 NICs. Chelsio NICs are a bit unique compared to some other NICs in that they expose different functionality on different physical functions. In particular, PF4 is used to manage the NIC interfaces ('t4nex' and 't5nex'). However, PF4 is not able to create VF devices. Instead, VFs are only supported by physical functions 0 through 3. This commit adds 't4iov' and 't5iov' drivers that attach to PF0-3. One extra wrinkle is that the iov devices cannot enable SR-IOV until the firwmare has been initialized by the main PF4 driver. To handle this case, a new t4_if kobj interface has been added to permit cross-calls between the PF drivers. The PF4 driver notifies sibling drivers when it is fully attached. It also requests sibling drivers to detach before it detaches. Sibling drivers query the PF4 driver during their attach routine to see if it is attached. If not, the sibling drivers defer their attach actions until the PF4 driver informs them it is attached. VF devices are associated with a single port on the NIC. VF devices created from PF0 are associated with the first port on the NIC, VFs from PF1 are associated with the second port, etc. VF devices can only be created from a PF device that has an associated port. Thus, on a 2-port card, VFs are only supported on PF0 and PF1. Reviewed by: np (earlier versions) MFC after: 1 month Sponsored by: Chelsio Communications
2016-07-22 22:46:41 +00:00
DEVMETHOD(t4_is_main_ready, t4_ready),
Use the port device name for the iov device for Chelsio T4/T5 cards. Chelsio T4/T5 adapters are multifunction cards. The main driver uses physical function 4 (PF4). However, VF devices for SR-IOV are only supported on physical functions 0 through 3, where PF0 creates VFs tied to port 0, etc. The t4iov/t5iov driver was previously added to create VF devices for ports that are present on each adapter. This change uses the recently added pci_iov_attach_name() function to name the character device in /dev/iov after the associated port on the card (e.g. /dev/iov/cxl0 is used to create VFs that share the cxl0 port). With this in place, mark the t4iov/t5iov devices quiet to prevent them from cluttering dmesg. Reviewed by: rstone Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7402
2016-08-03 17:11:08 +00:00
DEVMETHOD(t4_read_port_device, t4_read_port_device),
Add a driver to create VF devices on Chelsio T4/T5 NICs. Chelsio NICs are a bit unique compared to some other NICs in that they expose different functionality on different physical functions. In particular, PF4 is used to manage the NIC interfaces ('t4nex' and 't5nex'). However, PF4 is not able to create VF devices. Instead, VFs are only supported by physical functions 0 through 3. This commit adds 't4iov' and 't5iov' drivers that attach to PF0-3. One extra wrinkle is that the iov devices cannot enable SR-IOV until the firwmare has been initialized by the main PF4 driver. To handle this case, a new t4_if kobj interface has been added to permit cross-calls between the PF drivers. The PF4 driver notifies sibling drivers when it is fully attached. It also requests sibling drivers to detach before it detaches. Sibling drivers query the PF4 driver during their attach routine to see if it is attached. If not, the sibling drivers defer their attach actions until the PF4 driver informs them it is attached. VF devices are associated with a single port on the NIC. VF devices created from PF0 are associated with the first port on the NIC, VFs from PF1 are associated with the second port, etc. VF devices can only be created from a PF device that has an associated port. Thus, on a 2-port card, VFs are only supported on PF0 and PF1. Reviewed by: np (earlier versions) MFC after: 1 month Sponsored by: Chelsio Communications
2016-07-22 22:46:41 +00:00
- There's no need to overwrite the default device method with the default one. Interestingly, these are actually the default for quite some time (bus_generic_driver_added(9) since r52045 and bus_generic_print_child(9) since r52045) but even recently added device drivers do this unnecessarily. Discussed with: jhb, marcel - While at it, use DEVMETHOD_END. Discussed with: jhb - Also while at it, use __FBSDID.
2011-11-22 21:28:20 +00:00
DEVMETHOD_END
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
};
static driver_t t4_driver = {
"t4nex",
t4_methods,
sizeof(struct adapter)
};
/* T4 port (cxgbe) interface */
static int cxgbe_probe(device_t);
static int cxgbe_attach(device_t);
static int cxgbe_detach(device_t);
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
device_method_t cxgbe_methods[] = {
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
DEVMETHOD(device_probe, cxgbe_probe),
DEVMETHOD(device_attach, cxgbe_attach),
DEVMETHOD(device_detach, cxgbe_detach),
{ 0, 0 }
};
static driver_t cxgbe_driver = {
"cxgbe",
cxgbe_methods,
sizeof(struct port_info)
};
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
/* T4 VI (vcxgbe) interface */
static int vcxgbe_probe(device_t);
static int vcxgbe_attach(device_t);
static int vcxgbe_detach(device_t);
static device_method_t vcxgbe_methods[] = {
DEVMETHOD(device_probe, vcxgbe_probe),
DEVMETHOD(device_attach, vcxgbe_attach),
DEVMETHOD(device_detach, vcxgbe_detach),
{ 0, 0 }
};
static driver_t vcxgbe_driver = {
"vcxgbe",
vcxgbe_methods,
sizeof(struct vi_info)
};
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
static d_ioctl_t t4_ioctl;
static struct cdevsw t4_cdevsw = {
.d_version = D_VERSION,
.d_ioctl = t4_ioctl,
.d_name = "t4nex",
};
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
/* T5 bus driver interface */
static int t5_probe(device_t);
static device_method_t t5_methods[] = {
DEVMETHOD(device_probe, t5_probe),
DEVMETHOD(device_attach, t4_attach),
DEVMETHOD(device_detach, t4_detach),
Add support for port unit wiring to cxgbe(4). - Add a bus_child_location_str method to the nexus drivers that prints out 'port=N' as the location string exported via devinfo and the '%location' sysctl node. - We can't use a bus_hint_device_unit to wire the unit numbers of devices with a fixed devclass as the device gets assigned a unit in make_device() before the device creator can set softc, etc. Instead, when adding a child device, use a helper function much like a bus_hint_device_unit method to look for wiring hints or to return -1 to let the system choose a unit number. This function requires an "at" hint for the port pointing to the nexus device and a "port" hint listing the port number. For example: hint.cxl.4.at="t5nex0" hint.cxl.4.port="0" wires cxl4 to the first port on the t5nex0 adapter. Requested by: gallatin MFC after: 2 months
2018-11-01 21:46:37 +00:00
DEVMETHOD(bus_child_location_str, t4_child_location_str),
Add a driver to create VF devices on Chelsio T4/T5 NICs. Chelsio NICs are a bit unique compared to some other NICs in that they expose different functionality on different physical functions. In particular, PF4 is used to manage the NIC interfaces ('t4nex' and 't5nex'). However, PF4 is not able to create VF devices. Instead, VFs are only supported by physical functions 0 through 3. This commit adds 't4iov' and 't5iov' drivers that attach to PF0-3. One extra wrinkle is that the iov devices cannot enable SR-IOV until the firwmare has been initialized by the main PF4 driver. To handle this case, a new t4_if kobj interface has been added to permit cross-calls between the PF drivers. The PF4 driver notifies sibling drivers when it is fully attached. It also requests sibling drivers to detach before it detaches. Sibling drivers query the PF4 driver during their attach routine to see if it is attached. If not, the sibling drivers defer their attach actions until the PF4 driver informs them it is attached. VF devices are associated with a single port on the NIC. VF devices created from PF0 are associated with the first port on the NIC, VFs from PF1 are associated with the second port, etc. VF devices can only be created from a PF device that has an associated port. Thus, on a 2-port card, VFs are only supported on PF0 and PF1. Reviewed by: np (earlier versions) MFC after: 1 month Sponsored by: Chelsio Communications
2016-07-22 22:46:41 +00:00
DEVMETHOD(t4_is_main_ready, t4_ready),
Use the port device name for the iov device for Chelsio T4/T5 cards. Chelsio T4/T5 adapters are multifunction cards. The main driver uses physical function 4 (PF4). However, VF devices for SR-IOV are only supported on physical functions 0 through 3, where PF0 creates VFs tied to port 0, etc. The t4iov/t5iov driver was previously added to create VF devices for ports that are present on each adapter. This change uses the recently added pci_iov_attach_name() function to name the character device in /dev/iov after the associated port on the card (e.g. /dev/iov/cxl0 is used to create VFs that share the cxl0 port). With this in place, mark the t4iov/t5iov devices quiet to prevent them from cluttering dmesg. Reviewed by: rstone Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7402
2016-08-03 17:11:08 +00:00
DEVMETHOD(t4_read_port_device, t4_read_port_device),
Add a driver to create VF devices on Chelsio T4/T5 NICs. Chelsio NICs are a bit unique compared to some other NICs in that they expose different functionality on different physical functions. In particular, PF4 is used to manage the NIC interfaces ('t4nex' and 't5nex'). However, PF4 is not able to create VF devices. Instead, VFs are only supported by physical functions 0 through 3. This commit adds 't4iov' and 't5iov' drivers that attach to PF0-3. One extra wrinkle is that the iov devices cannot enable SR-IOV until the firwmare has been initialized by the main PF4 driver. To handle this case, a new t4_if kobj interface has been added to permit cross-calls between the PF drivers. The PF4 driver notifies sibling drivers when it is fully attached. It also requests sibling drivers to detach before it detaches. Sibling drivers query the PF4 driver during their attach routine to see if it is attached. If not, the sibling drivers defer their attach actions until the PF4 driver informs them it is attached. VF devices are associated with a single port on the NIC. VF devices created from PF0 are associated with the first port on the NIC, VFs from PF1 are associated with the second port, etc. VF devices can only be created from a PF device that has an associated port. Thus, on a 2-port card, VFs are only supported on PF0 and PF1. Reviewed by: np (earlier versions) MFC after: 1 month Sponsored by: Chelsio Communications
2016-07-22 22:46:41 +00:00
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
DEVMETHOD_END
};
static driver_t t5_driver = {
"t5nex",
t5_methods,
sizeof(struct adapter)
};
/* T5 port (cxl) interface */
static driver_t cxl_driver = {
"cxl",
cxgbe_methods,
sizeof(struct port_info)
};
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
/* T5 VI (vcxl) interface */
static driver_t vcxl_driver = {
"vcxl",
vcxgbe_methods,
sizeof(struct vi_info)
};
cxgbe(4): Attach to cards with the Terminator 6 ASIC. T6 cards will come up as 't6nex' nexus devices with 'cc' ports hanging off them. The T6 firmware and configuration files will be added as soon as they are released. For now the driver will try to work with whatever firmware and configuration is on the card's flash. Sponsored by: Chelsio Communications
2016-09-16 00:08:37 +00:00
/* T6 bus driver interface */
static int t6_probe(device_t);
static device_method_t t6_methods[] = {
DEVMETHOD(device_probe, t6_probe),
DEVMETHOD(device_attach, t4_attach),
DEVMETHOD(device_detach, t4_detach),
Add support for port unit wiring to cxgbe(4). - Add a bus_child_location_str method to the nexus drivers that prints out 'port=N' as the location string exported via devinfo and the '%location' sysctl node. - We can't use a bus_hint_device_unit to wire the unit numbers of devices with a fixed devclass as the device gets assigned a unit in make_device() before the device creator can set softc, etc. Instead, when adding a child device, use a helper function much like a bus_hint_device_unit method to look for wiring hints or to return -1 to let the system choose a unit number. This function requires an "at" hint for the port pointing to the nexus device and a "port" hint listing the port number. For example: hint.cxl.4.at="t5nex0" hint.cxl.4.port="0" wires cxl4 to the first port on the t5nex0 adapter. Requested by: gallatin MFC after: 2 months
2018-11-01 21:46:37 +00:00
DEVMETHOD(bus_child_location_str, t4_child_location_str),
cxgbe(4): Attach to cards with the Terminator 6 ASIC. T6 cards will come up as 't6nex' nexus devices with 'cc' ports hanging off them. The T6 firmware and configuration files will be added as soon as they are released. For now the driver will try to work with whatever firmware and configuration is on the card's flash. Sponsored by: Chelsio Communications
2016-09-16 00:08:37 +00:00
DEVMETHOD(t4_is_main_ready, t4_ready),
DEVMETHOD(t4_read_port_device, t4_read_port_device),
DEVMETHOD_END
};
static driver_t t6_driver = {
"t6nex",
t6_methods,
sizeof(struct adapter)
};
/* T6 port (cc) interface */
static driver_t cc_driver = {
"cc",
cxgbe_methods,
sizeof(struct port_info)
};
/* T6 VI (vcc) interface */
static driver_t vcc_driver = {
"vcc",
vcxgbe_methods,
sizeof(struct vi_info)
};
cxgbev(4): Updates to the VF driver to cope with recent ifmedia and ctrlq changes in the base driver. MFC after: 1 week Sponsored by: Chelsio Communications
2018-08-23 00:58:10 +00:00
/* ifnet interface */
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
static void cxgbe_init(void *);
static int cxgbe_ioctl(struct ifnet *, unsigned long, caddr_t);
static int cxgbe_transmit(struct ifnet *, struct mbuf *);
static void cxgbe_qflush(struct ifnet *);
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
MALLOC_DEFINE(M_CXGBE, "cxgbe", "Chelsio T4/T5 Ethernet driver and services");
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
/*
* Correct lock order when you need to acquire multiple locks is t4_list_lock,
* then ADAPTER_LOCK, then t4_uld_list_lock.
*/
Change t4_list_lock and t4_uld_list_lock from mutexes to sx'es. - tom_uninit had to be reworked not to hold the adapter lock (a mutex) around t4_deactivate_uld, which acquires the uld_list_lock. - the ifc_match for the interface cloner that creates the tracer ifnet had to be reworked as the kernel calls ifc_match with the global if_cloners_mtx held.
2013-08-28 20:59:22 +00:00
static struct sx t4_list_lock;
Unstaticize t4_list and t4_uld_list. This works around a clang annoyance[1] and allows kgdb to find these symbols. [1] http://lists.freebsd.org/pipermail/freebsd-hackers/2012-November/041166.html MFC after: 3 days
2013-12-09 23:33:57 +00:00
SLIST_HEAD(, adapter) t4_list;
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
#ifdef TCP_OFFLOAD
Change t4_list_lock and t4_uld_list_lock from mutexes to sx'es. - tom_uninit had to be reworked not to hold the adapter lock (a mutex) around t4_deactivate_uld, which acquires the uld_list_lock. - the ifc_match for the interface cloner that creates the tracer ifnet had to be reworked as the kernel calls ifc_match with the global if_cloners_mtx held.
2013-08-28 20:59:22 +00:00
static struct sx t4_uld_list_lock;
Unstaticize t4_list and t4_uld_list. This works around a clang annoyance[1] and allows kgdb to find these symbols. [1] http://lists.freebsd.org/pipermail/freebsd-hackers/2012-November/041166.html MFC after: 3 days
2013-12-09 23:33:57 +00:00
SLIST_HEAD(, uld_info) t4_uld_list;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
#endif
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
/*
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
* Tunables. See tweak_tunables() too.
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
*
* Each tunable is set to a default value here if it's known at compile-time.
cxgbe(4): Allow tunables that control the number of queues to be set to '-n' to tell the driver to create _up to_ 'n' queues if enough cores are available. For example, setting hw.cxgbe.nrxq10g="-32" will result in 16 queues if the system has 16 cores, 32 if it has 32. There is no change in the default number of queues of any type. MFC after: 1 week Sponsored by: Chelsio Communications
2017-02-06 05:19:29 +00:00
* Otherwise it is set to -n as an indication to tweak_tunables() that it should
* provide a reasonable default (upto n) when the driver is loaded.
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
*
* Tunables applicable to both T4 and T5 are under hw.cxgbe. Those specific to
* T5 are under hw.cxl.
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
*/
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_NODE(_hw, OID_AUTO, cxgbe, CTLFLAG_RD, 0, "cxgbe(4) parameters");
SYSCTL_NODE(_hw, OID_AUTO, cxl, CTLFLAG_RD, 0, "cxgbe(4) T5+ parameters");
SYSCTL_NODE(_hw_cxgbe, OID_AUTO, toe, CTLFLAG_RD, 0, "cxgbe(4) TOE parameters");
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
/*
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
* Number of queues for tx and rx, NIC and offload.
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
*/
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
#define NTXQ 16
int t4_ntxq = -NTXQ;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, ntxq, CTLFLAG_RDTUN, &t4_ntxq, 0,
"Number of TX queues per port");
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
TUNABLE_INT("hw.cxgbe.ntxq10g", &t4_ntxq); /* Old name, undocumented */
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
#define NRXQ 8
int t4_nrxq = -NRXQ;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, nrxq, CTLFLAG_RDTUN, &t4_nrxq, 0,
"Number of RX queues per port");
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
TUNABLE_INT("hw.cxgbe.nrxq10g", &t4_nrxq); /* Old name, undocumented */
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
#define NTXQ_VI 1
cxgbe(4): Allow tunables that control the number of queues to be set to '-n' to tell the driver to create _up to_ 'n' queues if enough cores are available. For example, setting hw.cxgbe.nrxq10g="-32" will result in 16 queues if the system has 16 cores, 32 if it has 32. There is no change in the default number of queues of any type. MFC after: 1 week Sponsored by: Chelsio Communications
2017-02-06 05:19:29 +00:00
static int t4_ntxq_vi = -NTXQ_VI;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, ntxq_vi, CTLFLAG_RDTUN, &t4_ntxq_vi, 0,
"Number of TX queues per VI");
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
#define NRXQ_VI 1
cxgbe(4): Allow tunables that control the number of queues to be set to '-n' to tell the driver to create _up to_ 'n' queues if enough cores are available. For example, setting hw.cxgbe.nrxq10g="-32" will result in 16 queues if the system has 16 cores, 32 if it has 32. There is no change in the default number of queues of any type. MFC after: 1 week Sponsored by: Chelsio Communications
2017-02-06 05:19:29 +00:00
static int t4_nrxq_vi = -NRXQ_VI;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, nrxq_vi, CTLFLAG_RDTUN, &t4_nrxq_vi, 0,
"Number of RX queues per VI");
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
Add a new sysctl, dev.cxgbe.N.rsrv_noflow, and a companion tunable, hw.cxgbe.rsrv_noflow. When set, queue 0 of the port is reserved for TX packets without a flowid. The hash value of packets with a flowid is bumped up by 1. The intent is to provide a private queue for link-level packets like LACP that is unlikely to overflow or suffer deep queue latency. Reviewed by: np Obtained from: Netflix MFC after: 3 days
2014-02-06 18:40:38 +00:00
static int t4_rsrv_noflowq = 0;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, rsrv_noflowq, CTLFLAG_RDTUN, &t4_rsrv_noflowq,
0, "Reserve TX queue 0 of each VI for non-flowid packets");
Add a new sysctl, dev.cxgbe.N.rsrv_noflow, and a companion tunable, hw.cxgbe.rsrv_noflow. When set, queue 0 of the port is reserved for TX packets without a flowid. The hash value of packets with a flowid is bumped up by 1. The intent is to provide a private queue for link-level packets like LACP that is unlikely to overflow or suffer deep queue latency. Reviewed by: np Obtained from: Netflix MFC after: 3 days
2014-02-06 18:40:38 +00:00
cxgbe(4): Allocate offload Tx queues when a card has resources provisioned for NIC_ETHOFLD and the kernel has option RATELIMIT. It is possible to use the chip's offload queues for normal NIC Tx and not just TOE Tx. The difference is that these queues support out of order processing of work requests and have a per-"flowid" mechanism for tracking credits between the driver and hardware. This allows Tx for any number of flows bound to different rate limits to be submitted to a single Tx queue and the work requests for slow flows won't cause HOL blocking for the rest. Sponsored by: Chelsio Communications
2018-05-17 01:42:18 +00:00
#if defined(TCP_OFFLOAD) || defined(RATELIMIT)
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
#define NOFLDTXQ 8
static int t4_nofldtxq = -NOFLDTXQ;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, nofldtxq, CTLFLAG_RDTUN, &t4_nofldtxq, 0,
"Number of offload TX queues per port");
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
#define NOFLDRXQ 2
static int t4_nofldrxq = -NOFLDRXQ;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, nofldrxq, CTLFLAG_RDTUN, &t4_nofldrxq, 0,
"Number of offload RX queues per port");
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
#define NOFLDTXQ_VI 1
cxgbe(4): Allow tunables that control the number of queues to be set to '-n' to tell the driver to create _up to_ 'n' queues if enough cores are available. For example, setting hw.cxgbe.nrxq10g="-32" will result in 16 queues if the system has 16 cores, 32 if it has 32. There is no change in the default number of queues of any type. MFC after: 1 week Sponsored by: Chelsio Communications
2017-02-06 05:19:29 +00:00
static int t4_nofldtxq_vi = -NOFLDTXQ_VI;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, nofldtxq_vi, CTLFLAG_RDTUN, &t4_nofldtxq_vi, 0,
"Number of offload TX queues per VI");
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
#define NOFLDRXQ_VI 1
cxgbe(4): Allow tunables that control the number of queues to be set to '-n' to tell the driver to create _up to_ 'n' queues if enough cores are available. For example, setting hw.cxgbe.nrxq10g="-32" will result in 16 queues if the system has 16 cores, 32 if it has 32. There is no change in the default number of queues of any type. MFC after: 1 week Sponsored by: Chelsio Communications
2017-02-06 05:19:29 +00:00
static int t4_nofldrxq_vi = -NOFLDRXQ_VI;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, nofldrxq_vi, CTLFLAG_RDTUN, &t4_nofldrxq_vi, 0,
"Number of offload RX queues per VI");
cxgbe(4): Allow the TOE timer tunables to be set with microsecond precision. These timers are already displayed in microseconds in the sysctl MIB. Add variables to track these tunables while here. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 15:57:10 +00:00
cxgbe(4): Provide knobs to set the holdoff parameters of TOE rx queues separately from NIC rx queues instead of using the same parameters for both types of queues. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-10-05 07:18:16 +00:00
#define TMR_IDX_OFLD 1
int t4_tmr_idx_ofld = TMR_IDX_OFLD;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, holdoff_timer_idx_ofld, CTLFLAG_RDTUN,
&t4_tmr_idx_ofld, 0, "Holdoff timer index for offload queues");
cxgbe(4): Provide knobs to set the holdoff parameters of TOE rx queues separately from NIC rx queues instead of using the same parameters for both types of queues. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-10-05 07:18:16 +00:00
#define PKTC_IDX_OFLD (-1)
int t4_pktc_idx_ofld = PKTC_IDX_OFLD;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, holdoff_pktc_idx_ofld, CTLFLAG_RDTUN,
&t4_pktc_idx_ofld, 0, "holdoff packet counter index for offload queues");
cxgbe(4): Provide knobs to set the holdoff parameters of TOE rx queues separately from NIC rx queues instead of using the same parameters for both types of queues. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-10-05 07:18:16 +00:00
cxgbe(4): Allow the TOE timer tunables to be set with microsecond precision. These timers are already displayed in microseconds in the sysctl MIB. Add variables to track these tunables while here. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 15:57:10 +00:00
/* 0 means chip/fw default, non-zero number is value in microseconds */
static u_long t4_toe_keepalive_idle = 0;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_ULONG(_hw_cxgbe_toe, OID_AUTO, keepalive_idle, CTLFLAG_RDTUN,
&t4_toe_keepalive_idle, 0, "TOE keepalive idle timer (us)");
cxgbe(4): Allow the TOE timer tunables to be set with microsecond precision. These timers are already displayed in microseconds in the sysctl MIB. Add variables to track these tunables while here. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 15:57:10 +00:00
/* 0 means chip/fw default, non-zero number is value in microseconds */
static u_long t4_toe_keepalive_interval = 0;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_ULONG(_hw_cxgbe_toe, OID_AUTO, keepalive_interval, CTLFLAG_RDTUN,
&t4_toe_keepalive_interval, 0, "TOE keepalive interval timer (us)");
cxgbe(4): Allow the TOE timer tunables to be set with microsecond precision. These timers are already displayed in microseconds in the sysctl MIB. Add variables to track these tunables while here. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 15:57:10 +00:00
/* 0 means chip/fw default, non-zero number is # of keepalives before abort */
static int t4_toe_keepalive_count = 0;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe_toe, OID_AUTO, keepalive_count, CTLFLAG_RDTUN,
&t4_toe_keepalive_count, 0, "Number of TOE keepalive probes before abort");
cxgbe(4): Allow the TOE timer tunables to be set with microsecond precision. These timers are already displayed in microseconds in the sysctl MIB. Add variables to track these tunables while here. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 15:57:10 +00:00
/* 0 means chip/fw default, non-zero number is value in microseconds */
static u_long t4_toe_rexmt_min = 0;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_ULONG(_hw_cxgbe_toe, OID_AUTO, rexmt_min, CTLFLAG_RDTUN,
&t4_toe_rexmt_min, 0, "Minimum TOE retransmit interval (us)");
cxgbe(4): Allow the TOE timer tunables to be set with microsecond precision. These timers are already displayed in microseconds in the sysctl MIB. Add variables to track these tunables while here. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 15:57:10 +00:00
/* 0 means chip/fw default, non-zero number is value in microseconds */
static u_long t4_toe_rexmt_max = 0;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_ULONG(_hw_cxgbe_toe, OID_AUTO, rexmt_max, CTLFLAG_RDTUN,
&t4_toe_rexmt_max, 0, "Maximum TOE retransmit interval (us)");
cxgbe(4): Allow the TOE timer tunables to be set with microsecond precision. These timers are already displayed in microseconds in the sysctl MIB. Add variables to track these tunables while here. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 15:57:10 +00:00
/* 0 means chip/fw default, non-zero number is # of rexmt before abort */
static int t4_toe_rexmt_count = 0;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe_toe, OID_AUTO, rexmt_count, CTLFLAG_RDTUN,
&t4_toe_rexmt_count, 0, "Number of TOE retransmissions before abort");
cxgbe(4): Allow the TOE timer tunables to be set with microsecond precision. These timers are already displayed in microseconds in the sysctl MIB. Add variables to track these tunables while here. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 15:57:10 +00:00
/* -1 means chip/fw default, other values are raw backoff values to use */
static int t4_toe_rexmt_backoff[16] = {
-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1
};
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_NODE(_hw_cxgbe_toe, OID_AUTO, rexmt_backoff, CTLFLAG_RD, 0,
"cxgbe(4) TOE retransmit backoff values");
SYSCTL_INT(_hw_cxgbe_toe_rexmt_backoff, OID_AUTO, 0, CTLFLAG_RDTUN,
&t4_toe_rexmt_backoff[0], 0, "");
SYSCTL_INT(_hw_cxgbe_toe_rexmt_backoff, OID_AUTO, 1, CTLFLAG_RDTUN,
&t4_toe_rexmt_backoff[1], 0, "");
SYSCTL_INT(_hw_cxgbe_toe_rexmt_backoff, OID_AUTO, 2, CTLFLAG_RDTUN,
&t4_toe_rexmt_backoff[2], 0, "");
SYSCTL_INT(_hw_cxgbe_toe_rexmt_backoff, OID_AUTO, 3, CTLFLAG_RDTUN,
&t4_toe_rexmt_backoff[3], 0, "");
SYSCTL_INT(_hw_cxgbe_toe_rexmt_backoff, OID_AUTO, 4, CTLFLAG_RDTUN,
&t4_toe_rexmt_backoff[4], 0, "");
SYSCTL_INT(_hw_cxgbe_toe_rexmt_backoff, OID_AUTO, 5, CTLFLAG_RDTUN,
&t4_toe_rexmt_backoff[5], 0, "");
SYSCTL_INT(_hw_cxgbe_toe_rexmt_backoff, OID_AUTO, 6, CTLFLAG_RDTUN,
&t4_toe_rexmt_backoff[6], 0, "");
SYSCTL_INT(_hw_cxgbe_toe_rexmt_backoff, OID_AUTO, 7, CTLFLAG_RDTUN,
&t4_toe_rexmt_backoff[7], 0, "");
SYSCTL_INT(_hw_cxgbe_toe_rexmt_backoff, OID_AUTO, 8, CTLFLAG_RDTUN,
&t4_toe_rexmt_backoff[8], 0, "");
SYSCTL_INT(_hw_cxgbe_toe_rexmt_backoff, OID_AUTO, 9, CTLFLAG_RDTUN,
&t4_toe_rexmt_backoff[9], 0, "");
SYSCTL_INT(_hw_cxgbe_toe_rexmt_backoff, OID_AUTO, 10, CTLFLAG_RDTUN,
&t4_toe_rexmt_backoff[10], 0, "");
SYSCTL_INT(_hw_cxgbe_toe_rexmt_backoff, OID_AUTO, 11, CTLFLAG_RDTUN,
&t4_toe_rexmt_backoff[11], 0, "");
SYSCTL_INT(_hw_cxgbe_toe_rexmt_backoff, OID_AUTO, 12, CTLFLAG_RDTUN,
&t4_toe_rexmt_backoff[12], 0, "");
SYSCTL_INT(_hw_cxgbe_toe_rexmt_backoff, OID_AUTO, 13, CTLFLAG_RDTUN,
&t4_toe_rexmt_backoff[13], 0, "");
SYSCTL_INT(_hw_cxgbe_toe_rexmt_backoff, OID_AUTO, 14, CTLFLAG_RDTUN,
&t4_toe_rexmt_backoff[14], 0, "");
SYSCTL_INT(_hw_cxgbe_toe_rexmt_backoff, OID_AUTO, 15, CTLFLAG_RDTUN,
&t4_toe_rexmt_backoff[15], 0, "");
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
#endif
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
#ifdef DEV_NETMAP
#define NNMTXQ_VI 2
cxgbe(4): Allow tunables that control the number of queues to be set to '-n' to tell the driver to create _up to_ 'n' queues if enough cores are available. For example, setting hw.cxgbe.nrxq10g="-32" will result in 16 queues if the system has 16 cores, 32 if it has 32. There is no change in the default number of queues of any type. MFC after: 1 week Sponsored by: Chelsio Communications
2017-02-06 05:19:29 +00:00
static int t4_nnmtxq_vi = -NNMTXQ_VI;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, nnmtxq_vi, CTLFLAG_RDTUN, &t4_nnmtxq_vi, 0,
"Number of netmap TX queues per VI");
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
#define NNMRXQ_VI 2
cxgbe(4): Allow tunables that control the number of queues to be set to '-n' to tell the driver to create _up to_ 'n' queues if enough cores are available. For example, setting hw.cxgbe.nrxq10g="-32" will result in 16 queues if the system has 16 cores, 32 if it has 32. There is no change in the default number of queues of any type. MFC after: 1 week Sponsored by: Chelsio Communications
2017-02-06 05:19:29 +00:00
static int t4_nnmrxq_vi = -NNMRXQ_VI;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, nnmrxq_vi, CTLFLAG_RDTUN, &t4_nnmrxq_vi, 0,
"Number of netmap RX queues per VI");
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
#endif
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
/*
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
* Holdoff parameters for ports.
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
*/
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
#define TMR_IDX 1
int t4_tmr_idx = TMR_IDX;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, holdoff_timer_idx, CTLFLAG_RDTUN, &t4_tmr_idx,
0, "Holdoff timer index");
cxgbe(4): Accept old names of a couple of tunables.
2018-01-26 00:45:40 +00:00
TUNABLE_INT("hw.cxgbe.holdoff_timer_idx_10G", &t4_tmr_idx); /* Old name */
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
#define PKTC_IDX (-1)
int t4_pktc_idx = PKTC_IDX;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, holdoff_pktc_idx, CTLFLAG_RDTUN, &t4_pktc_idx,
0, "Holdoff packet counter index");
cxgbe(4): Accept old names of a couple of tunables.
2018-01-26 00:45:40 +00:00
TUNABLE_INT("hw.cxgbe.holdoff_pktc_idx_10G", &t4_pktc_idx); /* Old name */
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
/*
* Size (# of entries) of each tx and rx queue.
*/
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
unsigned int t4_qsize_txq = TX_EQ_QSIZE;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, qsize_txq, CTLFLAG_RDTUN, &t4_qsize_txq, 0,
"Number of descriptors in each TX queue");
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
unsigned int t4_qsize_rxq = RX_IQ_QSIZE;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, qsize_rxq, CTLFLAG_RDTUN, &t4_qsize_rxq, 0,
"Number of descriptors in each RX queue");
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
/*
* Interrupt types allowed (bits 0, 1, 2 = INTx, MSI, MSI-X respectively).
*/
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
int t4_intr_types = INTR_MSIX | INTR_MSI | INTR_INTX;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, interrupt_types, CTLFLAG_RDTUN, &t4_intr_types,
0, "Interrupt types allowed (bit 0 = INTx, 1 = MSI, 2 = MSI-X)");
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
/*
cxgbe(4): Fall back to a failsafe configuration built into the firmware if an error is reported while pre-processing the configuration file that the driver attempted to use. Also, allow the user to explicitly use the built-in configuration with hw.cxgbe.config_file="built-in" MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-16 17:55:16 +00:00
* Configuration file. All the _CF names here are special.
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
*/
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
#define DEFAULT_CF "default"
cxgbe(4): Fall back to a failsafe configuration built into the firmware if an error is reported while pre-processing the configuration file that the driver attempted to use. Also, allow the user to explicitly use the built-in configuration with hw.cxgbe.config_file="built-in" MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-16 17:55:16 +00:00
#define BUILTIN_CF "built-in"
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
#define FLASH_CF "flash"
#define UWIRE_CF "uwire"
- Explain clearly why a different firmware is being installed (if/when it is being installed). Improve other error messages while here. - Select special FPGA specific configuration profile when appropriate. MFC after: 3 days
2013-04-11 19:39:40 +00:00
#define FPGA_CF "fpga"
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
static char t4_cfg_file[32] = DEFAULT_CF;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_STRING(_hw_cxgbe, OID_AUTO, config_file, CTLFLAG_RDTUN, t4_cfg_file,
sizeof(t4_cfg_file), "Firmware configuration file");
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): knobs to enable/disable PAUSE frame based flow control. MFC after: 1 week
2014-09-12 05:25:56 +00:00
/*
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
* PAUSE settings (bit 0, 1, 2 = rx_pause, tx_pause, pause_autoneg respectively).
cxgbe(4): knobs to enable/disable PAUSE frame based flow control. MFC after: 1 week
2014-09-12 05:25:56 +00:00
* rx_pause = 1 to heed incoming PAUSE frames, 0 to ignore them.
* tx_pause = 1 to emit PAUSE frames when the rx FIFO reaches its high water
* mark or when signalled to do so, 0 to never emit PAUSE.
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
* pause_autoneg = 1 means PAUSE will be negotiated if possible and the
* negotiated settings will override rx_pause/tx_pause.
* Otherwise rx_pause/tx_pause are applied forcibly.
cxgbe(4): knobs to enable/disable PAUSE frame based flow control. MFC after: 1 week
2014-09-12 05:25:56 +00:00
*/
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
static int t4_pause_settings = PAUSE_RX | PAUSE_TX | PAUSE_AUTONEG;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, pause_settings, CTLFLAG_RDTUN,
&t4_pause_settings, 0,
"PAUSE settings (bit 0 = rx_pause, 1 = tx_pause, 2 = pause_autoneg)");
cxgbe(4): knobs to enable/disable PAUSE frame based flow control. MFC after: 1 week
2014-09-12 05:25:56 +00:00
cxgbe(4): Updates to link configuration. - Update struct link_settings and associated shared code. - Add tunables to control FEC and autonegotiation. All ports inherit these values as their initial settings. hw.cxgbe.fec hw.cxgbe.autoneg - Add per-port sysctls to control FEC and autonegotiation. These can be modified at any time. dev.<port>.<n>.fec dev.<port>.<n>.autoneg MFC after: 3 days Sponsored by: Chelsio Communications
2016-12-30 08:59:49 +00:00
/*
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
* Forward Error Correction settings (bit 0, 1 = RS, BASER respectively).
* -1 to run with the firmware default. Same as FEC_AUTO (bit 5)
cxgbe(4): Updates to link configuration. - Update struct link_settings and associated shared code. - Add tunables to control FEC and autonegotiation. All ports inherit these values as their initial settings. hw.cxgbe.fec hw.cxgbe.autoneg - Add per-port sysctls to control FEC and autonegotiation. These can be modified at any time. dev.<port>.<n>.fec dev.<port>.<n>.autoneg MFC after: 3 days Sponsored by: Chelsio Communications
2016-12-30 08:59:49 +00:00
* 0 to disable FEC.
*/
static int t4_fec = -1;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, fec, CTLFLAG_RDTUN, &t4_fec, 0,
"Forward Error Correction (bit 0 = RS, bit 1 = BASER_RS)");
cxgbe(4): Updates to link configuration. - Update struct link_settings and associated shared code. - Add tunables to control FEC and autonegotiation. All ports inherit these values as their initial settings. hw.cxgbe.fec hw.cxgbe.autoneg - Add per-port sysctls to control FEC and autonegotiation. These can be modified at any time. dev.<port>.<n>.fec dev.<port>.<n>.autoneg MFC after: 3 days Sponsored by: Chelsio Communications
2016-12-30 08:59:49 +00:00
/*
* Link autonegotiation.
* -1 to run with the firmware default.
* 0 to disable.
* 1 to enable.
*/
static int t4_autoneg = -1;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, autoneg, CTLFLAG_RDTUN, &t4_autoneg, 0,
"Link autonegotiation");
cxgbe(4): Updates to link configuration. - Update struct link_settings and associated shared code. - Add tunables to control FEC and autonegotiation. All ports inherit these values as their initial settings. hw.cxgbe.fec hw.cxgbe.autoneg - Add per-port sysctls to control FEC and autonegotiation. These can be modified at any time. dev.<port>.<n>.fec dev.<port>.<n>.autoneg MFC after: 3 days Sponsored by: Chelsio Communications
2016-12-30 08:59:49 +00:00
cxgbe(4): Consider all the API versions of the interfaces exported by the firmware (instead of just the main firmware version) when evaluating firmware compatibility. Document the new "hw.cxgbe.fw_install" knob being introduced here. This should fix kern/173584 too. Setting hw.cxgbe.fw_install=2 will mostly do what was requested in the PR but it's a bit more intelligent in that it won't reinstall the same firmware repeatedly if the knob is left set. PR: kern/173584 MFC after: 5 days
2013-02-26 20:35:54 +00:00
/*
* Firmware auto-install by driver during attach (0, 1, 2 = prohibited, allowed,
cxgbe(4): Allow negative values in hw.cxgbe.fw_install and take them to mean that the driver should taste the firmware in the KLD and use that firmware's version for all its fw_install checks. The driver gets firmware version information from compiled-in values by default and this change allows custom (or older/newer) firmware modules to be used with the stock driver. There is no change in default behavior. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 18:42:16 +00:00
* encouraged respectively). '-n' is the same as 'n' except the firmware
* version used in the checks is read from the firmware bundled with the driver.
cxgbe(4): Consider all the API versions of the interfaces exported by the firmware (instead of just the main firmware version) when evaluating firmware compatibility. Document the new "hw.cxgbe.fw_install" knob being introduced here. This should fix kern/173584 too. Setting hw.cxgbe.fw_install=2 will mostly do what was requested in the PR but it's a bit more intelligent in that it won't reinstall the same firmware repeatedly if the knob is left set. PR: kern/173584 MFC after: 5 days
2013-02-26 20:35:54 +00:00
*/
cxgbe(4): Allow negative values in hw.cxgbe.fw_install and take them to mean that the driver should taste the firmware in the KLD and use that firmware's version for all its fw_install checks. The driver gets firmware version information from compiled-in values by default and this change allows custom (or older/newer) firmware modules to be used with the stock driver. There is no change in default behavior. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 18:42:16 +00:00
static int t4_fw_install = 1;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, fw_install, CTLFLAG_RDTUN, &t4_fw_install, 0,
"Firmware auto-install (0 = prohibited, 1 = allowed, 2 = encouraged)");
cxgbe(4): Consider all the API versions of the interfaces exported by the firmware (instead of just the main firmware version) when evaluating firmware compatibility. Document the new "hw.cxgbe.fw_install" knob being introduced here. This should fix kern/173584 too. Setting hw.cxgbe.fw_install=2 will mostly do what was requested in the PR but it's a bit more intelligent in that it won't reinstall the same firmware repeatedly if the knob is left set. PR: kern/173584 MFC after: 5 days
2013-02-26 20:35:54 +00:00
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
/*
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
* ASIC features that will be used. Disable the ones you don't want so that the
* chip resources aren't wasted on features that will not be used.
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
*/
cxgbe(4): Catch up with the latest list of card capabilities as reported by the firmware.
2016-03-12 02:54:55 +00:00
static int t4_nbmcaps_allowed = 0;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, nbmcaps_allowed, CTLFLAG_RDTUN,
&t4_nbmcaps_allowed, 0, "Default NBM capabilities");
cxgbe(4): Catch up with the latest list of card capabilities as reported by the firmware.
2016-03-12 02:54:55 +00:00
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
static int t4_linkcaps_allowed = 0; /* No DCBX, PPP, etc. by default */
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, linkcaps_allowed, CTLFLAG_RDTUN,
&t4_linkcaps_allowed, 0, "Default link capabilities");
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): Enable additional capabilities in the default configuration files. All features with FreeBSD drivers of some kind are now in the default configuration.
2016-03-16 19:43:44 +00:00
static int t4_switchcaps_allowed = FW_CAPS_CONFIG_SWITCH_INGRESS |
FW_CAPS_CONFIG_SWITCH_EGRESS;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, switchcaps_allowed, CTLFLAG_RDTUN,
&t4_switchcaps_allowed, 0, "Default switch capabilities");
cxgbe(4): Catch up with the latest list of card capabilities as reported by the firmware.
2016-03-12 02:54:55 +00:00
cxgbe(4): Enable support for per-connection rate limiting in the default firmware configuration files. Approved by: re@ (gjb@) Sponsored by: Chelsio Communications
2018-09-26 21:16:07 +00:00
#ifdef RATELIMIT
cxgbe(4): Add support for hash filters. These filters reside in the card's memory instead of its TCAM and can be configured via a new "hashfilter" subcommand in cxgbetool. Hash and normal TCAM filters can be used together. The hardware does an exact-match of packet fields for hash filters, unlike the masked match performed for TCAM filters. Any T5/T6 card with memory can support at least half a million hash filters. The sample config file with the driver configures 512K of these, it is possible to double this to 1 million+ in some cases. The chip does an exact-match of fields of incoming datagrams with hash filters and performs the action configured for the filter if it matches. The fields to match are specified in a "filter mask" in the firmware config file. The filter mask always includes the 5-tuple (sip, dip, sport, dport, ipproto). It can, optionally, also include any subset of the filter mode (see filterMode and filterMask in the firmware config file). For example: filterMode = fragmentation, mpshittype, protocol, vlan, port, fcoe filterMask = protocol, port, vlan Exact values of the 5-tuple, the physical port, and VLAN tag would have to be provided while setting up a hash filter with the chip configuration above. Hash filters support all actions supported by TCAM filters. A packet that hits a hash filter can be dropped, let through (with optional steering to a specific queue or RSS region), switched out of another port (with optional L2 rewrite of DMAC, SMAC, VLAN tag), or get NAT'ed. (Support for some of these will show up in the driver in a follow-up commit very shortly). Sponsored by: Chelsio Communications
2018-05-09 04:09:49 +00:00
static int t4_niccaps_allowed = FW_CAPS_CONFIG_NIC |
cxgbe(4): Add NIC_ETHOFLD to the NIC capabilities allowed by the driver by default. This is the first of a series of commits that will add support for RATELIMIT kernel option to the base if_cxgbe driver, for use with ordinary NIC traffic "flows". RATELIMIT is already supported by t4_tom for the fully-offloaded TCP connections that it handles. Sponsored by: Chelsio Communications
2018-05-17 00:52:48 +00:00
FW_CAPS_CONFIG_NIC_HASHFILTER | FW_CAPS_CONFIG_NIC_ETHOFLD;
cxgbe(4): Enable support for per-connection rate limiting in the default firmware configuration files. Approved by: re@ (gjb@) Sponsored by: Chelsio Communications
2018-09-26 21:16:07 +00:00
#else
static int t4_niccaps_allowed = FW_CAPS_CONFIG_NIC |
FW_CAPS_CONFIG_NIC_HASHFILTER;
#endif
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, niccaps_allowed, CTLFLAG_RDTUN,
&t4_niccaps_allowed, 0, "Default NIC capabilities");
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
Instruct the firmware not to provision resources for TCP offload if the kernel is being built without TCP_OFFLOAD. But never override toecaps_allowed if it has been set manually.
2012-07-02 20:42:43 +00:00
static int t4_toecaps_allowed = -1;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, toecaps_allowed, CTLFLAG_RDTUN,
&t4_toecaps_allowed, 0, "Default TCP offload capabilities");
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): Enable additional capabilities in the default configuration files. All features with FreeBSD drivers of some kind are now in the default configuration.
2016-03-16 19:43:44 +00:00
static int t4_rdmacaps_allowed = -1;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, rdmacaps_allowed, CTLFLAG_RDTUN,
&t4_rdmacaps_allowed, 0, "Default RDMA capabilities");
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
Enable support for lookaside crypto operations by default. This permits ccr(4) to be used with the default firmware configuration file. Discussed with: np Sponsored by: Chelsio Communications
2017-09-18 23:50:34 +00:00
static int t4_cryptocaps_allowed = -1;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, cryptocaps_allowed, CTLFLAG_RDTUN,
&t4_cryptocaps_allowed, 0, "Default crypto capabilities");
cxgbe(4): Catch up with the latest list of card capabilities as reported by the firmware.
2016-03-12 02:54:55 +00:00
cxgbe(4): Enable additional capabilities in the default configuration files. All features with FreeBSD drivers of some kind are now in the default configuration.
2016-03-16 19:43:44 +00:00
static int t4_iscsicaps_allowed = -1;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, iscsicaps_allowed, CTLFLAG_RDTUN,
&t4_iscsicaps_allowed, 0, "Default iSCSI capabilities");
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
static int t4_fcoecaps_allowed = 0;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, fcoecaps_allowed, CTLFLAG_RDTUN,
&t4_fcoecaps_allowed, 0, "Default FCoE capabilities");
T4 packet filtering/steering. - Enable 5-tuple and every-packet lookup. - Setup the default filter mode to allow filtering/steering based on IP protocol, ingress port, inner VLAN ID, IP frag, FCoE, and MPS match type; all combined together. You can also filter based on MAC index, Ethernet type, IP TOS/IPv6 Traffic Class, and outer VLAN ID but you'll have to modify the default filter mode and exclude some of the match-fields in it. IPv4 and IPv6 SIP/DIP/SPORT/DPORT are always available in all filter rules. - Add driver ioctls to get/set the global filter mode. - Add driver ioctls to program and delete hardware filters. A couple of the "switch" actions that rewrite Ethernet and VLAN information and switch the packet out of another port may not work as the L2 code is not yet in place. Everything else, including all "drop" and "pass" rules with RSS or absolute qid, should work. Obtained from: Chelsio Communications
2011-05-05 02:04:56 +00:00
cxgbe(4): Disable write-combined doorbells by default. This had been the default behavior but was changed accidentally as part of the recent iw_cxgbe+OFED overhaul. Fix another bug in that change while here: the global knob affects all the adapters in the system and should be left alone by per-adapter code. MFC after: 3 days Sponsored by: Chelsio Communications
2018-05-10 06:33:54 +00:00
static int t5_write_combine = 0;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxl, OID_AUTO, write_combine, CTLFLAG_RDTUN, &t5_write_combine,
0, "Use WC instead of UC for BAR2");
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
static int t4_num_vis = 1;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, num_vis, CTLFLAG_RDTUN, &t4_num_vis, 0,
"Number of VIs per port");
cxgbe(4): Add a knob to enable/disable PCIe relaxed ordering. Disable it by default when running on Intel CPUs. This is a crude fix for the performance issues alluded to in these Linux commits: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=87e09cdec4dae08acdb4aa49beb793c19d73e73e https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=a99b646afa8a02571ea298bedca6592d818229cd MFC after: 1 week Sponsored by: Chelsio Communications
2018-01-03 19:24:57 +00:00
/*
* PCIe Relaxed Ordering.
* -1: driver should figure out a good value.
* 0: disable RO.
* 1: enable RO.
* 2: leave RO alone.
*/
static int pcie_relaxed_ordering = -1;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, pcie_relaxed_ordering, CTLFLAG_RDTUN,
&pcie_relaxed_ordering, 0,
"PCIe Relaxed Ordering: 0 = disable, 1 = enable, 2 = leave alone");
cxgbe(4): Add a knob to enable/disable PCIe relaxed ordering. Disable it by default when running on Intel CPUs. This is a crude fix for the performance issues alluded to in these Linux commits: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=87e09cdec4dae08acdb4aa49beb793c19d73e73e https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=a99b646afa8a02571ea298bedca6592d818229cd MFC after: 1 week Sponsored by: Chelsio Communications
2018-01-03 19:24:57 +00:00
Optionally panic when cxgbe encounters a fatal error Sometimes it is better to panic than to leave a machine unreachable. Reviewed by: np Sponsored by: Netflix
2018-05-01 15:33:21 +00:00
static int t4_panic_on_fatal_err = 0;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, panic_on_fatal_err, CTLFLAG_RDTUN,
cxgbe(4): Delay the panic due to a fatal error by 30s. This lets information logged by the interrupt handler reach the system log before the system goes down.
2019-02-09 01:49:53 +00:00
&t4_panic_on_fatal_err, 0, "panic on fatal errors");
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
cxgbe(4): Add support for Connection Offload Policy (aka COP). COP allows fine-grained control on whether to offload a TCP connection using t4_tom, and what settings to apply to a connection selected for offload. t4_tom must still be loaded and IFCAP_TOE must still be enabled for full TCP offload to take place on an interface. The difference is that IFCAP_TOE used to be the only knob and would enable TOE for all new connections on the inteface, but now the driver will also consult the COP, if any, before offloading to the hardware TOE. A policy is a plain text file with any number of rules, one per line. Each rule has a "match" part consisting of a socket-type (L = listen, A = active open, P = passive open, D = don't care) and a pcap-filter(7) expression, and a "settings" part that specifies whether to offload the connection or not and the parameters to use if so. The general format of a rule is: [socket-type] expr => settings Example. See cxgbetool(8) for more information. [L] ip && port http => offload [L] port 443 => !offload [L] port ssh => offload [P] src net 192.168/16 && dst port ssh => offload !nagle !timestamp cong newreno [P] dst port ssh => offload !nagle ecn cong tahoe [P] dst port http => offload [A] dst port 443 => offload tls [A] dst net 192.168/16 => offload !timestamp cong highspeed The driver processes the rules for each new listen, active open, or passive open and stops at the first match. There is an implicit rule at the end of every policy that prohibits offload when no rule in the policy matches: [D] all => !offload This is a reworked and expanded version of a patch submitted by Krishnamraju Eraparaju @ Chelsio. Sponsored by: Chelsio Communications
2018-04-14 19:07:56 +00:00
#ifdef TCP_OFFLOAD
/*
* TOE tunables.
*/
static int t4_cop_managed_offloading = 0;
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
SYSCTL_INT(_hw_cxgbe, OID_AUTO, cop_managed_offloading, CTLFLAG_RDTUN,
&t4_cop_managed_offloading, 0,
"COP (Connection Offload Policy) controls all TOE offload");
cxgbe(4): Add support for Connection Offload Policy (aka COP). COP allows fine-grained control on whether to offload a TCP connection using t4_tom, and what settings to apply to a connection selected for offload. t4_tom must still be loaded and IFCAP_TOE must still be enabled for full TCP offload to take place on an interface. The difference is that IFCAP_TOE used to be the only knob and would enable TOE for all new connections on the inteface, but now the driver will also consult the COP, if any, before offloading to the hardware TOE. A policy is a plain text file with any number of rules, one per line. Each rule has a "match" part consisting of a socket-type (L = listen, A = active open, P = passive open, D = don't care) and a pcap-filter(7) expression, and a "settings" part that specifies whether to offload the connection or not and the parameters to use if so. The general format of a rule is: [socket-type] expr => settings Example. See cxgbetool(8) for more information. [L] ip && port http => offload [L] port 443 => !offload [L] port ssh => offload [P] src net 192.168/16 && dst port ssh => offload !nagle !timestamp cong newreno [P] dst port ssh => offload !nagle ecn cong tahoe [P] dst port http => offload [A] dst port 443 => offload tls [A] dst net 192.168/16 => offload !timestamp cong highspeed The driver processes the rules for each new listen, active open, or passive open and stops at the first match. There is an implicit rule at the end of every policy that prohibits offload when no rule in the policy matches: [D] all => !offload This is a reworked and expanded version of a patch submitted by Krishnamraju Eraparaju @ Chelsio. Sponsored by: Chelsio Communications
2018-04-14 19:07:56 +00:00
#endif
cxgbe(4): vi_mac_funcs should include the base Ethernet function. It is already used in the driver as if it does. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-28 07:50:54 +00:00
/* Functions used by VIs to obtain unique MAC addresses for each VI. */
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
static int vi_mac_funcs[] = {
cxgbe(4): vi_mac_funcs should include the base Ethernet function. It is already used in the driver as if it does. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-28 07:50:54 +00:00
FW_VI_FUNC_ETH,
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
FW_VI_FUNC_OFLD,
FW_VI_FUNC_IWARP,
FW_VI_FUNC_OPENISCSI,
FW_VI_FUNC_OPENFCOE,
FW_VI_FUNC_FOISCSI,
FW_VI_FUNC_FOFCOE,
};
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
struct intrs_and_queues {
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
uint16_t intr_type; /* INTx, MSI, or MSI-X */
cxgbe(4): Sanitize t4_num_vis during MOD_LOAD like all other t4_* tunables. Add num_vis to the intrs_and_queues structure as it affects the number of interrupts requested and queues created. In future cfg_itype_and_nqueues might lower it incrementally instead of going straight to 1 when enough interrupts aren't available. Sponsored by: Chelsio Communications
2017-11-16 01:33:53 +00:00
uint16_t num_vis; /* number of VIs for each port */
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
uint16_t nirq; /* Total # of vectors */
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
uint16_t ntxq; /* # of NIC txq's for each port */
uint16_t nrxq; /* # of NIC rxq's for each port */
cxgbe(4): Allocate offload Tx queues when a card has resources provisioned for NIC_ETHOFLD and the kernel has option RATELIMIT. It is possible to use the chip's offload queues for normal NIC Tx and not just TOE Tx. The difference is that these queues support out of order processing of work requests and have a per-"flowid" mechanism for tracking credits between the driver and hardware. This allows Tx for any number of flows bound to different rate limits to be submitted to a single Tx queue and the work requests for slow flows won't cause HOL blocking for the rest. Sponsored by: Chelsio Communications
2018-05-17 01:42:18 +00:00
uint16_t nofldtxq; /* # of TOE/ETHOFLD txq's for each port */
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
uint16_t nofldrxq; /* # of TOE rxq's for each port */
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
/* The vcxgbe/vcxl interfaces use these and not the ones above. */
uint16_t ntxq_vi; /* # of NIC txq's */
uint16_t nrxq_vi; /* # of NIC rxq's */
uint16_t nofldtxq_vi; /* # of TOE txq's */
uint16_t nofldrxq_vi; /* # of TOE rxq's */
uint16_t nnmtxq_vi; /* # of netmap txq's */
uint16_t nnmrxq_vi; /* # of netmap rxq's */
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
};
static void setup_memwin(struct adapter *);
cxgbe(4): Add general purpose routines that offer safe access to the chip's memory windows. Convert existing users of these windows to the new routines.
2016-03-10 06:15:31 +00:00
static void position_memwin(struct adapter *, int, uint32_t);
Treat the memory lengths for CHELSIO_T4_GET_MEM as unsigned. Previously attempts to read the MC region were failing since the length was greater than 2^31. Reviewed by: np MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D17857
2018-11-06 22:33:36 +00:00
static int validate_mem_range(struct adapter *, uint32_t, uint32_t);
cxgbe(4): Update T4 and T5 firmwares to 1.9.12.0
2013-10-14 21:25:07 +00:00
static int fwmtype_to_hwmtype(int);
Treat the memory lengths for CHELSIO_T4_GET_MEM as unsigned. Previously attempts to read the MC region were failing since the length was greater than 2^31. Reviewed by: np MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D17857
2018-11-06 22:33:36 +00:00
static int validate_mt_off_len(struct adapter *, int, uint32_t, uint32_t,
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
uint32_t *);
cxgbe(4): Improvements to the code that deals with the firmware's log. - Query the location of the log very early during attach. Refresh the location later after establishing contact with the firmware. - Save the log's location as a flat address in devlog_params. - Use a memory window instead of backdoor access to the EDC/MC to read the log.
2016-03-10 23:17:26 +00:00
static int fixup_devlog_params(struct adapter *);
cxgbe(4): Sanitize t4_num_vis during MOD_LOAD like all other t4_* tunables. Add num_vis to the intrs_and_queues structure as it affects the number of interrupts requested and queues created. In future cfg_itype_and_nqueues might lower it incrementally instead of going straight to 1 when enough interrupts aren't available. Sponsored by: Chelsio Communications
2017-11-16 01:33:53 +00:00
static int cfg_itype_and_nqueues(struct adapter *, struct intrs_and_queues *);
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
static int contact_firmware(struct adapter *);
static int partition_resources(struct adapter *);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
static int get_params__pre_init(struct adapter *);
cxgbe(4): Request high priority filter support explicitly, as required by recent firmwares. MFC after: 1 week Sponsored by: Chelsio Communications
2019-02-28 05:45:14 +00:00
static int set_params__pre_init(struct adapter *);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
static int get_params__post_init(struct adapter *);
cxgbe(4): Ask the card's firmware to pad up tiny CPLs by encapsulating them in a firmware message if it is able to do so. This works out better for one of the FIFOs in the chip. MFC after: 5 days
2013-02-26 00:27:27 +00:00
static int set_params__post_init(struct adapter *);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
static void t4_set_desc(struct adapter *);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
static bool fixed_ifmedia(struct port_info *);
static void build_medialist(struct port_info *);
static void init_link_config(struct port_info *);
static int fixup_link_config(struct port_info *);
static int apply_link_config(struct port_info *);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
static int cxgbe_init_synchronized(struct vi_info *);
static int cxgbe_uninit_synchronized(struct vi_info *);
cxgbe(4): major tx rework. a) Front load as much work as possible in if_transmit, before any driver lock or software queue has to get involved. b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This is specifically for the tx multiqueue model where one of the if_transmit producer threads becomes the consumer and other producers carry on as usual. mp_ring is implemented as standalone code and it should be possible to use it in any driver with tx multiqueue. It also has: - the ability to enqueue/dequeue multiple items. This might become significant if packet batching is ever implemented. - an abdication mechanism to allow a thread to give up writing tx descriptors and have another if_transmit thread take over. A thread that's writing tx descriptors can end up doing so for an unbounded time period if a) there are other if_transmit threads continuously feeding the sofware queue, and b) the chip keeps up with whatever the thread is throwing at it. - accurate statistics about interesting events even when the stats come at the expense of additional branches/conditional code. The NIC txq lock is uncontested on the fast path at this point. I've left it there for synchronization with the control events (interface up/down, modload/unload). c) Add support for "type 1" coalescing work request in the normal NIC tx path. This work request is optimized for frames with a single item in the DMA gather list. These are very common when forwarding packets. Note that netmap tx in cxgbe already uses these "type 1" work requests. d) Do not request automatic cidx updates every 32 descriptors. Instead, request updates via bits in individual work requests (still every 32 descriptors approximately). Also, request an automatic final update when the queue idles after activity. This means NIC tx reclaim is still performed lazily but it will catch up quickly as soon as the queue idles. This seems to be the best middle ground and I'll probably do something similar for netmap tx as well. e) Implement a faster tx path for WRQs (used by TOE tx and control queues, _not_ by the normal NIC tx). Allow work requests to be written directly to the hardware descriptor ring if room is available. I will convert t4_tom and iw_cxgbe modules to this faster style gradually. MFC after: 2 months
2014-12-31 23:19:16 +00:00
static void quiesce_txq(struct adapter *, struct sge_txq *);
static void quiesce_wrq(struct adapter *, struct sge_wrq *);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
static void quiesce_iq(struct adapter *, struct sge_iq *);
static void quiesce_fl(struct adapter *, struct sge_fl *);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
static int t4_alloc_irq(struct adapter *, struct irq *, int rid,
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
driver_intr_t *, void *, char *);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
static int t4_free_irq(struct adapter *, struct irq *);
cxgbe(4): Updated register dumps. - Get the list of registers to read during a regdump from the shared code instead of the OS specific code. This follows a similar move internally. The shared code includes the list for T6. - Update cxgbetool to be able to decode T5 VF, T6, and T6 VF register dumps (and catch up with some updates to T4 and T5 register decode). Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-07 21:11:35 +00:00
static void get_regs(struct adapter *, struct t4_regdump *, uint8_t *);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
static void vi_refresh_stats(struct adapter *, struct vi_info *);
cxgbe(4): implement if_get_counter.
2014-09-27 05:50:31 +00:00
static void cxgbe_refresh_stats(struct adapter *, struct port_info *);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
static void cxgbe_tick(void *);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
static void cxgbe_sysctls(struct port_info *);
Display holdoff timers and packet counts as a list of numbers. MFC after: 1 week
2011-03-09 21:07:09 +00:00
static int sysctl_int_array(SYSCTL_HANDLER_ARGS);
cxgbe(4): Break up sysctl_bitfield into 8 bit and 16 bit variants. Have them display the current value of the bitfield rather than the fixed value that was provided when the sysctl node was created. MFC after: 1 week Sponsored by: Chelsio Communications
2018-08-06 21:54:51 +00:00
static int sysctl_bitfield_8b(SYSCTL_HANDLER_ARGS);
static int sysctl_bitfield_16b(SYSCTL_HANDLER_ARGS);
- Show the reason why link is down if this information is available. - Display the temperature and PHY firmware version of the BT PHY. MFC after: 1 day
2013-07-05 01:53:51 +00:00
static int sysctl_btphy(SYSCTL_HANDLER_ARGS);
Add a new sysctl, dev.cxgbe.N.rsrv_noflow, and a companion tunable, hw.cxgbe.rsrv_noflow. When set, queue 0 of the port is reserved for TX packets without a flowid. The hash value of packets with a flowid is bumped up by 1. The intent is to provide a private queue for link-level packets like LACP that is unlikely to overflow or suffer deep queue latency. Reviewed by: np Obtained from: Netflix MFC after: 3 days
2014-02-06 18:40:38 +00:00
static int sysctl_noflowq(SYSCTL_HANDLER_ARGS);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
static int sysctl_holdoff_tmr_idx(SYSCTL_HANDLER_ARGS);
static int sysctl_holdoff_pktc_idx(SYSCTL_HANDLER_ARGS);
static int sysctl_qsize_rxq(SYSCTL_HANDLER_ARGS);
static int sysctl_qsize_txq(SYSCTL_HANDLER_ARGS);
cxgbe(4): knobs to enable/disable PAUSE frame based flow control. MFC after: 1 week
2014-09-12 05:25:56 +00:00
static int sysctl_pause_settings(SYSCTL_HANDLER_ARGS);
cxgbe(4): Updates to link configuration. - Update struct link_settings and associated shared code. - Add tunables to control FEC and autonegotiation. All ports inherit these values as their initial settings. hw.cxgbe.fec hw.cxgbe.autoneg - Add per-port sysctls to control FEC and autonegotiation. These can be modified at any time. dev.<port>.<n>.fec dev.<port>.<n>.autoneg MFC after: 3 days Sponsored by: Chelsio Communications
2016-12-30 08:59:49 +00:00
static int sysctl_fec(SYSCTL_HANDLER_ARGS);
static int sysctl_autoneg(SYSCTL_HANDLER_ARGS);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
static int sysctl_handle_t4_reg64(SYSCTL_HANDLER_ARGS);
Display temperature sensor data. Shows -1 if sensor not available on the card. # sysctl dev.t4nex.0.temperature # sysctl dev.t5nex.0.temperature
2013-08-02 18:05:42 +00:00
static int sysctl_temperature(SYSCTL_HANDLER_ARGS);
cxgbe(4): Add a sysctl to report the chip's microprocessor's load averages. This works with debug or custom firmwares only. sysctl dev.<nexus>.<instance>.loadavg sysctl dev.t6nex.0.loadavg MFC after: 1 month Sponsored by: Chelsio Communications
2018-07-10 03:03:10 +00:00
static int sysctl_loadavg(SYSCTL_HANDLER_ARGS);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
static int sysctl_cctrl(SYSCTL_HANDLER_ARGS);
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
static int sysctl_cim_ibq_obq(SYSCTL_HANDLER_ARGS);
static int sysctl_cim_la(SYSCTL_HANDLER_ARGS);
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
static int sysctl_cim_ma_la(SYSCTL_HANDLER_ARGS);
static int sysctl_cim_pif_la(SYSCTL_HANDLER_ARGS);
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
static int sysctl_cim_qcfg(SYSCTL_HANDLER_ARGS);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
static int sysctl_cpl_stats(SYSCTL_HANDLER_ARGS);
static int sysctl_ddp_stats(SYSCTL_HANDLER_ARGS);
Firmware device log. # sysctl dev.t4nex.0.devlog MFC after: mdf's sysctl+sbuf changes are MFC'd
2011-05-31 23:49:13 +00:00
static int sysctl_devlog(SYSCTL_HANDLER_ARGS);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
static int sysctl_fcoe_stats(SYSCTL_HANDLER_ARGS);
static int sysctl_hw_sched(SYSCTL_HANDLER_ARGS);
static int sysctl_lb_stats(SYSCTL_HANDLER_ARGS);
Display a string instead of a numeric code in the linkdnrc sysctl. Submitted by: gnn@
2013-07-27 07:43:43 +00:00
static int sysctl_linkdnrc(SYSCTL_HANDLER_ARGS);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
static int sysctl_meminfo(SYSCTL_HANDLER_ARGS);
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
static int sysctl_mps_tcam(SYSCTL_HANDLER_ARGS);
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
static int sysctl_mps_tcam_t6(SYSCTL_HANDLER_ARGS);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
static int sysctl_path_mtus(SYSCTL_HANDLER_ARGS);
static int sysctl_pm_stats(SYSCTL_HANDLER_ARGS);
static int sysctl_rdma_stats(SYSCTL_HANDLER_ARGS);
static int sysctl_tcp_stats(SYSCTL_HANDLER_ARGS);
static int sysctl_tids(SYSCTL_HANDLER_ARGS);
static int sysctl_tp_err_stats(SYSCTL_HANDLER_ARGS);
cxgbe(4): Add a sysctl for the event capture mask of the TP block's logic analyzer. dev.t5nex.<n>.misc.tp_la_mask dev.t4nex.<n>.misc.tp_la_mask
2016-03-11 01:54:43 +00:00
static int sysctl_tp_la_mask(SYSCTL_HANDLER_ARGS);
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
static int sysctl_tp_la(SYSCTL_HANDLER_ARGS);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
static int sysctl_tx_rate(SYSCTL_HANDLER_ARGS);
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
static int sysctl_ulprx_la(SYSCTL_HANDLER_ARGS);
Cosmetic change (s/wrwc/wcwr/;s/WRWC/WCWR/). MFC after: 3 days.
2013-04-11 22:49:29 +00:00
static int sysctl_wcwr_stats(SYSCTL_HANDLER_ARGS);
cxgbe(4): sysctls to display the local and intr CPUs for the adapter. The driver assumes the list can change (even though it does't right now) and queries it every time the sysctl runs. sysctl dev.<nexus>.<inst>.local_cpus sysctl dev.<nexus>.<inst>.intr_cpus sysctl dev.t6nex.0.local_cpus sysctl dev.t6nex.0.intr_cpus Sponsored by: Chelsio Communications
2018-06-15 18:04:44 +00:00
static int sysctl_cpus(SYSCTL_HANDLER_ARGS);
Fix the following gcc warnings on sparc64, when TCP_OFFLOAD is not defined: sys/dev/cxgbe/t4_main.c:7474: warning: 'sysctl_tp_tick' defined but not used sys/dev/cxgbe/t4_main.c:7505: warning: 'sysctl_tp_dack_timer' defined but not used sys/dev/cxgbe/t4_main.c:7519: warning: 'sysctl_tp_timer' defined but not used This just adds a bunch of #ifdef TCP_OFFLOAD in the right places. Reviewed by: np Differential Revision: https://reviews.freebsd.org/D5620
2016-03-12 18:38:51 +00:00
#ifdef TCP_OFFLOAD
Support for TLS offload of TOE connections on T6 adapters. The TOE engine in Chelsio T6 adapters supports offloading of TLS encryption and TCP segmentation for offloaded connections. Sockets using TLS are required to use a set of custom socket options to upload RX and TX keys to the NIC and to enable RX processing. Currently these socket options are implemented as TCP options in the vendor specific range. A patched OpenSSL library will be made available in a port / package for use with the TLS TOE support. TOE sockets can either offload both transmit and reception of TLS records or just transmit. TLS offload (both RX and TX) is enabled by setting the dev.t6nex.<x>.tls sysctl to 1 and requires TOE to be enabled on the relevant interface. Transmit offload can be used on any "normal" or TLS TOE socket by using the custom socket option to program a transmit key. This permits most TOE sockets to transparently offload TLS when applications use a patched SSL library (e.g. using LD_LIBRARY_PATH to request use of a patched OpenSSL library). Receive offload can only be used with TOE sockets using the TLS mode. The dev.t6nex.0.toe.tls_rx_ports sysctl can be set to a list of TCP port numbers. Any connection with either a local or remote port number in that list will be created as a TLS socket rather than a plain TOE socket. Note that although this sysctl accepts an arbitrary list of port numbers, the sysctl(8) tool is only able to set sysctl nodes to a single value. A TLS socket will hang without receiving data if used by an application that is not using a patched SSL library. Thus, the tls_rx_ports node should be used with care. For a server mostly concerned with offloading TLS transmit, this node is not needed as plain TOE sockets will fall back to software crypto when using an unpatched SSL library. New per-interface statistics nodes are added giving counts of TLS packets and payload bytes (payload bytes do not include TLS headers or authentication tags/MACs) offloaded via the TOE engine, e.g.: dev.cc.0.stats.rx_tls_octets: 149 dev.cc.0.stats.rx_tls_records: 13 dev.cc.0.stats.tx_tls_octets: 26501823 dev.cc.0.stats.tx_tls_records: 1620 TLS transmit work requests are constructed by a new variant of t4_push_frames() called t4_push_tls_records() in tom/t4_tls.c. TLS transmit work requests require a buffer containing IVs. If the IVs are too large to fit into the work request, a separate buffer is allocated when constructing a work request. This buffer is associated with the transmit descriptor and freed when the descriptor is ACKed by the adapter. Received TLS frames use two new CPL messages. The first message is a CPL_TLS_DATA containing the decryped payload of a single TLS record. The handler places the mbuf containing the received payload on an mbufq in the TOE pcb. The second message is a CPL_RX_TLS_CMP message which includes a copy of the TLS header and indicates if there were any errors. The handler for this message places the TLS header into the socket buffer followed by the saved mbuf with the payload data. Both of these handlers are contained in tom/t4_tls.c. A few routines were exposed from t4_cpl_io.c for use by t4_tls.c including send_rx_credits(), a new send_rx_modulate(), and t4_close_conn(). TLS keys for both transmit and receive are stored in onboard memory in the NIC in the "TLS keys" memory region. In some cases a TLS socket can hang with pending data available in the NIC that is not delivered to the host. As a workaround, TLS sockets are more aggressive about sending CPL_RX_DATA_ACK messages anytime that any data is read from a TLS socket. In addition, a fallback timer will periodically send CPL_RX_DATA_ACK messages to the NIC for connections that are still in the handshake phase. Once the connection has finished the handshake and programmed RX keys via the socket option, the timer is stopped. A new function select_ulp_mode() is used to determine what sub-mode a given TOE socket should use (plain TOE, DDP, or TLS). The existing set_tcpddp_ulp_mode() function has been renamed to set_ulp_mode() and handles initialization of TLS-specific state when necessary in addition to DDP-specific state. Since TLS sockets do not receive individual TCP segments but always receive full TLS records, they can receive more data than is available in the current window (e.g. if a 16k TLS record is received but the socket buffer is itself 16k). To cope with this, just drop the window to 0 when this happens, but track the overage and "eat" the overage as it is read from the socket buffer not opening the window (or adding rx_credits) for the overage bytes. Reviewed by: np (earlier version) Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D14529
2018-03-13 23:05:51 +00:00
static int sysctl_tls_rx_ports(SYSCTL_HANDLER_ARGS);
cxgbe(4): sysctls to display the TOE's TCP timers. cask:~# sysctl -d dev.t5nex.0.toe dev.t5nex.0.toe.finwait2_timer: FINWAIT2 timer (us) dev.t5nex.0.toe.initial_srtt: Initial SRTT (us) dev.t5nex.0.toe.keepalive_intvl: Keepidle interval (us) dev.t5nex.0.toe.keepalive_idle: Keepidle idle timer (us) dev.t5nex.0.toe.persist_max: Persist timer max (us) dev.t5nex.0.toe.persist_min: Persist timer min (us) dev.t5nex.0.toe.rexmt_max: Retransmit max (us) dev.t5nex.0.toe.rexmt_min: Retransmit min (us) dev.t5nex.0.toe.dack_timer: DACK timer (us) dev.t5nex.0.toe.dack_tick: DACK tick (us) dev.t5nex.0.toe.timestamp_tick: TCP timestamp tick (us) dev.t5nex.0.toe.timer_tick: TP timer tick (us) ... cask:~# sysctl dev.t5nex.0.toe dev.t5nex.0.toe.finwait2_timer: 9765440 dev.t5nex.0.toe.initial_srtt: 244128 dev.t5nex.0.toe.keepalive_intvl: 73240800 dev.t5nex.0.toe.keepalive_idle: 7031116800 dev.t5nex.0.toe.persist_max: 9765440 dev.t5nex.0.toe.persist_min: 976544 dev.t5nex.0.toe.rexmt_max: 9765440 dev.t5nex.0.toe.rexmt_min: 244128 dev.t5nex.0.toe.dack_timer: 19520 dev.t5nex.0.toe.dack_tick: 32.768 dev.t5nex.0.toe.timestamp_tick: 1048.576 dev.t5nex.0.toe.timer_tick: 32.768 ...
2016-03-11 23:24:04 +00:00
static int sysctl_tp_tick(SYSCTL_HANDLER_ARGS);
static int sysctl_tp_dack_timer(SYSCTL_HANDLER_ARGS);
static int sysctl_tp_timer(SYSCTL_HANDLER_ARGS);
cxgbe(4): Display some more TOE parameters related to retransmission and keepalive in the sysctl MIB. Provide tunables to change some of these parameters. These are supposed to be setup by the firmware so these tunables are for experimentation only. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-24 19:17:13 +00:00
static int sysctl_tp_shift_cnt(SYSCTL_HANDLER_ARGS);
static int sysctl_tp_backoff(SYSCTL_HANDLER_ARGS);
cxgbe(4): Provide knobs to set the holdoff parameters of TOE rx queues separately from NIC rx queues instead of using the same parameters for both types of queues. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-10-05 07:18:16 +00:00
static int sysctl_holdoff_tmr_idx_ofld(SYSCTL_HANDLER_ARGS);
static int sysctl_holdoff_pktc_idx_ofld(SYSCTL_HANDLER_ARGS);
Fix the following gcc warnings on sparc64, when TCP_OFFLOAD is not defined: sys/dev/cxgbe/t4_main.c:7474: warning: 'sysctl_tp_tick' defined but not used sys/dev/cxgbe/t4_main.c:7505: warning: 'sysctl_tp_dack_timer' defined but not used sys/dev/cxgbe/t4_main.c:7519: warning: 'sysctl_tp_timer' defined but not used This just adds a bunch of #ifdef TCP_OFFLOAD in the right places. Reviewed by: np Differential Revision: https://reviews.freebsd.org/D5620
2016-03-12 18:38:51 +00:00
#endif
- driver ioctl to get SGE context for any given queue. - sysctls to display the context id, cidx, and pidx of all kinds of queues. MFC after: 3 days
2011-06-11 04:50:54 +00:00
static int get_sge_context(struct adapter *, struct t4_sge_context *);
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
static int load_fw(struct adapter *, struct t4_data *);
cxgbe(4): Add an ioctl to copy a firmware config file to the card's flash.
2016-10-07 19:02:39 +00:00
static int load_cfg(struct adapter *, struct t4_data *);
cxgbe(4): New ioctls to flash bootrom and boot config to the card. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-18 00:50:58 +00:00
static int load_boot(struct adapter *, struct t4_bootrom *);
static int load_bootcfg(struct adapter *, struct t4_data *);
cxgbe(4): Initial import of the "collect" component of Chelsio unified debug (cudbg) code, hooked up to the main driver via an ioctl. The ioctl can be used to collect the chip's internal state in a compressed dump file. These dumps can be decoded with the "view" component of cudbg. Obtained from: Chelsio Communications MFC after: 2 months Sponsored by: Chelsio Communications
2017-08-03 14:43:30 +00:00
static int cudbg_dump(struct adapter *, struct t4_cudbg_dump *);
cxgbe(4): Add support for Connection Offload Policy (aka COP). COP allows fine-grained control on whether to offload a TCP connection using t4_tom, and what settings to apply to a connection selected for offload. t4_tom must still be loaded and IFCAP_TOE must still be enabled for full TCP offload to take place on an interface. The difference is that IFCAP_TOE used to be the only knob and would enable TOE for all new connections on the inteface, but now the driver will also consult the COP, if any, before offloading to the hardware TOE. A policy is a plain text file with any number of rules, one per line. Each rule has a "match" part consisting of a socket-type (L = listen, A = active open, P = passive open, D = don't care) and a pcap-filter(7) expression, and a "settings" part that specifies whether to offload the connection or not and the parameters to use if so. The general format of a rule is: [socket-type] expr => settings Example. See cxgbetool(8) for more information. [L] ip && port http => offload [L] port 443 => !offload [L] port ssh => offload [P] src net 192.168/16 && dst port ssh => offload !nagle !timestamp cong newreno [P] dst port ssh => offload !nagle ecn cong tahoe [P] dst port http => offload [A] dst port 443 => offload tls [A] dst net 192.168/16 => offload !timestamp cong highspeed The driver processes the rules for each new listen, active open, or passive open and stops at the first match. There is an implicit rule at the end of every policy that prohibits offload when no rule in the policy matches: [D] all => !offload This is a reworked and expanded version of a patch submitted by Krishnamraju Eraparaju @ Chelsio. Sponsored by: Chelsio Communications
2018-04-14 19:07:56 +00:00
static void free_offload_policy(struct t4_offload_policy *);
static int set_offload_policy(struct adapter *, struct t4_offload_policy *);
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
static int read_card_mem(struct adapter *, int, struct t4_mem_range *);
Add a driver ioctl to read a byte from any device on a port's i2c bus. This lets userspace read arbitrary information from the SFP+ modules etc. on this bus. Reading multiple bytes in the same transaction isn't possible right now. I'll update the driver once the chip's firmware supports this. MFC after: 3 days
2012-10-10 17:13:46 +00:00
static int read_i2c(struct adapter *, struct t4_i2c_data *);
cxgbe(4): Clear the freelist statistics in the clearstats ioctl. Move all clearstats code into its own function while here. MFC after: 1 week Sponsored by: Chelsio Communications
2019-07-09 22:24:22 +00:00
static int clear_stats(struct adapter *, u_int);
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
#ifdef TCP_OFFLOAD
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
static int toe_capability(struct vi_info *, int);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
#endif
cxgbe(4): Ensure that the MOD_LOAD handler runs before either t4nex or t5nex attach to their devices. MFC after: 3 days
2013-04-11 17:50:50 +00:00
static int mod_event(module_t, int, void *);
Add a driver to create VF devices on Chelsio T4/T5 NICs. Chelsio NICs are a bit unique compared to some other NICs in that they expose different functionality on different physical functions. In particular, PF4 is used to manage the NIC interfaces ('t4nex' and 't5nex'). However, PF4 is not able to create VF devices. Instead, VFs are only supported by physical functions 0 through 3. This commit adds 't4iov' and 't5iov' drivers that attach to PF0-3. One extra wrinkle is that the iov devices cannot enable SR-IOV until the firwmare has been initialized by the main PF4 driver. To handle this case, a new t4_if kobj interface has been added to permit cross-calls between the PF drivers. The PF4 driver notifies sibling drivers when it is fully attached. It also requests sibling drivers to detach before it detaches. Sibling drivers query the PF4 driver during their attach routine to see if it is attached. If not, the sibling drivers defer their attach actions until the PF4 driver informs them it is attached. VF devices are associated with a single port on the NIC. VF devices created from PF0 are associated with the first port on the NIC, VFs from PF1 are associated with the second port, etc. VF devices can only be created from a PF device that has an associated port. Thus, on a 2-port card, VFs are only supported on PF0 and PF1. Reviewed by: np (earlier versions) MFC after: 1 month Sponsored by: Chelsio Communications
2016-07-22 22:46:41 +00:00
static int notify_siblings(device_t, int);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
struct {
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
uint16_t device;
char *desc;
} t4_pciids[] = {
Allow cxgbe(4) running within a VM to attach to its devices that have been exported via PCI passthrough. - Do not check for a specific physical function (PF) before claiming a device. Different PFs have different device-ids so this check is redundant anyway. - Obtain the PF# from the WHOAMI register instead of pci_get_function(). - Setup the memory windows using the real BAR0 address, not what the VM says it is. Obtained from: Chelsio Communications
2012-06-26 00:34:34 +00:00
{0xa000, "Chelsio Terminator 4 FPGA"},
{0x4400, "Chelsio T440-dbg"},
{0x4401, "Chelsio T420-CR"},
{0x4402, "Chelsio T422-CR"},
{0x4403, "Chelsio T440-CR"},
{0x4404, "Chelsio T420-BCH"},
{0x4405, "Chelsio T440-BCH"},
{0x4406, "Chelsio T440-CH"},
{0x4407, "Chelsio T420-SO"},
{0x4408, "Chelsio T420-CX"},
{0x4409, "Chelsio T420-BT"},
{0x440a, "Chelsio T404-BT"},
cxgbe(4): Add support for the T440-LP-CR card. This is the 4x10G low profile card with a QSFP+ transceiver. MFC after: 3 days
2012-12-22 07:47:07 +00:00
{0x440e, "Chelsio T440-LP-CR"},
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
}, t5_pciids[] = {
{0xb000, "Chelsio Terminator 5 FPGA"},
{0x5400, "Chelsio T580-dbg"},
Expand the list of devices claimed by cxgbe(4).
2013-07-27 00:53:07 +00:00
{0x5401, "Chelsio T520-CR"}, /* 2 x 10G */
{0x5402, "Chelsio T522-CR"}, /* 2 x 10G, 2 X 1G */
Attach to the 4x10G T540-CR card.
2013-07-11 19:09:31 +00:00
{0x5403, "Chelsio T540-CR"}, /* 4 x 10G */
Expand the list of devices claimed by cxgbe(4).
2013-07-27 00:53:07 +00:00
{0x5407, "Chelsio T520-SO"}, /* 2 x 10G, nomem */
{0x5409, "Chelsio T520-BT"}, /* 2 x 10GBaseT */
{0x540a, "Chelsio T504-BT"}, /* 4 x 1G */
{0x540d, "Chelsio T580-CR"}, /* 2 x 40G */
{0x540e, "Chelsio T540-LP-CR"}, /* 4 x 10G */
Attach to the T580 (2 x 40G) card. MFC after: 1 week.
2013-04-30 06:30:21 +00:00
{0x5410, "Chelsio T580-LP-CR"}, /* 2 x 40G */
Expand the list of devices claimed by cxgbe(4).
2013-07-27 00:53:07 +00:00
{0x5411, "Chelsio T520-LL-CR"}, /* 2 x 10G */
{0x5412, "Chelsio T560-CR"}, /* 1 x 40G, 2 x 10G */
{0x5414, "Chelsio T580-LP-SO-CR"}, /* 2 x 40G, nomem */
Add device ID for the T502-BT (dual-port 1G) adapter. Reviewed by: np MFC after: 1 week
2014-11-11 20:05:50 +00:00
{0x5415, "Chelsio T502-BT"}, /* 2 x 1G */
cxgbe(4): Claim some more T5 and T6 boards. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-15 21:54:59 +00:00
{0x5418, "Chelsio T540-BT"}, /* 4 x 10GBaseT */
{0x5419, "Chelsio T540-LP-BT"}, /* 4 x 10GBaseT */
{0x541a, "Chelsio T540-SO-BT"}, /* 4 x 10GBaseT, nomem */
{0x541b, "Chelsio T540-SO-CR"}, /* 4 x 10G, nomem */
cxgbe(4): Attach to two T540 variants. MFC after: 1 week
2018-12-30 01:57:11 +00:00
/* Custom */
{0x5483, "Custom T540-CR"},
{0x5484, "Custom T540-BT"},
cxgbe(4): Attach to cards with the Terminator 6 ASIC. T6 cards will come up as 't6nex' nexus devices with 'cc' ports hanging off them. The T6 firmware and configuration files will be added as soon as they are released. For now the driver will try to work with whatever firmware and configuration is on the card's flash. Sponsored by: Chelsio Communications
2016-09-16 00:08:37 +00:00
}, t6_pciids[] = {
{0xc006, "Chelsio Terminator 6 FPGA"}, /* T6 PE10K6 FPGA (PF0) */
cxgbe(4): Update the list of PCIe devices claimed by the driver. At this point any board with a T6 should just work. Obtained from: Chelsio Communications MFC after: 1 week Sponsored by: Chelsio Communications
2017-05-05 00:54:23 +00:00
{0x6400, "Chelsio T6-DBG-25"}, /* 2 x 10/25G, debug */
cxgbe(4): Attach to cards with the Terminator 6 ASIC. T6 cards will come up as 't6nex' nexus devices with 'cc' ports hanging off them. The T6 firmware and configuration files will be added as soon as they are released. For now the driver will try to work with whatever firmware and configuration is on the card's flash. Sponsored by: Chelsio Communications
2016-09-16 00:08:37 +00:00
{0x6401, "Chelsio T6225-CR"}, /* 2 x 10/25G */
{0x6402, "Chelsio T6225-SO-CR"}, /* 2 x 10/25G, nomem */
cxgbe(4): Update the list of PCIe devices claimed by the driver. At this point any board with a T6 should just work. Obtained from: Chelsio Communications MFC after: 1 week Sponsored by: Chelsio Communications
2017-05-05 00:54:23 +00:00
{0x6403, "Chelsio T6425-CR"}, /* 4 x 10/25G */
{0x6404, "Chelsio T6425-SO-CR"}, /* 4 x 10/25G, nomem */
{0x6405, "Chelsio T6225-OCP-SO"}, /* 2 x 10/25G, nomem */
{0x6406, "Chelsio T62100-OCP-SO"}, /* 2 x 40/50/100G, nomem */
cxgbe(4): Attach to cards with the Terminator 6 ASIC. T6 cards will come up as 't6nex' nexus devices with 'cc' ports hanging off them. The T6 firmware and configuration files will be added as soon as they are released. For now the driver will try to work with whatever firmware and configuration is on the card's flash. Sponsored by: Chelsio Communications
2016-09-16 00:08:37 +00:00
{0x6407, "Chelsio T62100-LP-CR"}, /* 2 x 40/50/100G */
{0x6408, "Chelsio T62100-SO-CR"}, /* 2 x 40/50/100G, nomem */
cxgbe(4): Update the list of PCIe devices claimed by the driver. At this point any board with a T6 should just work. Obtained from: Chelsio Communications MFC after: 1 week Sponsored by: Chelsio Communications
2017-05-05 00:54:23 +00:00
{0x6409, "Chelsio T6210-BT"}, /* 2 x 10GBASE-T */
cxgbe(4): Attach to cards with the Terminator 6 ASIC. T6 cards will come up as 't6nex' nexus devices with 'cc' ports hanging off them. The T6 firmware and configuration files will be added as soon as they are released. For now the driver will try to work with whatever firmware and configuration is on the card's flash. Sponsored by: Chelsio Communications
2016-09-16 00:08:37 +00:00
{0x640d, "Chelsio T62100-CR"}, /* 2 x 40/50/100G */
cxgbe(4): Update the list of PCIe devices claimed by the driver. At this point any board with a T6 should just work. Obtained from: Chelsio Communications MFC after: 1 week Sponsored by: Chelsio Communications
2017-05-05 00:54:23 +00:00
{0x6410, "Chelsio T6-DBG-100"}, /* 2 x 40/50/100G, debug */
{0x6411, "Chelsio T6225-LL-CR"}, /* 2 x 10/25G */
{0x6414, "Chelsio T61100-OCP-SO"}, /* 1 x 40/50/100G, nomem */
{0x6415, "Chelsio T6201-BT"}, /* 2 x 1000BASE-T */
/* Custom */
cxgbe(4): Claim some more T5 and T6 boards. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-15 21:54:59 +00:00
{0x6480, "Custom T6225-CR"},
{0x6481, "Custom T62100-CR"},
{0x6482, "Custom T6225-CR"},
{0x6483, "Custom T62100-CR"},
{0x6484, "Custom T64100-CR"},
{0x6485, "Custom T6240-SO"},
{0x6486, "Custom T6225-SO-CR"},
{0x6487, "Custom T6225-CR"},
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
};
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
#ifdef TCP_OFFLOAD
/*
cxgbe(4): Create two variants of service_iq, one for queues with freelists and one for those without. MFH: 3 weeks Sponsored by: Chelsio Communications
2018-08-11 04:55:47 +00:00
* service_iq_fl() has an iq and needs the fl. Offset of fl from the iq should
* be exactly the same for both rxq and ofld_rxq.
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
*/
CTASSERT(offsetof(struct sge_ofld_rxq, iq) == offsetof(struct sge_rxq, iq));
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
CTASSERT(offsetof(struct sge_ofld_rxq, fl) == offsetof(struct sge_rxq, fl));
#endif
cxgbe(4): significant rx rework. - More flexible cluster size selection, including the ability to fall back to a safe cluster size (PAGE_SIZE from zone_jumbop by default) in case an allocation of a larger size fails. - A single get_fl_payload() function that assembles the payload into an mbuf chain for any kind of freelist. This replaces two variants: one for freelists with buffer packing enabled and another for those without. - Buffer packing with any sized cluster. It was limited to 4K clusters only before this change. - Enable buffer packing for TOE rx queues as well. - Statistics and tunables to go with all these changes. The driver's man page will be updated separately. MFC after: 5 weeks
2014-03-18 20:14:13 +00:00
CTASSERT(sizeof(struct cluster_metadata) <= CL_METADATA_SIZE);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
static int
t4_probe(device_t dev)
{
int i;
uint16_t v = pci_get_vendor(dev);
uint16_t d = pci_get_device(dev);
Allow cxgbe(4) running within a VM to attach to its devices that have been exported via PCI passthrough. - Do not check for a specific physical function (PF) before claiming a device. Different PFs have different device-ids so this check is redundant anyway. - Obtain the PF# from the WHOAMI register instead of pci_get_function(). - Setup the memory windows using the real BAR0 address, not what the VM says it is. Obtained from: Chelsio Communications
2012-06-26 00:34:34 +00:00
uint8_t f = pci_get_function(dev);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (v != PCI_VENDOR_ID_CHELSIO)
return (ENXIO);
Allow cxgbe(4) running within a VM to attach to its devices that have been exported via PCI passthrough. - Do not check for a specific physical function (PF) before claiming a device. Different PFs have different device-ids so this check is redundant anyway. - Obtain the PF# from the WHOAMI register instead of pci_get_function(). - Setup the memory windows using the real BAR0 address, not what the VM says it is. Obtained from: Chelsio Communications
2012-06-26 00:34:34 +00:00
/* Attach only to PF0 of the FPGA */
if (d == 0xa000 && f != 0)
return (ENXIO);
Use native FreeBSD facilities everywhere except the shared code in common/ MFC after: 1 week
2012-09-13 09:10:10 +00:00
for (i = 0; i < nitems(t4_pciids); i++) {
Allow cxgbe(4) running within a VM to attach to its devices that have been exported via PCI passthrough. - Do not check for a specific physical function (PF) before claiming a device. Different PFs have different device-ids so this check is redundant anyway. - Obtain the PF# from the WHOAMI register instead of pci_get_function(). - Setup the memory windows using the real BAR0 address, not what the VM says it is. Obtained from: Chelsio Communications
2012-06-26 00:34:34 +00:00
if (d == t4_pciids[i].device) {
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
device_set_desc(dev, t4_pciids[i].desc);
return (BUS_PROBE_DEFAULT);
}
}
return (ENXIO);
}
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
static int
t5_probe(device_t dev)
{
int i;
uint16_t v = pci_get_vendor(dev);
uint16_t d = pci_get_device(dev);
uint8_t f = pci_get_function(dev);
if (v != PCI_VENDOR_ID_CHELSIO)
return (ENXIO);
/* Attach only to PF0 of the FPGA */
if (d == 0xb000 && f != 0)
return (ENXIO);
for (i = 0; i < nitems(t5_pciids); i++) {
if (d == t5_pciids[i].device) {
device_set_desc(dev, t5_pciids[i].desc);
return (BUS_PROBE_DEFAULT);
}
}
return (ENXIO);
}
cxgbe(4): Attach to cards with the Terminator 6 ASIC. T6 cards will come up as 't6nex' nexus devices with 'cc' ports hanging off them. The T6 firmware and configuration files will be added as soon as they are released. For now the driver will try to work with whatever firmware and configuration is on the card's flash. Sponsored by: Chelsio Communications
2016-09-16 00:08:37 +00:00
static int
t6_probe(device_t dev)
{
int i;
uint16_t v = pci_get_vendor(dev);
uint16_t d = pci_get_device(dev);
if (v != PCI_VENDOR_ID_CHELSIO)
return (ENXIO);
for (i = 0; i < nitems(t6_pciids); i++) {
if (d == t6_pciids[i].device) {
device_set_desc(dev, t6_pciids[i].desc);
return (BUS_PROBE_DEFAULT);
}
}
return (ENXIO);
}
Chelsio T5 chips do not properly echo the No Snoop and Relaxed Ordering attributes when replying to a TLP from a Root Port. As a workaround, disable No Snoop and Relaxed Ordering in the Root Port of each T5 adapter during attach so that CPU-initiated requests do not contain these flags. Note that this affects CPU-initiated requests to all devices under this root port. Reviewed by: np MFC after: 1 week Sponsored by: Chelsio
2015-11-05 21:33:15 +00:00
static void
t5_attribute_workaround(device_t dev)
{
device_t root_port;
uint32_t v;
/*
* The T5 chips do not properly echo the No Snoop and Relaxed
* Ordering attributes when replying to a TLP from a Root
* Port. As a workaround, find the parent Root Port and
* disable No Snoop and Relaxed Ordering. Note that this
* affects all devices under this root port.
*/
root_port = pci_find_pcie_root_port(dev);
if (root_port == NULL) {
device_printf(dev, "Unable to find parent root port\n");
return;
}
v = pcie_adjust_config(root_port, PCIER_DEVICE_CTL,
PCIEM_CTL_RELAXED_ORD_ENABLE | PCIEM_CTL_NOSNOOP_ENABLE, 0, 2);
if ((v & (PCIEM_CTL_RELAXED_ORD_ENABLE | PCIEM_CTL_NOSNOOP_ENABLE)) !=
0)
device_printf(dev, "Disabled No Snoop/Relaxed Ordering on %s\n",
device_get_nameunit(root_port));
}
cxgbe(4): Attach to cards with the Terminator 6 ASIC. T6 cards will come up as 't6nex' nexus devices with 'cc' ports hanging off them. The T6 firmware and configuration files will be added as soon as they are released. For now the driver will try to work with whatever firmware and configuration is on the card's flash. Sponsored by: Chelsio Communications
2016-09-16 00:08:37 +00:00
static const struct devnames devnames[] = {
{
.nexus_name = "t4nex",
.ifnet_name = "cxgbe",
.vi_ifnet_name = "vcxgbe",
.pf03_drv_name = "t4iov",
.vf_nexus_name = "t4vf",
.vf_ifnet_name = "cxgbev"
}, {
.nexus_name = "t5nex",
.ifnet_name = "cxl",
.vi_ifnet_name = "vcxl",
.pf03_drv_name = "t5iov",
.vf_nexus_name = "t5vf",
.vf_ifnet_name = "cxlv"
}, {
.nexus_name = "t6nex",
.ifnet_name = "cc",
.vi_ifnet_name = "vcc",
.pf03_drv_name = "t6iov",
.vf_nexus_name = "t6vf",
.vf_ifnet_name = "ccv"
}
};
void
t4_init_devnames(struct adapter *sc)
{
int id;
id = chip_id(sc);
if (id >= CHELSIO_T4 && id - CHELSIO_T4 < nitems(devnames))
sc->names = &devnames[id - CHELSIO_T4];
else {
device_printf(sc->dev, "chip id %d is not supported.\n", id);
sc->names = NULL;
}
}
Add support for port unit wiring to cxgbe(4). - Add a bus_child_location_str method to the nexus drivers that prints out 'port=N' as the location string exported via devinfo and the '%location' sysctl node. - We can't use a bus_hint_device_unit to wire the unit numbers of devices with a fixed devclass as the device gets assigned a unit in make_device() before the device creator can set softc, etc. Instead, when adding a child device, use a helper function much like a bus_hint_device_unit method to look for wiring hints or to return -1 to let the system choose a unit number. This function requires an "at" hint for the port pointing to the nexus device and a "port" hint listing the port number. For example: hint.cxl.4.at="t5nex0" hint.cxl.4.port="0" wires cxl4 to the first port on the t5nex0 adapter. Requested by: gallatin MFC after: 2 months
2018-11-01 21:46:37 +00:00
static int
t4_ifnet_unit(struct adapter *sc, struct port_info *pi)
{
const char *parent, *name;
long value;
int line, unit;
line = 0;
parent = device_get_nameunit(sc->dev);
name = sc->names->ifnet_name;
while (resource_find_dev(&line, name, &unit, "at", parent) == 0) {
if (resource_long_value(name, unit, "port", &value) == 0 &&
value == pi->port_id)
return (unit);
}
return (-1);
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
static int
t4_attach(device_t dev)
{
struct adapter *sc;
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
int rc = 0, i, j, rqidx, tqidx, nports;
Various fixes to the t4/5nex character device. - Remove null open/close methods. - Don't set d_flags to 0 explicitly. - Remove t5_cdevsw as the .d_name member isn't really used and doesn't warrant a separate cdevsw just for the name. - Use ENOTTY as the error value for an unknown ioctl request. - Use make_dev_s() to close race with setting si_drv1. Sponsored by: Chelsio Communications
2016-07-29 22:11:29 +00:00
struct make_dev_args mda;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
struct intrs_and_queues iaq;
struct sge *s;
cxgbe(4): Read the MFG diags version from the VPD and make it available in the sysctl MIB. MFC after: 1 week Sponsored by: Chelsio Communications
2017-12-21 15:19:43 +00:00
uint32_t *buf;
cxgbe(4): Allocate offload Tx queues when a card has resources provisioned for NIC_ETHOFLD and the kernel has option RATELIMIT. It is possible to use the chip's offload queues for normal NIC Tx and not just TOE Tx. The difference is that these queues support out of order processing of work requests and have a per-"flowid" mechanism for tracking credits between the driver and hardware. This allows Tx for any number of flows bound to different rate limits to be submitted to a single Tx queue and the work requests for slow flows won't cause HOL blocking for the rest. Sponsored by: Chelsio Communications
2018-05-17 01:42:18 +00:00
#if defined(TCP_OFFLOAD) || defined(RATELIMIT)
int ofld_tqidx;
#endif
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
#ifdef TCP_OFFLOAD
cxgbe(4): Allocate offload Tx queues when a card has resources provisioned for NIC_ETHOFLD and the kernel has option RATELIMIT. It is possible to use the chip's offload queues for normal NIC Tx and not just TOE Tx. The difference is that these queues support out of order processing of work requests and have a per-"flowid" mechanism for tracking credits between the driver and hardware. This allows Tx for any number of flows bound to different rate limits to be submitted to a single Tx queue and the work requests for slow flows won't cause HOL blocking for the rest. Sponsored by: Chelsio Communications
2018-05-17 01:42:18 +00:00
int ofld_rqidx;
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
#endif
#ifdef DEV_NETMAP
int nm_rqidx, nm_tqidx;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
#endif
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
int num_vis;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
sc = device_get_softc(dev);
sc->dev = dev;
cxgbe(4): Rename the debug_flags driver tunable/sysctl to dflags. Tunables that end with _flags are special. Sponsored by: Chelsio Communications
2016-09-11 18:05:37 +00:00
TUNABLE_INT_FETCH("hw.cxgbe.dflags", &sc->debug_flags);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Chelsio T5 chips do not properly echo the No Snoop and Relaxed Ordering attributes when replying to a TLP from a Root Port. As a workaround, disable No Snoop and Relaxed Ordering in the Root Port of each T5 adapter during attach so that CPU-initiated requests do not contain these flags. Note that this affects CPU-initiated requests to all devices under this root port. Reviewed by: np MFC after: 1 week Sponsored by: Chelsio
2015-11-05 21:33:15 +00:00
if ((pci_get_device(dev) & 0xff00) == 0x5400)
t5_attribute_workaround(dev);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
pci_enable_busmaster(dev);
- Enable per-channel congestion notification. - Enable PCIe relaxed ordering for all egress queues and rx data buffers. MFC after: 3 days
2011-05-18 22:09:04 +00:00
if (pci_find_cap(dev, PCIY_EXPRESS, &i) == 0) {
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
uint32_t v;
- Enable per-channel congestion notification. - Enable PCIe relaxed ordering for all egress queues and rx data buffers. MFC after: 3 days
2011-05-18 22:09:04 +00:00
pci_set_max_read_req(dev, 4096);
Align the PCI Express #defines with the style used for the PCI-X #defines. This also has the advantage that it makes the names more compact, iand also allows us to correct the non-uniform naming of the PCIM_LINK_* defines, making them all consistent amongst themselves. This is a mostly mechanical rename: s/PCIR_EXPRESS_/PCIER_/g s/PCIM_EXP_/PCIEM_/g s/PCIM_LINK_/PCIEM_LINK_/g When this is MFC'd, #defines will be added for the old names to assist out-of-tree drivers. Discussed with: jhb MFC after: 1 week
2012-09-18 22:04:59 +00:00
v = pci_read_config(dev, i + PCIER_DEVICE_CTL, 2);
cxgbe(4): figure out the max payload size and save it for later. MFC after: 1 week
2014-11-19 20:16:56 +00:00
sc->params.pci.mps = 128 << ((v & PCIEM_CTL_MAX_PAYLOAD) >> 5);
cxgbe(4): Add a knob to enable/disable PCIe relaxed ordering. Disable it by default when running on Intel CPUs. This is a crude fix for the performance issues alluded to in these Linux commits: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=87e09cdec4dae08acdb4aa49beb793c19d73e73e https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=a99b646afa8a02571ea298bedca6592d818229cd MFC after: 1 week Sponsored by: Chelsio Communications
2018-01-03 19:24:57 +00:00
if (pcie_relaxed_ordering == 0 &&
cxgbe(4): Check the RO bit properly before disabling relaxed ordering. Reported by: Coverity (CID 1384286) MFC after: 1 week Sponsored by: Chelsio Communications
2018-08-21 21:32:51 +00:00
(v & PCIEM_CTL_RELAXED_ORD_ENABLE) != 0) {
cxgbe(4): Add a knob to enable/disable PCIe relaxed ordering. Disable it by default when running on Intel CPUs. This is a crude fix for the performance issues alluded to in these Linux commits: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=87e09cdec4dae08acdb4aa49beb793c19d73e73e https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=a99b646afa8a02571ea298bedca6592d818229cd MFC after: 1 week Sponsored by: Chelsio Communications
2018-01-03 19:24:57 +00:00
v &= ~PCIEM_CTL_RELAXED_ORD_ENABLE;
pci_write_config(dev, i + PCIER_DEVICE_CTL, v, 2);
} else if (pcie_relaxed_ordering == 1 &&
(v & PCIEM_CTL_RELAXED_ORD_ENABLE) == 0) {
v |= PCIEM_CTL_RELAXED_ORD_ENABLE;
pci_write_config(dev, i + PCIER_DEVICE_CTL, v, 2);
}
- Enable per-channel congestion notification. - Enable PCIe relaxed ordering for all egress queues and rx data buffers. MFC after: 3 days
2011-05-18 22:09:04 +00:00
}
Store the offset of the KDOORBELL and GTS registers in the softc. VF devices use a different register layout than PF devices. Storing the offset in a value in the softc allows code to be shared between the PF and VF drivers. Reviewed by: np Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7389
2016-08-01 22:39:51 +00:00
sc->sge_gts_reg = MYPF_REG(A_SGE_PF_GTS);
sc->sge_kdoorbell_reg = MYPF_REG(A_SGE_PF_KDOORBELL);
Add support for packet-sniffing tracers to cxgbe(4). This works with all T4 and T5 based cards and is useful for analyzing TSO, LRO, TOE, and for general purpose monitoring without tapping any cxgbe or cxl ifnet directly. Tracers on the T4/T5 chips provide access to Ethernet frames exactly as they were received from or transmitted on the wire. On transmit, a tracer will capture a frame after TSO segmentation, hw VLAN tag insertion, hw L3 & L4 checksum insertion, etc. It will also capture frames generated by the TCP offload engine (TOE traffic is normally invisible to the kernel). On receive, a tracer will capture a frame before hw VLAN extraction, runt filtering, other badness filtering, before the steering/drop/L2-rewrite filters or the TOE have had a go at it, and of course before sw LRO in the driver. There are 4 tracers on a chip. A tracer can trace only in one direction (tx or rx). For now cxgbetool will set up tracers to capture the first 128B of every transmitted or received frame on a given port. This is a small subset of what the hardware can do. A pseudo ifnet with the same name as the nexus driver (t4nex0 or t5nex0) will be created for tracing. The data delivered to this ifnet is an additional copy made inside the chip. Normal delivery to cxgbe<n> or cxl<n> will be made as usual. /* watch cxl0, which is the first port hanging off t5nex0. */ # cxgbetool t5nex0 tracer 0 tx0 (watch what cxl0 is transmitting) # cxgbetool t5nex0 tracer 1 rx0 (watch what cxl0 is receiving) # cxgbetool t5nex0 tracer list # tcpdump -i t5nex0 <== all that cxl0 sees and puts on the wire If you were doing TSO, a tcpdump on cxl0 may have shown you ~64K "frames" with no L3/L4 checksum but this will show you the frames that were actually transmitted. /* all done */ # cxgbetool t5nex0 tracer 0 disable # cxgbetool t5nex0 tracer 1 disable # cxgbetool t5nex0 tracer list # ifconfig t5nex0 destroy
2013-07-26 22:04:11 +00:00
sc->traceq = -1;
mtx_init(&sc->ifp_lock, sc->ifp_lockname, 0, MTX_DEF);
snprintf(sc->ifp_lockname, sizeof(sc->ifp_lockname), "%s tracer",
device_get_nameunit(dev));
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
snprintf(sc->lockname, sizeof(sc->lockname), "%s",
device_get_nameunit(dev));
mtx_init(&sc->sc_lock, sc->lockname, 0, MTX_DEF);
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
t4_add_adapter(sc);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
mtx_init(&sc->sfl_lock, "starving freelists", 0, MTX_DEF);
TAILQ_INIT(&sc->sfl);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
callout_init_mtx(&sc->sfl_callout, &sc->sfl_lock, 0);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Rename regwin_lock to reg_lock. It is used to protect access to indirect registers only.
2016-03-08 22:23:30 +00:00
mtx_init(&sc->reg_lock, "indirect register access", 0, MTX_DEF);
cxgbe(4): implement if_get_counter.
2014-09-27 05:50:31 +00:00
cxgbe(4): Add support for Connection Offload Policy (aka COP). COP allows fine-grained control on whether to offload a TCP connection using t4_tom, and what settings to apply to a connection selected for offload. t4_tom must still be loaded and IFCAP_TOE must still be enabled for full TCP offload to take place on an interface. The difference is that IFCAP_TOE used to be the only knob and would enable TOE for all new connections on the inteface, but now the driver will also consult the COP, if any, before offloading to the hardware TOE. A policy is a plain text file with any number of rules, one per line. Each rule has a "match" part consisting of a socket-type (L = listen, A = active open, P = passive open, D = don't care) and a pcap-filter(7) expression, and a "settings" part that specifies whether to offload the connection or not and the parameters to use if so. The general format of a rule is: [socket-type] expr => settings Example. See cxgbetool(8) for more information. [L] ip && port http => offload [L] port 443 => !offload [L] port ssh => offload [P] src net 192.168/16 && dst port ssh => offload !nagle !timestamp cong newreno [P] dst port ssh => offload !nagle ecn cong tahoe [P] dst port http => offload [A] dst port 443 => offload tls [A] dst net 192.168/16 => offload !timestamp cong highspeed The driver processes the rules for each new listen, active open, or passive open and stops at the first match. There is an implicit rule at the end of every policy that prohibits offload when no rule in the policy matches: [D] all => !offload This is a reworked and expanded version of a patch submitted by Krishnamraju Eraparaju @ Chelsio. Sponsored by: Chelsio Communications
2018-04-14 19:07:56 +00:00
sc->policy = NULL;
rw_init(&sc->policy_lock, "connection offload policy");
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
rc = t4_map_bars_0_and_4(sc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (rc != 0)
goto done; /* error message displayed already */
memset(sc->chan_map, 0xff, sizeof(sc->chan_map));
cxgbe(4): Updates to the shared routines that deal with the serial EEPROM, flash, and VPD. Obtained from: Chelsio Communications
2016-03-08 07:48:55 +00:00
/* Prepare the adapter for operation. */
buf = malloc(PAGE_SIZE, M_CXGBE, M_ZERO | M_WAITOK);
rc = -t4_prep_adapter(sc, buf);
free(buf, M_CXGBE);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (rc != 0) {
device_printf(dev, "failed to prepare adapter: %d.\n", rc);
goto done;
}
cxgbe(4): Catch up with the different layout of WHOAMI in T6. Note that the code moved below t4_prep_adapter() as part of this change because now it needs a working chip_id().
2016-09-22 18:47:07 +00:00
/*
* This is the real PF# to which we're attaching. Works from within PCI
* passthrough environments too, where pci_get_function() could return a
* different PF# depending on the passthrough configuration. We need to
* use the real PF# in all our communication with the firmware.
*/
j = t4_read_reg(sc, A_PL_WHOAMI);
sc->pf = chip_id(sc) <= CHELSIO_T5 ? G_SOURCEPF(j) : G_T6_SOURCEPF(j);
sc->mbox = sc->pf;
cxgbe(4): Attach to cards with the Terminator 6 ASIC. T6 cards will come up as 't6nex' nexus devices with 'cc' ports hanging off them. The T6 firmware and configuration files will be added as soon as they are released. For now the driver will try to work with whatever firmware and configuration is on the card's flash. Sponsored by: Chelsio Communications
2016-09-16 00:08:37 +00:00
t4_init_devnames(sc);
if (sc->names == NULL) {
rc = ENOTSUP;
goto done; /* error message displayed already */
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
/*
* Do this really early, with the memory windows set up even before the
* character device. The userland tool's register i/o and mem read
* will work even in "recovery mode".
*/
setup_memwin(sc);
cxgbe(4): Improvements to the code that deals with the firmware's log. - Query the location of the log very early during attach. Refresh the location later after establishing contact with the firmware. - Save the log's location as a flat address in devlog_params. - Use a memory window instead of backdoor access to the EDC/MC to read the log.
2016-03-10 23:17:26 +00:00
if (t4_init_devlog_params(sc, 0) == 0)
fixup_devlog_params(sc);
Various fixes to the t4/5nex character device. - Remove null open/close methods. - Don't set d_flags to 0 explicitly. - Remove t5_cdevsw as the .d_name member isn't really used and doesn't warrant a separate cdevsw just for the name. - Use ENOTTY as the error value for an unknown ioctl request. - Use make_dev_s() to close race with setting si_drv1. Sponsored by: Chelsio Communications
2016-07-29 22:11:29 +00:00
make_dev_args_init(&mda);
mda.mda_devsw = &t4_cdevsw;
mda.mda_uid = UID_ROOT;
mda.mda_gid = GID_WHEEL;
mda.mda_mode = 0600;
mda.mda_si_drv1 = sc;
rc = make_dev_s(&mda, &sc->cdev, "%s", device_get_nameunit(dev));
if (rc != 0)
device_printf(dev, "failed to create nexus char device: %d.\n",
rc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
/* Go no further if recovery mode has been requested. */
if (TUNABLE_INT_FETCH("hw.cxgbe.sos", &i) && i != 0) {
device_printf(dev, "recovery mode.\n");
goto done;
}
Allow cxgbe(4) to be built on i386. Driver attach will succeed only on a subset of i386 systems.
2015-01-16 01:32:40 +00:00
#if defined(__i386__)
if ((cpu_feature & CPUID_CX8) == 0) {
device_printf(dev, "64 bit atomics not available.\n");
rc = ENOTSUP;
goto done;
}
#endif
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
/* Contact the firmware and try to become the master driver. */
rc = contact_firmware(sc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (rc != 0)
goto done; /* error message displayed already */
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
MPASS(sc->flags & FW_OK);
rc = get_params__pre_init(sc);
if (rc != 0)
goto done; /* error message displayed already */
if (sc->flags & MASTER_PF) {
rc = partition_resources(sc);
if (rc != 0)
goto done; /* error message displayed already */
cxgbe(4): Improved error reporting and diagnostics. "slow" interrupt handler: - Expand the list of INT_CAUSE registers known to the driver. - Add decode information for many more bits but decouple it from the rest of intr_info so that it is entirely optional. - Call t4_fatal_err exactly once, and from the top level PL intr handler. t4_fatal_err: - Use t4_shutdown_adapter from the common code to stop the adapter. - Stop servicing slow interrupts after the first fatal one. Driver/firmware interaction: - CH_DUMP_MBOX: note whether the mailbox being dumped is a command or a reply or something else. - Log the raw value of pcie_fw for some errors. - Use correct log levels (debug vs. error). Sponsored by: Chelsio Communications
2019-02-01 20:42:49 +00:00
t4_intr_clear(sc);
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
rc = get_params__post_init(sc);
if (rc != 0)
goto done; /* error message displayed already */
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Ask the card's firmware to pad up tiny CPLs by encapsulating them in a firmware message if it is able to do so. This works out better for one of the FIFOs in the chip. MFC after: 5 days
2013-02-26 00:27:27 +00:00
rc = set_params__post_init(sc);
if (rc != 0)
goto done; /* error message displayed already */
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
rc = t4_map_bar_2(sc);
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
if (rc != 0)
goto done; /* error message displayed already */
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
rc = t4_create_dma_tag(sc);
if (rc != 0)
goto done; /* error message displayed already */
/*
* First pass over all the ports - allocate VIs and initialize some
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
* basic parameters like mac address, port type, etc.
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
*/
for_each_port(sc, i) {
struct port_info *pi;
pi = malloc(sizeof(*pi), M_CXGBE, M_ZERO | M_WAITOK);
sc->port[i] = pi;
/* These must be set before t4_port_init */
pi->adapter = sc;
pi->port_id = i;
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
/*
* XXX: vi[0] is special so we can't delay this allocation until
* pi->nvi's final value is known.
*/
cxgbe(4): Sanitize t4_num_vis during MOD_LOAD like all other t4_* tunables. Add num_vis to the intrs_and_queues structure as it affects the number of interrupts requested and queues created. In future cfg_itype_and_nqueues might lower it incrementally instead of going straight to 1 when enough interrupts aren't available. Sponsored by: Chelsio Communications
2017-11-16 01:33:53 +00:00
pi->vi = malloc(sizeof(struct vi_info) * t4_num_vis, M_CXGBE,
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
M_ZERO | M_WAITOK);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
/*
* Allocate the "main" VI and initialize parameters
* like mac addr.
*/
cxgbe(4): Updates to the shared routines that deal with the serial EEPROM, flash, and VPD. Obtained from: Chelsio Communications
2016-03-08 07:48:55 +00:00
rc = -t4_port_init(sc, sc->mbox, sc->pf, 0, i);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (rc != 0) {
device_printf(dev, "unable to initialize port %d: %d\n",
i, rc);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
free(pi->vi, M_CXGBE);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
free(pi, M_CXGBE);
- Specialized ingress queues that take interrupts for other ingress queues. Try to have a set of these per port when possible, fall back to sharing a common pool between all ports otherwise. - One control queue per port (used to be one per hardware channel). - t4_eth_rx now handles Ethernet rx only. - sysctls to display pidx/cidx for some queues. MFC after: 1 week
2011-05-30 21:34:44 +00:00
sc->port[i] = NULL;
goto done;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
cxgbe(4): knobs to enable/disable PAUSE frame based flow control. MFC after: 1 week
2014-09-12 05:25:56 +00:00
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
snprintf(pi->lockname, sizeof(pi->lockname), "%sp%d",
device_get_nameunit(dev), i);
mtx_init(&pi->pi_lock, pi->lockname, 0, MTX_DEF);
Add support for packet-sniffing tracers to cxgbe(4). This works with all T4 and T5 based cards and is useful for analyzing TSO, LRO, TOE, and for general purpose monitoring without tapping any cxgbe or cxl ifnet directly. Tracers on the T4/T5 chips provide access to Ethernet frames exactly as they were received from or transmitted on the wire. On transmit, a tracer will capture a frame after TSO segmentation, hw VLAN tag insertion, hw L3 & L4 checksum insertion, etc. It will also capture frames generated by the TCP offload engine (TOE traffic is normally invisible to the kernel). On receive, a tracer will capture a frame before hw VLAN extraction, runt filtering, other badness filtering, before the steering/drop/L2-rewrite filters or the TOE have had a go at it, and of course before sw LRO in the driver. There are 4 tracers on a chip. A tracer can trace only in one direction (tx or rx). For now cxgbetool will set up tracers to capture the first 128B of every transmitted or received frame on a given port. This is a small subset of what the hardware can do. A pseudo ifnet with the same name as the nexus driver (t4nex0 or t5nex0) will be created for tracing. The data delivered to this ifnet is an additional copy made inside the chip. Normal delivery to cxgbe<n> or cxl<n> will be made as usual. /* watch cxl0, which is the first port hanging off t5nex0. */ # cxgbetool t5nex0 tracer 0 tx0 (watch what cxl0 is transmitting) # cxgbetool t5nex0 tracer 1 rx0 (watch what cxl0 is receiving) # cxgbetool t5nex0 tracer list # tcpdump -i t5nex0 <== all that cxl0 sees and puts on the wire If you were doing TSO, a tcpdump on cxl0 may have shown you ~64K "frames" with no L3/L4 checksum but this will show you the frames that were actually transmitted. /* all done */ # cxgbetool t5nex0 tracer 0 disable # cxgbetool t5nex0 tracer 1 disable # cxgbetool t5nex0 tracer list # ifconfig t5nex0 destroy
2013-07-26 22:04:11 +00:00
sc->chan_map[pi->tx_chan] = i;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Maintain one ifmedia per physical port instead of one per Virtual Interface (VI). All autonomous VIs that share a port share the same media. MFC after: 1 week Sponsored by: Chelsio Communications
2017-08-28 21:44:25 +00:00
/* All VIs on this port share this media. */
ifmedia_init(&pi->media, IFM_IMASK, cxgbe_media_change,
cxgbe_media_status);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
PORT_LOCK(pi);
init_link_config(pi);
fixup_link_config(pi);
build_medialist(pi);
if (fixed_ifmedia(pi))
pi->flags |= FIXED_IFMEDIA;
PORT_UNLOCK(pi);
Add support for port unit wiring to cxgbe(4). - Add a bus_child_location_str method to the nexus drivers that prints out 'port=N' as the location string exported via devinfo and the '%location' sysctl node. - We can't use a bus_hint_device_unit to wire the unit numbers of devices with a fixed devclass as the device gets assigned a unit in make_device() before the device creator can set softc, etc. Instead, when adding a child device, use a helper function much like a bus_hint_device_unit method to look for wiring hints or to return -1 to let the system choose a unit number. This function requires an "at" hint for the port pointing to the nexus device and a "port" hint listing the port number. For example: hint.cxl.4.at="t5nex0" hint.cxl.4.port="0" wires cxl4 to the first port on the t5nex0 adapter. Requested by: gallatin MFC after: 2 months
2018-11-01 21:46:37 +00:00
pi->dev = device_add_child(dev, sc->names->ifnet_name,
t4_ifnet_unit(sc, pi));
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (pi->dev == NULL) {
device_printf(dev,
"failed to add device for port %d.\n", i);
rc = ENXIO;
goto done;
}
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
pi->vi[0].dev = pi->dev;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
device_set_softc(pi->dev, pi);
}
/*
* Interrupt type, # of interrupts, # of rx/tx queues, etc.
*/
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
nports = sc->params.nports;
cxgbe(4): Sanitize t4_num_vis during MOD_LOAD like all other t4_* tunables. Add num_vis to the intrs_and_queues structure as it affects the number of interrupts requested and queues created. In future cfg_itype_and_nqueues might lower it incrementally instead of going straight to 1 when enough interrupts aren't available. Sponsored by: Chelsio Communications
2017-11-16 01:33:53 +00:00
rc = cfg_itype_and_nqueues(sc, &iaq);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (rc != 0)
goto done; /* error message displayed already */
cxgbe(4): Sanitize t4_num_vis during MOD_LOAD like all other t4_* tunables. Add num_vis to the intrs_and_queues structure as it affects the number of interrupts requested and queues created. In future cfg_itype_and_nqueues might lower it incrementally instead of going straight to 1 when enough interrupts aren't available. Sponsored by: Chelsio Communications
2017-11-16 01:33:53 +00:00
num_vis = iaq.num_vis;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
sc->intr_type = iaq.intr_type;
sc->intr_count = iaq.nirq;
s = &sc->sge;
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
s->nrxq = nports * iaq.nrxq;
s->ntxq = nports * iaq.ntxq;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (num_vis > 1) {
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
s->nrxq += nports * (num_vis - 1) * iaq.nrxq_vi;
s->ntxq += nports * (num_vis - 1) * iaq.ntxq_vi;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
}
- Move all Ethernet specific items from sge_eq to sge_txq. sge_eq is now a suitable base for all kinds of egress queues. - Add control queues (sge_ctrlq) and allocate one of these per hardware channel. They can be used to program filters and steer traffic (and more). MFC after: 1 week
2011-04-19 22:08:28 +00:00
s->neq = s->ntxq + s->nrxq; /* the free list in an rxq is an eq */
cxgbe(4): Move all control queues to the adapter. There used to be one control queue per adapter (the mgmtq) that was initialized during adapter init and one per port that was initialized later during port init. This change moves all the control queues (one per port/channel) to the adapter so that they are initialized during adapter init and are available before any port is up. This allows the driver to issue ctrlq work requests over any channel without having to bring up any port. MFH: 2 weeks Sponsored by: Chelsio Communications
2018-08-11 21:10:08 +00:00
s->neq += nports; /* ctrl queues: 1 per port */
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
s->niq = s->nrxq + 1; /* 1 extra for firmware event queue */
cxgbe(4): Allocate offload Tx queues when a card has resources provisioned for NIC_ETHOFLD and the kernel has option RATELIMIT. It is possible to use the chip's offload queues for normal NIC Tx and not just TOE Tx. The difference is that these queues support out of order processing of work requests and have a per-"flowid" mechanism for tracking credits between the driver and hardware. This allows Tx for any number of flows bound to different rate limits to be submitted to a single Tx queue and the work requests for slow flows won't cause HOL blocking for the rest. Sponsored by: Chelsio Communications
2018-05-17 01:42:18 +00:00
#if defined(TCP_OFFLOAD) || defined(RATELIMIT)
if (is_offload(sc) || is_ethoffload(sc)) {
s->nofldtxq = nports * iaq.nofldtxq;
if (num_vis > 1)
s->nofldtxq += nports * (num_vis - 1) * iaq.nofldtxq_vi;
cxgbe(4): Fix s->neq miscalculation in r333698.
2018-05-17 06:04:50 +00:00
s->neq += s->nofldtxq;
cxgbe(4): Allocate offload Tx queues when a card has resources provisioned for NIC_ETHOFLD and the kernel has option RATELIMIT. It is possible to use the chip's offload queues for normal NIC Tx and not just TOE Tx. The difference is that these queues support out of order processing of work requests and have a per-"flowid" mechanism for tracking credits between the driver and hardware. This allows Tx for any number of flows bound to different rate limits to be submitted to a single Tx queue and the work requests for slow flows won't cause HOL blocking for the rest. Sponsored by: Chelsio Communications
2018-05-17 01:42:18 +00:00
s->ofld_txq = malloc(s->nofldtxq * sizeof(struct sge_wrq),
M_CXGBE, M_ZERO | M_WAITOK);
}
#endif
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
#ifdef TCP_OFFLOAD
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
if (is_offload(sc)) {
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
s->nofldrxq = nports * iaq.nofldrxq;
cxgbe(4): Fix s->neq miscalculation in r333698.
2018-05-17 06:04:50 +00:00
if (num_vis > 1)
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
s->nofldrxq += nports * (num_vis - 1) * iaq.nofldrxq_vi;
cxgbe(4): Fix s->neq miscalculation in r333698.
2018-05-17 06:04:50 +00:00
s->neq += s->nofldrxq; /* free list */
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
s->niq += s->nofldrxq;
s->ofld_rxq = malloc(s->nofldrxq * sizeof(struct sge_ofld_rxq),
M_CXGBE, M_ZERO | M_WAITOK);
}
#endif
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
#ifdef DEV_NETMAP
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
if (num_vis > 1) {
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
s->nnmrxq = nports * (num_vis - 1) * iaq.nnmrxq_vi;
s->nnmtxq = nports * (num_vis - 1) * iaq.nnmtxq_vi;
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
}
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
s->neq += s->nnmtxq + s->nnmrxq;
s->niq += s->nnmrxq;
s->nm_rxq = malloc(s->nnmrxq * sizeof(struct sge_nm_rxq),
M_CXGBE, M_ZERO | M_WAITOK);
s->nm_txq = malloc(s->nnmtxq * sizeof(struct sge_nm_txq),
M_CXGBE, M_ZERO | M_WAITOK);
#endif
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
s->ctrlq = malloc(nports * sizeof(struct sge_wrq), M_CXGBE,
- Move all Ethernet specific items from sge_eq to sge_txq. sge_eq is now a suitable base for all kinds of egress queues. - Add control queues (sge_ctrlq) and allocate one of these per hardware channel. They can be used to program filters and steer traffic (and more). MFC after: 1 week
2011-04-19 22:08:28 +00:00
M_ZERO | M_WAITOK);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
s->rxq = malloc(s->nrxq * sizeof(struct sge_rxq), M_CXGBE,
M_ZERO | M_WAITOK);
s->txq = malloc(s->ntxq * sizeof(struct sge_txq), M_CXGBE,
M_ZERO | M_WAITOK);
s->iqmap = malloc(s->niq * sizeof(struct sge_iq *), M_CXGBE,
M_ZERO | M_WAITOK);
s->eqmap = malloc(s->neq * sizeof(struct sge_eq *), M_CXGBE,
M_ZERO | M_WAITOK);
sc->irq = malloc(sc->intr_count * sizeof(struct irq), M_CXGBE,
M_ZERO | M_WAITOK);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
t4_init_l2t(sc, M_WAITOK);
cxgbe(4): Add code to deal with the chip's source MAC table (aka SMT). Submitted by: Krishnamraju Eraparaju @ Chelsio Sponsored by: Chelsio Communications
2018-05-31 21:31:08 +00:00
t4_init_smt(sc, M_WAITOK);
cxgbe(4): Support routines for Tx traffic scheduling. - Create a new file, t4_sched.c, and move all of the code related to traffic management from t4_main.c and t4_sge.c to this file. - Track both Channel Rate Limiter (ch_rl) and Class Rate Limiter (cl_rl) parameters in the PF driver. - Initialize all the cl_rl limiters with somewhat arbitrary default rates and provide routines to update them on the fly. - Provide routines to reserve and release traffic classes. MFC after: 1 month Sponsored by: Chelsio Communications
2017-05-02 20:38:10 +00:00
t4_init_tx_sched(sc);
cxgbe(4): Implement ifnet callbacks that deal with send tags. An etid (ethoffload tid) is allocated for a send tag and it acquires a reference on the traffic class that matches the send parameters associated with the tag. Sponsored by: Chelsio Communications
2018-05-18 06:09:15 +00:00
#ifdef RATELIMIT
t4_init_etid_table(sc);
#endif
Make most of the CLIP code conditional on #ifdef INET6. This fixes builds of kernels without INET6 such as LINT-NOINET6. Reported by: arybchik Reviewed by: np Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18384
2018-11-29 23:14:54 +00:00
#ifdef INET6
Move CLIP table handling out of TOM and into the base driver. - Store the clip table in 'struct adapter' instead of in the TOM softc. - Init the clip table during attach and teardown during detach. - While here, add a dev.<nexus>.<unit>.misc.clip sysctl to dump the CLIP table. This does mean that we update the clip table even if TOE is not enabled, but non-TOE things need the CLIP table anyway. Reviewed by: np, Krishnamraju Eraparaju @ Chelsio Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18010
2018-11-29 01:15:53 +00:00
t4_init_clip_table(sc);
Make most of the CLIP code conditional on #ifdef INET6. This fixes builds of kernels without INET6 such as LINT-NOINET6. Reported by: arybchik Reviewed by: np Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18384
2018-11-29 23:14:54 +00:00
#endif
Move the TLS key map into the adapter softc so non-TOE code can use it. Sponsored by: Chelsio Communications
2018-11-15 23:00:30 +00:00
if (sc->vres.key.size != 0)
sc->key_map = vmem_create("T4TLS key map", sc->vres.key.start,
Change the quantum for TLS key addresses to 32 bytes. The addresses passed when reading and writing keys are always shifted right by 5 as the memory locations are addressed in 32-byte chunks, so the quantum needs to be 32, not 8. MFC after: 1 month Sponsored by: Chelsio Communications
2018-11-15 23:10:46 +00:00
sc->vres.key.size, 32, 0, M_FIRSTFIT | M_WAITOK);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
/*
* Second pass over the ports. This time we know the number of rx and
* tx queues that each port should get.
*/
rqidx = tqidx = 0;
cxgbe(4): Allocate offload Tx queues when a card has resources provisioned for NIC_ETHOFLD and the kernel has option RATELIMIT. It is possible to use the chip's offload queues for normal NIC Tx and not just TOE Tx. The difference is that these queues support out of order processing of work requests and have a per-"flowid" mechanism for tracking credits between the driver and hardware. This allows Tx for any number of flows bound to different rate limits to be submitted to a single Tx queue and the work requests for slow flows won't cause HOL blocking for the rest. Sponsored by: Chelsio Communications
2018-05-17 01:42:18 +00:00
#if defined(TCP_OFFLOAD) || defined(RATELIMIT)
ofld_tqidx = 0;
#endif
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
#ifdef TCP_OFFLOAD
cxgbe(4): Allocate offload Tx queues when a card has resources provisioned for NIC_ETHOFLD and the kernel has option RATELIMIT. It is possible to use the chip's offload queues for normal NIC Tx and not just TOE Tx. The difference is that these queues support out of order processing of work requests and have a per-"flowid" mechanism for tracking credits between the driver and hardware. This allows Tx for any number of flows bound to different rate limits to be submitted to a single Tx queue and the work requests for slow flows won't cause HOL blocking for the rest. Sponsored by: Chelsio Communications
2018-05-17 01:42:18 +00:00
ofld_rqidx = 0;
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
#endif
#ifdef DEV_NETMAP
nm_rqidx = nm_tqidx = 0;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
#endif
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
for_each_port(sc, i) {
struct port_info *pi = sc->port[i];
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct vi_info *vi;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (pi == NULL)
continue;
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
pi->nvi = num_vis;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
for_each_vi(pi, j, vi) {
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
vi->pi = pi;
vi->qsize_rxq = t4_qsize_rxq;
vi->qsize_txq = t4_qsize_txq;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
vi->first_rxq = rqidx;
vi->first_txq = tqidx;
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
vi->tmr_idx = t4_tmr_idx;
vi->pktc_idx = t4_pktc_idx;
vi->nrxq = j == 0 ? iaq.nrxq : iaq.nrxq_vi;
vi->ntxq = j == 0 ? iaq.ntxq : iaq.ntxq_vi;
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
rqidx += vi->nrxq;
tqidx += vi->ntxq;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
if (j == 0 && vi->ntxq > 1)
cxgbe(4): Remove rsrv_noflowq from intrs_and_queues structure as it does not influence or get affected by the number of interrupts or queues. Sponsored by: Chelsio Communications
2017-11-16 02:42:37 +00:00
vi->rsrv_noflowq = t4_rsrv_noflowq ? 1 : 0;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
else
vi->rsrv_noflowq = 0;
cxgbe(4): Allocate offload Tx queues when a card has resources provisioned for NIC_ETHOFLD and the kernel has option RATELIMIT. It is possible to use the chip's offload queues for normal NIC Tx and not just TOE Tx. The difference is that these queues support out of order processing of work requests and have a per-"flowid" mechanism for tracking credits between the driver and hardware. This allows Tx for any number of flows bound to different rate limits to be submitted to a single Tx queue and the work requests for slow flows won't cause HOL blocking for the rest. Sponsored by: Chelsio Communications
2018-05-17 01:42:18 +00:00
#if defined(TCP_OFFLOAD) || defined(RATELIMIT)
vi->first_ofld_txq = ofld_tqidx;
vi->nofldtxq = j == 0 ? iaq.nofldtxq : iaq.nofldtxq_vi;
ofld_tqidx += vi->nofldtxq;
#endif
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
#ifdef TCP_OFFLOAD
cxgbe(4): Provide knobs to set the holdoff parameters of TOE rx queues separately from NIC rx queues instead of using the same parameters for both types of queues. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-10-05 07:18:16 +00:00
vi->ofld_tmr_idx = t4_tmr_idx_ofld;
vi->ofld_pktc_idx = t4_pktc_idx_ofld;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
vi->first_ofld_rxq = ofld_rqidx;
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
vi->nofldrxq = j == 0 ? iaq.nofldrxq : iaq.nofldrxq_vi;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
ofld_rqidx += vi->nofldrxq;
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
#endif
#ifdef DEV_NETMAP
if (j > 0) {
vi->first_nm_rxq = nm_rqidx;
vi->first_nm_txq = nm_tqidx;
vi->nnmrxq = iaq.nnmrxq_vi;
vi->nnmtxq = iaq.nnmtxq_vi;
nm_rqidx += vi->nnmrxq;
nm_tqidx += vi->nnmtxq;
}
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
#endif
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
rc = t4_setup_intr_handlers(sc);
Install interrupt handlers early, during attach, for the reason explained in r239913 by jhb. MFC after: 1 week
2012-09-13 09:18:13 +00:00
if (rc != 0) {
device_printf(dev,
"failed to setup interrupt handlers: %d\n", rc);
goto done;
}
Add support for child devices that aren't ports. Invoke any identify routines of child drivers during attach before attaching children, and delete any remaining devices after deleting ports. MFC after: 1 month Sponsored by: Chelsio Communications
2017-05-16 23:18:50 +00:00
rc = bus_generic_probe(dev);
if (rc != 0) {
device_printf(dev, "failed to probe child drivers: %d\n", rc);
goto done;
}
cxgbe(4): Verify that the driver accesses the firmware mailbox in a thread-safe manner. MFC after: 3 days
2017-08-28 03:13:16 +00:00
/*
* Ensure thread-safe mailbox access (in debug builds).
*
* So far this was the only thread accessing the mailbox but various
* ifnets and sysctls are about to be created and their handlers/ioctls
* will access the mailbox from different threads.
*/
sc->flags |= CHK_MBOX_ACCESS;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
rc = bus_generic_attach(dev);
if (rc != 0) {
device_printf(dev,
"failed to attach all child ports: %d\n", rc);
goto done;
}
device_printf(dev,
cxgbe: no need to display the per-lane GT/s rating of the pcie link. MFC after: 1 week
2015-06-01 03:24:39 +00:00
"PCIe gen%d x%d, %d ports, %d %s interrupt%s, %d eq, %d iq\n",
sc->params.pci.speed, sc->params.pci.width, sc->params.nports,
sc->intr_count, sc->intr_type == INTR_MSIX ? "MSI-X" :
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
(sc->intr_type == INTR_MSI ? "MSI" : "INTx"),
sc->intr_count > 1 ? "s" : "", sc->sge.neq, sc->sge.niq);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
t4_set_desc(sc);
Add a driver to create VF devices on Chelsio T4/T5 NICs. Chelsio NICs are a bit unique compared to some other NICs in that they expose different functionality on different physical functions. In particular, PF4 is used to manage the NIC interfaces ('t4nex' and 't5nex'). However, PF4 is not able to create VF devices. Instead, VFs are only supported by physical functions 0 through 3. This commit adds 't4iov' and 't5iov' drivers that attach to PF0-3. One extra wrinkle is that the iov devices cannot enable SR-IOV until the firwmare has been initialized by the main PF4 driver. To handle this case, a new t4_if kobj interface has been added to permit cross-calls between the PF drivers. The PF4 driver notifies sibling drivers when it is fully attached. It also requests sibling drivers to detach before it detaches. Sibling drivers query the PF4 driver during their attach routine to see if it is attached. If not, the sibling drivers defer their attach actions until the PF4 driver informs them it is attached. VF devices are associated with a single port on the NIC. VF devices created from PF0 are associated with the first port on the NIC, VFs from PF1 are associated with the second port, etc. VF devices can only be created from a PF device that has an associated port. Thus, on a 2-port card, VFs are only supported on PF0 and PF1. Reviewed by: np (earlier versions) MFC after: 1 month Sponsored by: Chelsio Communications
2016-07-22 22:46:41 +00:00
notify_siblings(dev, 0);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
done:
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
if (rc != 0 && sc->cdev) {
/* cdev was created and so cxgbetool works; recover that way. */
device_printf(dev,
"error during attach, adapter is now in recovery mode.\n");
rc = 0;
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (rc != 0)
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
t4_detach_common(dev);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
else
t4_sysctls(sc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
return (rc);
}
Add support for port unit wiring to cxgbe(4). - Add a bus_child_location_str method to the nexus drivers that prints out 'port=N' as the location string exported via devinfo and the '%location' sysctl node. - We can't use a bus_hint_device_unit to wire the unit numbers of devices with a fixed devclass as the device gets assigned a unit in make_device() before the device creator can set softc, etc. Instead, when adding a child device, use a helper function much like a bus_hint_device_unit method to look for wiring hints or to return -1 to let the system choose a unit number. This function requires an "at" hint for the port pointing to the nexus device and a "port" hint listing the port number. For example: hint.cxl.4.at="t5nex0" hint.cxl.4.port="0" wires cxl4 to the first port on the t5nex0 adapter. Requested by: gallatin MFC after: 2 months
2018-11-01 21:46:37 +00:00
static int
t4_child_location_str(device_t bus, device_t dev, char *buf, size_t buflen)
{
Don't assume all children of a nexus are ports. Specifically, ccr(4) devices are also children of cxgbe nexus devices. Rather than making assumptions about the child device's softc, walk the list of ports from the nexus' softc to determine if a child is a port in t4_child_location_str(). This fixes a panic when detaching a ccr device. Reviewed by: np MFC after: 1 week Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D19399
2019-02-28 22:10:19 +00:00
struct adapter *sc;
Add support for port unit wiring to cxgbe(4). - Add a bus_child_location_str method to the nexus drivers that prints out 'port=N' as the location string exported via devinfo and the '%location' sysctl node. - We can't use a bus_hint_device_unit to wire the unit numbers of devices with a fixed devclass as the device gets assigned a unit in make_device() before the device creator can set softc, etc. Instead, when adding a child device, use a helper function much like a bus_hint_device_unit method to look for wiring hints or to return -1 to let the system choose a unit number. This function requires an "at" hint for the port pointing to the nexus device and a "port" hint listing the port number. For example: hint.cxl.4.at="t5nex0" hint.cxl.4.port="0" wires cxl4 to the first port on the t5nex0 adapter. Requested by: gallatin MFC after: 2 months
2018-11-01 21:46:37 +00:00
struct port_info *pi;
Don't assume all children of a nexus are ports. Specifically, ccr(4) devices are also children of cxgbe nexus devices. Rather than making assumptions about the child device's softc, walk the list of ports from the nexus' softc to determine if a child is a port in t4_child_location_str(). This fixes a panic when detaching a ccr device. Reviewed by: np MFC after: 1 week Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D19399
2019-02-28 22:10:19 +00:00
int i;
Add support for port unit wiring to cxgbe(4). - Add a bus_child_location_str method to the nexus drivers that prints out 'port=N' as the location string exported via devinfo and the '%location' sysctl node. - We can't use a bus_hint_device_unit to wire the unit numbers of devices with a fixed devclass as the device gets assigned a unit in make_device() before the device creator can set softc, etc. Instead, when adding a child device, use a helper function much like a bus_hint_device_unit method to look for wiring hints or to return -1 to let the system choose a unit number. This function requires an "at" hint for the port pointing to the nexus device and a "port" hint listing the port number. For example: hint.cxl.4.at="t5nex0" hint.cxl.4.port="0" wires cxl4 to the first port on the t5nex0 adapter. Requested by: gallatin MFC after: 2 months
2018-11-01 21:46:37 +00:00
Don't assume all children of a nexus are ports. Specifically, ccr(4) devices are also children of cxgbe nexus devices. Rather than making assumptions about the child device's softc, walk the list of ports from the nexus' softc to determine if a child is a port in t4_child_location_str(). This fixes a panic when detaching a ccr device. Reviewed by: np MFC after: 1 week Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D19399
2019-02-28 22:10:19 +00:00
sc = device_get_softc(bus);
buf[0] = '\0';
for_each_port(sc, i) {
pi = sc->port[i];
if (pi != NULL && pi->dev == dev) {
snprintf(buf, buflen, "port=%d", pi->port_id);
break;
}
}
Add support for port unit wiring to cxgbe(4). - Add a bus_child_location_str method to the nexus drivers that prints out 'port=N' as the location string exported via devinfo and the '%location' sysctl node. - We can't use a bus_hint_device_unit to wire the unit numbers of devices with a fixed devclass as the device gets assigned a unit in make_device() before the device creator can set softc, etc. Instead, when adding a child device, use a helper function much like a bus_hint_device_unit method to look for wiring hints or to return -1 to let the system choose a unit number. This function requires an "at" hint for the port pointing to the nexus device and a "port" hint listing the port number. For example: hint.cxl.4.at="t5nex0" hint.cxl.4.port="0" wires cxl4 to the first port on the t5nex0 adapter. Requested by: gallatin MFC after: 2 months
2018-11-01 21:46:37 +00:00
return (0);
}
Add a driver to create VF devices on Chelsio T4/T5 NICs. Chelsio NICs are a bit unique compared to some other NICs in that they expose different functionality on different physical functions. In particular, PF4 is used to manage the NIC interfaces ('t4nex' and 't5nex'). However, PF4 is not able to create VF devices. Instead, VFs are only supported by physical functions 0 through 3. This commit adds 't4iov' and 't5iov' drivers that attach to PF0-3. One extra wrinkle is that the iov devices cannot enable SR-IOV until the firwmare has been initialized by the main PF4 driver. To handle this case, a new t4_if kobj interface has been added to permit cross-calls between the PF drivers. The PF4 driver notifies sibling drivers when it is fully attached. It also requests sibling drivers to detach before it detaches. Sibling drivers query the PF4 driver during their attach routine to see if it is attached. If not, the sibling drivers defer their attach actions until the PF4 driver informs them it is attached. VF devices are associated with a single port on the NIC. VF devices created from PF0 are associated with the first port on the NIC, VFs from PF1 are associated with the second port, etc. VF devices can only be created from a PF device that has an associated port. Thus, on a 2-port card, VFs are only supported on PF0 and PF1. Reviewed by: np (earlier versions) MFC after: 1 month Sponsored by: Chelsio Communications
2016-07-22 22:46:41 +00:00
static int
t4_ready(device_t dev)
{
struct adapter *sc;
sc = device_get_softc(dev);
if (sc->flags & FW_OK)
return (0);
return (ENXIO);
}
static int
Use the port device name for the iov device for Chelsio T4/T5 cards. Chelsio T4/T5 adapters are multifunction cards. The main driver uses physical function 4 (PF4). However, VF devices for SR-IOV are only supported on physical functions 0 through 3, where PF0 creates VFs tied to port 0, etc. The t4iov/t5iov driver was previously added to create VF devices for ports that are present on each adapter. This change uses the recently added pci_iov_attach_name() function to name the character device in /dev/iov after the associated port on the card (e.g. /dev/iov/cxl0 is used to create VFs that share the cxl0 port). With this in place, mark the t4iov/t5iov devices quiet to prevent them from cluttering dmesg. Reviewed by: rstone Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7402
2016-08-03 17:11:08 +00:00
t4_read_port_device(device_t dev, int port, device_t *child)
Add a driver to create VF devices on Chelsio T4/T5 NICs. Chelsio NICs are a bit unique compared to some other NICs in that they expose different functionality on different physical functions. In particular, PF4 is used to manage the NIC interfaces ('t4nex' and 't5nex'). However, PF4 is not able to create VF devices. Instead, VFs are only supported by physical functions 0 through 3. This commit adds 't4iov' and 't5iov' drivers that attach to PF0-3. One extra wrinkle is that the iov devices cannot enable SR-IOV until the firwmare has been initialized by the main PF4 driver. To handle this case, a new t4_if kobj interface has been added to permit cross-calls between the PF drivers. The PF4 driver notifies sibling drivers when it is fully attached. It also requests sibling drivers to detach before it detaches. Sibling drivers query the PF4 driver during their attach routine to see if it is attached. If not, the sibling drivers defer their attach actions until the PF4 driver informs them it is attached. VF devices are associated with a single port on the NIC. VF devices created from PF0 are associated with the first port on the NIC, VFs from PF1 are associated with the second port, etc. VF devices can only be created from a PF device that has an associated port. Thus, on a 2-port card, VFs are only supported on PF0 and PF1. Reviewed by: np (earlier versions) MFC after: 1 month Sponsored by: Chelsio Communications
2016-07-22 22:46:41 +00:00
{
struct adapter *sc;
struct port_info *pi;
sc = device_get_softc(dev);
if (port < 0 || port >= MAX_NPORTS)
return (EINVAL);
pi = sc->port[port];
if (pi == NULL || pi->dev == NULL)
return (ENXIO);
Use the port device name for the iov device for Chelsio T4/T5 cards. Chelsio T4/T5 adapters are multifunction cards. The main driver uses physical function 4 (PF4). However, VF devices for SR-IOV are only supported on physical functions 0 through 3, where PF0 creates VFs tied to port 0, etc. The t4iov/t5iov driver was previously added to create VF devices for ports that are present on each adapter. This change uses the recently added pci_iov_attach_name() function to name the character device in /dev/iov after the associated port on the card (e.g. /dev/iov/cxl0 is used to create VFs that share the cxl0 port). With this in place, mark the t4iov/t5iov devices quiet to prevent them from cluttering dmesg. Reviewed by: rstone Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7402
2016-08-03 17:11:08 +00:00
*child = pi->dev;
Add a driver to create VF devices on Chelsio T4/T5 NICs. Chelsio NICs are a bit unique compared to some other NICs in that they expose different functionality on different physical functions. In particular, PF4 is used to manage the NIC interfaces ('t4nex' and 't5nex'). However, PF4 is not able to create VF devices. Instead, VFs are only supported by physical functions 0 through 3. This commit adds 't4iov' and 't5iov' drivers that attach to PF0-3. One extra wrinkle is that the iov devices cannot enable SR-IOV until the firwmare has been initialized by the main PF4 driver. To handle this case, a new t4_if kobj interface has been added to permit cross-calls between the PF drivers. The PF4 driver notifies sibling drivers when it is fully attached. It also requests sibling drivers to detach before it detaches. Sibling drivers query the PF4 driver during their attach routine to see if it is attached. If not, the sibling drivers defer their attach actions until the PF4 driver informs them it is attached. VF devices are associated with a single port on the NIC. VF devices created from PF0 are associated with the first port on the NIC, VFs from PF1 are associated with the second port, etc. VF devices can only be created from a PF device that has an associated port. Thus, on a 2-port card, VFs are only supported on PF0 and PF1. Reviewed by: np (earlier versions) MFC after: 1 month Sponsored by: Chelsio Communications
2016-07-22 22:46:41 +00:00
return (0);
}
static int
notify_siblings(device_t dev, int detaching)
{
device_t sibling;
int error, i;
error = 0;
for (i = 0; i < PCI_FUNCMAX; i++) {
if (i == pci_get_function(dev))
continue;
sibling = pci_find_dbsf(pci_get_domain(dev), pci_get_bus(dev),
pci_get_slot(dev), i);
if (sibling == NULL || !device_is_attached(sibling))
continue;
if (detaching)
error = T4_DETACH_CHILD(sibling);
else
(void)T4_ATTACH_CHILD(sibling);
if (error)
break;
}
return (error);
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
/*
* Idempotent
*/
static int
t4_detach(device_t dev)
{
struct adapter *sc;
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
int rc;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
sc = device_get_softc(dev);
Add a driver to create VF devices on Chelsio T4/T5 NICs. Chelsio NICs are a bit unique compared to some other NICs in that they expose different functionality on different physical functions. In particular, PF4 is used to manage the NIC interfaces ('t4nex' and 't5nex'). However, PF4 is not able to create VF devices. Instead, VFs are only supported by physical functions 0 through 3. This commit adds 't4iov' and 't5iov' drivers that attach to PF0-3. One extra wrinkle is that the iov devices cannot enable SR-IOV until the firwmare has been initialized by the main PF4 driver. To handle this case, a new t4_if kobj interface has been added to permit cross-calls between the PF drivers. The PF4 driver notifies sibling drivers when it is fully attached. It also requests sibling drivers to detach before it detaches. Sibling drivers query the PF4 driver during their attach routine to see if it is attached. If not, the sibling drivers defer their attach actions until the PF4 driver informs them it is attached. VF devices are associated with a single port on the NIC. VF devices created from PF0 are associated with the first port on the NIC, VFs from PF1 are associated with the second port, etc. VF devices can only be created from a PF device that has an associated port. Thus, on a 2-port card, VFs are only supported on PF0 and PF1. Reviewed by: np (earlier versions) MFC after: 1 month Sponsored by: Chelsio Communications
2016-07-22 22:46:41 +00:00
rc = notify_siblings(dev, 1);
if (rc) {
device_printf(dev,
"failed to detach sibling devices: %d\n", rc);
return (rc);
}
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
return (t4_detach_common(dev));
}
int
t4_detach_common(device_t dev)
{
struct adapter *sc;
struct port_info *pi;
int i, rc;
sc = device_get_softc(dev);
cxgbe(4): Destroy the cdev before disabling interrupts in driver detach. Filter work requests are submitted in the nexus cdev's ioctl which then blocks waiting for a reply. If driver detach runs in this state and disables interrupts the ioctl will never complete and detach will hang in destroy_cdev. Sponsored by: Chelsio Communications
2018-05-01 14:59:38 +00:00
if (sc->cdev) {
destroy_dev(sc->cdev);
sc->cdev = NULL;
}
Remove adapters from t4_list earlier during detach. This ensures the clip task won't race with t4_destroy_clip_table. While here, make some mutex destroys unconditional since attach always initializes them. Reviewed by: np MFC after: 1 week Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D21952
2019-10-09 21:08:51 +00:00
sx_xlock(&t4_list_lock);
SLIST_REMOVE(&t4_list, sc, adapter, link);
sx_xunlock(&t4_list_lock);
cxgbe(4): Verify that the driver accesses the firmware mailbox in a thread-safe manner. MFC after: 3 days
2017-08-28 03:13:16 +00:00
sc->flags &= ~CHK_MBOX_ACCESS;
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
if (sc->flags & FULL_INIT_DONE) {
if (!(sc->flags & IS_VF))
t4_intr_disable(sc);
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
if (device_is_attached(dev)) {
rc = bus_generic_detach(dev);
if (rc) {
device_printf(dev,
"failed to detach child devices: %d\n", rc);
return (rc);
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Install interrupt handlers early, during attach, for the reason explained in r239913 by jhb. MFC after: 1 week
2012-09-13 09:18:13 +00:00
for (i = 0; i < sc->intr_count; i++)
t4_free_irq(sc, &sc->irq[i]);
cxgbe(4): Support routines for Tx traffic scheduling. - Create a new file, t4_sched.c, and move all of the code related to traffic management from t4_main.c and t4_sge.c to this file. - Track both Channel Rate Limiter (ch_rl) and Class Rate Limiter (cl_rl) parameters in the PF driver. - Initialize all the cl_rl limiters with somewhat arbitrary default rates and provide routines to update them on the fly. - Provide routines to reserve and release traffic classes. MFC after: 1 month Sponsored by: Chelsio Communications
2017-05-02 20:38:10 +00:00
if ((sc->flags & (IS_VF | FW_OK)) == FW_OK)
t4_free_tx_sched(sc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
for (i = 0; i < MAX_NPORTS; i++) {
pi = sc->port[i];
if (pi) {
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
t4_free_vi(sc, sc->mbox, sc->pf, 0, pi->vi[0].viid);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (pi->dev)
device_delete_child(dev, pi->dev);
mtx_destroy(&pi->pi_lock);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
free(pi->vi, M_CXGBE);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
free(pi, M_CXGBE);
}
}
Add support for child devices that aren't ports. Invoke any identify routines of child drivers during attach before attaching children, and delete any remaining devices after deleting ports. MFC after: 1 month Sponsored by: Chelsio Communications
2017-05-16 23:18:50 +00:00
device_delete_children(dev);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
if (sc->flags & FULL_INIT_DONE)
adapter_full_uninit(sc);
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
if ((sc->flags & (IS_VF | FW_OK)) == FW_OK)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
t4_fw_bye(sc, sc->mbox);
Do not over-allocate MSI interrupts for the case where each ingress queue has its own interrupt. If the exact number that we need is not a power of 2 and we're using MSI, then switch to interrupt multiplexing. While here, replace the magic numbers with something more readable. MFC after: 3 days
2011-03-24 01:03:01 +00:00
if (sc->intr_type == INTR_MSI || sc->intr_type == INTR_MSIX)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
pci_release_msi(dev);
if (sc->regs_res)
bus_release_resource(dev, SYS_RES_MEMORY, sc->regs_rid,
sc->regs_res);
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
if (sc->udbs_res)
bus_release_resource(dev, SYS_RES_MEMORY, sc->udbs_rid,
sc->udbs_res);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (sc->msix_res)
bus_release_resource(dev, SYS_RES_MEMORY, sc->msix_rid,
sc->msix_res);
L2 table code. This is enough to get the T4's switch + L2 rewrite filters working. (All other filters - switch without L2 info rewrite, steer, and drop - were already fully-functional). Some contrived examples of "switch" filters with L2 rewriting: # cxgbetool t4nex0 iport 0 dport 80 action switch vlan +9 eport 3 Intercept all packets received on physical port 0 with TCP port 80 as destination, insert a vlan tag with VID 9, and send them out of port 3. # cxgbetool t4nex0 sip 192.168.1.1/32 ivlan 5 action switch \ vlan =9 smac aa:bb:cc:dd:ee:ff eport 0 Intercept all packets (received on any port) with source IP address 192.168.1.1 and VLAN id 5, rewrite the VLAN id to 9, rewrite source mac to aa:bb:cc:dd:ee:ff, and send it out of port 0. MFC after: 1 week
2011-05-30 21:07:26 +00:00
if (sc->l2t)
t4_free_l2t(sc->l2t);
cxgbe(4): Add code to deal with the chip's source MAC table (aka SMT). Submitted by: Krishnamraju Eraparaju @ Chelsio Sponsored by: Chelsio Communications
2018-05-31 21:31:08 +00:00
if (sc->smt)
t4_free_smt(sc->smt);
cxgbe(4): Implement ifnet callbacks that deal with send tags. An etid (ethoffload tid) is allocated for a send tag and it acquires a reference on the traffic class that matches the send parameters associated with the tag. Sponsored by: Chelsio Communications
2018-05-18 06:09:15 +00:00
#ifdef RATELIMIT
t4_free_etid_table(sc);
#endif
Move the TLS key map into the adapter softc so non-TOE code can use it. Sponsored by: Chelsio Communications
2018-11-15 23:00:30 +00:00
if (sc->key_map)
vmem_destroy(sc->key_map);
Make most of the CLIP code conditional on #ifdef INET6. This fixes builds of kernels without INET6 such as LINT-NOINET6. Reported by: arybchik Reviewed by: np Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18384
2018-11-29 23:14:54 +00:00
#ifdef INET6
Move CLIP table handling out of TOM and into the base driver. - Store the clip table in 'struct adapter' instead of in the TOM softc. - Init the clip table during attach and teardown during detach. - While here, add a dev.<nexus>.<unit>.misc.clip sysctl to dump the CLIP table. This does mean that we update the clip table even if TOE is not enabled, but non-TOE things need the CLIP table anyway. Reviewed by: np, Krishnamraju Eraparaju @ Chelsio Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18010
2018-11-29 01:15:53 +00:00
t4_destroy_clip_table(sc);
Make most of the CLIP code conditional on #ifdef INET6. This fixes builds of kernels without INET6 such as LINT-NOINET6. Reported by: arybchik Reviewed by: np Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18384
2018-11-29 23:14:54 +00:00
#endif
L2 table code. This is enough to get the T4's switch + L2 rewrite filters working. (All other filters - switch without L2 info rewrite, steer, and drop - were already fully-functional). Some contrived examples of "switch" filters with L2 rewriting: # cxgbetool t4nex0 iport 0 dport 80 action switch vlan +9 eport 3 Intercept all packets received on physical port 0 with TCP port 80 as destination, insert a vlan tag with VID 9, and send them out of port 3. # cxgbetool t4nex0 sip 192.168.1.1/32 ivlan 5 action switch \ vlan =9 smac aa:bb:cc:dd:ee:ff eport 0 Intercept all packets (received on any port) with source IP address 192.168.1.1 and VLAN id 5, rewrite the VLAN id to 9, rewrite source mac to aa:bb:cc:dd:ee:ff, and send it out of port 0. MFC after: 1 week
2011-05-30 21:07:26 +00:00
cxgbe(4): Allocate offload Tx queues when a card has resources provisioned for NIC_ETHOFLD and the kernel has option RATELIMIT. It is possible to use the chip's offload queues for normal NIC Tx and not just TOE Tx. The difference is that these queues support out of order processing of work requests and have a per-"flowid" mechanism for tracking credits between the driver and hardware. This allows Tx for any number of flows bound to different rate limits to be submitted to a single Tx queue and the work requests for slow flows won't cause HOL blocking for the rest. Sponsored by: Chelsio Communications
2018-05-17 01:42:18 +00:00
#if defined(TCP_OFFLOAD) || defined(RATELIMIT)
free(sc->sge.ofld_txq, M_CXGBE);
#endif
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
#ifdef TCP_OFFLOAD
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
free(sc->sge.ofld_rxq, M_CXGBE);
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
#endif
#ifdef DEV_NETMAP
free(sc->sge.nm_rxq, M_CXGBE);
free(sc->sge.nm_txq, M_CXGBE);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
#endif
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
free(sc->irq, M_CXGBE);
free(sc->sge.rxq, M_CXGBE);
free(sc->sge.txq, M_CXGBE);
- Move all Ethernet specific items from sge_eq to sge_txq. sge_eq is now a suitable base for all kinds of egress queues. - Add control queues (sge_ctrlq) and allocate one of these per hardware channel. They can be used to program filters and steer traffic (and more). MFC after: 1 week
2011-04-19 22:08:28 +00:00
free(sc->sge.ctrlq, M_CXGBE);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
free(sc->sge.iqmap, M_CXGBE);
free(sc->sge.eqmap, M_CXGBE);
T4 packet filtering/steering. - Enable 5-tuple and every-packet lookup. - Setup the default filter mode to allow filtering/steering based on IP protocol, ingress port, inner VLAN ID, IP frag, FCoE, and MPS match type; all combined together. You can also filter based on MAC index, Ethernet type, IP TOS/IPv6 Traffic Class, and outer VLAN ID but you'll have to modify the default filter mode and exclude some of the match-fields in it. IPv4 and IPv6 SIP/DIP/SPORT/DPORT are always available in all filter rules. - Add driver ioctls to get/set the global filter mode. - Add driver ioctls to program and delete hardware filters. A couple of the "switch" actions that rewrite Ethernet and VLAN information and switch the packet out of another port may not work as the L2 code is not yet in place. Everything else, including all "drop" and "pass" rules with RSS or absolute qid, should work. Obtained from: Chelsio Communications
2011-05-05 02:04:56 +00:00
free(sc->tids.ftid_tab, M_CXGBE);
cxgbe(4): Add support for high priority filters on T6+. They have their own region in the TCAM starting with T6, unlike previous chips where they were in the same region as normal filters. These filters "hit" before anything else in the LE's lookup. The exact order is: a) High priority filters b) TOE's active region (TCAM and/or hash) c) Servers (TOE hw listeners) d) Normal filters MFC after: 1 week Sponsored by: Chelsio Communications
2018-08-09 14:19:47 +00:00
free(sc->tids.hpftid_tab, M_CXGBE);
cxgbe(4): Use two hashes instead of a table to keep track of hashfilters. Two because the driver needs to look up a hashfilter by its 4-tuple or tid. A couple of fixes while here: - Reject attempts to add duplicate hashfilters. - Do not assume that any part of the 4-tuple that isn't specified is 0. This makes it consistent with all other mandatory parameters that already require explicit user input. MFC after: 2 weeks Sponsored by: Chelsio Communications
2018-08-15 03:03:01 +00:00
free_hftid_hash(&sc->tids);
cxgbe(4): Add support for hash filters. These filters reside in the card's memory instead of its TCAM and can be configured via a new "hashfilter" subcommand in cxgbetool. Hash and normal TCAM filters can be used together. The hardware does an exact-match of packet fields for hash filters, unlike the masked match performed for TCAM filters. Any T5/T6 card with memory can support at least half a million hash filters. The sample config file with the driver configures 512K of these, it is possible to double this to 1 million+ in some cases. The chip does an exact-match of fields of incoming datagrams with hash filters and performs the action configured for the filter if it matches. The fields to match are specified in a "filter mask" in the firmware config file. The filter mask always includes the 5-tuple (sip, dip, sport, dport, ipproto). It can, optionally, also include any subset of the filter mode (see filterMode and filterMask in the firmware config file). For example: filterMode = fragmentation, mpshittype, protocol, vlan, port, fcoe filterMask = protocol, port, vlan Exact values of the 5-tuple, the physical port, and VLAN tag would have to be provided while setting up a hash filter with the chip configuration above. Hash filters support all actions supported by TCAM filters. A packet that hits a hash filter can be dropped, let through (with optional steering to a specific queue or RSS region), switched out of another port (with optional L2 rewrite of DMAC, SMAC, VLAN tag), or get NAT'ed. (Support for some of these will show up in the driver in a follow-up commit very shortly). Sponsored by: Chelsio Communications
2018-05-09 04:09:49 +00:00
free(sc->tids.atid_tab, M_CXGBE);
free(sc->tids.tid_tab, M_CXGBE);
Support for TLS offload of TOE connections on T6 adapters. The TOE engine in Chelsio T6 adapters supports offloading of TLS encryption and TCP segmentation for offloaded connections. Sockets using TLS are required to use a set of custom socket options to upload RX and TX keys to the NIC and to enable RX processing. Currently these socket options are implemented as TCP options in the vendor specific range. A patched OpenSSL library will be made available in a port / package for use with the TLS TOE support. TOE sockets can either offload both transmit and reception of TLS records or just transmit. TLS offload (both RX and TX) is enabled by setting the dev.t6nex.<x>.tls sysctl to 1 and requires TOE to be enabled on the relevant interface. Transmit offload can be used on any "normal" or TLS TOE socket by using the custom socket option to program a transmit key. This permits most TOE sockets to transparently offload TLS when applications use a patched SSL library (e.g. using LD_LIBRARY_PATH to request use of a patched OpenSSL library). Receive offload can only be used with TOE sockets using the TLS mode. The dev.t6nex.0.toe.tls_rx_ports sysctl can be set to a list of TCP port numbers. Any connection with either a local or remote port number in that list will be created as a TLS socket rather than a plain TOE socket. Note that although this sysctl accepts an arbitrary list of port numbers, the sysctl(8) tool is only able to set sysctl nodes to a single value. A TLS socket will hang without receiving data if used by an application that is not using a patched SSL library. Thus, the tls_rx_ports node should be used with care. For a server mostly concerned with offloading TLS transmit, this node is not needed as plain TOE sockets will fall back to software crypto when using an unpatched SSL library. New per-interface statistics nodes are added giving counts of TLS packets and payload bytes (payload bytes do not include TLS headers or authentication tags/MACs) offloaded via the TOE engine, e.g.: dev.cc.0.stats.rx_tls_octets: 149 dev.cc.0.stats.rx_tls_records: 13 dev.cc.0.stats.tx_tls_octets: 26501823 dev.cc.0.stats.tx_tls_records: 1620 TLS transmit work requests are constructed by a new variant of t4_push_frames() called t4_push_tls_records() in tom/t4_tls.c. TLS transmit work requests require a buffer containing IVs. If the IVs are too large to fit into the work request, a separate buffer is allocated when constructing a work request. This buffer is associated with the transmit descriptor and freed when the descriptor is ACKed by the adapter. Received TLS frames use two new CPL messages. The first message is a CPL_TLS_DATA containing the decryped payload of a single TLS record. The handler places the mbuf containing the received payload on an mbufq in the TOE pcb. The second message is a CPL_RX_TLS_CMP message which includes a copy of the TLS header and indicates if there were any errors. The handler for this message places the TLS header into the socket buffer followed by the saved mbuf with the payload data. Both of these handlers are contained in tom/t4_tls.c. A few routines were exposed from t4_cpl_io.c for use by t4_tls.c including send_rx_credits(), a new send_rx_modulate(), and t4_close_conn(). TLS keys for both transmit and receive are stored in onboard memory in the NIC in the "TLS keys" memory region. In some cases a TLS socket can hang with pending data available in the NIC that is not delivered to the host. As a workaround, TLS sockets are more aggressive about sending CPL_RX_DATA_ACK messages anytime that any data is read from a TLS socket. In addition, a fallback timer will periodically send CPL_RX_DATA_ACK messages to the NIC for connections that are still in the handshake phase. Once the connection has finished the handshake and programmed RX keys via the socket option, the timer is stopped. A new function select_ulp_mode() is used to determine what sub-mode a given TOE socket should use (plain TOE, DDP, or TLS). The existing set_tcpddp_ulp_mode() function has been renamed to set_ulp_mode() and handles initialization of TLS-specific state when necessary in addition to DDP-specific state. Since TLS sockets do not receive individual TCP segments but always receive full TLS records, they can receive more data than is available in the current window (e.g. if a 16k TLS record is received but the socket buffer is itself 16k). To cope with this, just drop the window to 0 when this happens, but track the overage and "eat" the overage as it is read from the socket buffer not opening the window (or adding rx_credits) for the overage bytes. Reviewed by: np (earlier version) Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D14529
2018-03-13 23:05:51 +00:00
free(sc->tt.tls_rx_ports, M_CXGBE);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
t4_destroy_dma_tag(sc);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
callout_drain(&sc->sfl_callout);
cxgbe(4): Add support for hash filters. These filters reside in the card's memory instead of its TCAM and can be configured via a new "hashfilter" subcommand in cxgbetool. Hash and normal TCAM filters can be used together. The hardware does an exact-match of packet fields for hash filters, unlike the masked match performed for TCAM filters. Any T5/T6 card with memory can support at least half a million hash filters. The sample config file with the driver configures 512K of these, it is possible to double this to 1 million+ in some cases. The chip does an exact-match of fields of incoming datagrams with hash filters and performs the action configured for the filter if it matches. The fields to match are specified in a "filter mask" in the firmware config file. The filter mask always includes the 5-tuple (sip, dip, sport, dport, ipproto). It can, optionally, also include any subset of the filter mode (see filterMode and filterMask in the firmware config file). For example: filterMode = fragmentation, mpshittype, protocol, vlan, port, fcoe filterMask = protocol, port, vlan Exact values of the 5-tuple, the physical port, and VLAN tag would have to be provided while setting up a hash filter with the chip configuration above. Hash filters support all actions supported by TCAM filters. A packet that hits a hash filter can be dropped, let through (with optional steering to a specific queue or RSS region), switched out of another port (with optional L2 rewrite of DMAC, SMAC, VLAN tag), or get NAT'ed. (Support for some of these will show up in the driver in a follow-up commit very shortly). Sponsored by: Chelsio Communications
2018-05-09 04:09:49 +00:00
if (mtx_initialized(&sc->tids.ftid_lock)) {
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
mtx_destroy(&sc->tids.ftid_lock);
cxgbe(4): Add support for hash filters. These filters reside in the card's memory instead of its TCAM and can be configured via a new "hashfilter" subcommand in cxgbetool. Hash and normal TCAM filters can be used together. The hardware does an exact-match of packet fields for hash filters, unlike the masked match performed for TCAM filters. Any T5/T6 card with memory can support at least half a million hash filters. The sample config file with the driver configures 512K of these, it is possible to double this to 1 million+ in some cases. The chip does an exact-match of fields of incoming datagrams with hash filters and performs the action configured for the filter if it matches. The fields to match are specified in a "filter mask" in the firmware config file. The filter mask always includes the 5-tuple (sip, dip, sport, dport, ipproto). It can, optionally, also include any subset of the filter mode (see filterMode and filterMask in the firmware config file). For example: filterMode = fragmentation, mpshittype, protocol, vlan, port, fcoe filterMask = protocol, port, vlan Exact values of the 5-tuple, the physical port, and VLAN tag would have to be provided while setting up a hash filter with the chip configuration above. Hash filters support all actions supported by TCAM filters. A packet that hits a hash filter can be dropped, let through (with optional steering to a specific queue or RSS region), switched out of another port (with optional L2 rewrite of DMAC, SMAC, VLAN tag), or get NAT'ed. (Support for some of these will show up in the driver in a follow-up commit very shortly). Sponsored by: Chelsio Communications
2018-05-09 04:09:49 +00:00
cv_destroy(&sc->tids.ftid_cv);
}
if (mtx_initialized(&sc->tids.atid_lock))
mtx_destroy(&sc->tids.atid_lock);
Add support for packet-sniffing tracers to cxgbe(4). This works with all T4 and T5 based cards and is useful for analyzing TSO, LRO, TOE, and for general purpose monitoring without tapping any cxgbe or cxl ifnet directly. Tracers on the T4/T5 chips provide access to Ethernet frames exactly as they were received from or transmitted on the wire. On transmit, a tracer will capture a frame after TSO segmentation, hw VLAN tag insertion, hw L3 & L4 checksum insertion, etc. It will also capture frames generated by the TCP offload engine (TOE traffic is normally invisible to the kernel). On receive, a tracer will capture a frame before hw VLAN extraction, runt filtering, other badness filtering, before the steering/drop/L2-rewrite filters or the TOE have had a go at it, and of course before sw LRO in the driver. There are 4 tracers on a chip. A tracer can trace only in one direction (tx or rx). For now cxgbetool will set up tracers to capture the first 128B of every transmitted or received frame on a given port. This is a small subset of what the hardware can do. A pseudo ifnet with the same name as the nexus driver (t4nex0 or t5nex0) will be created for tracing. The data delivered to this ifnet is an additional copy made inside the chip. Normal delivery to cxgbe<n> or cxl<n> will be made as usual. /* watch cxl0, which is the first port hanging off t5nex0. */ # cxgbetool t5nex0 tracer 0 tx0 (watch what cxl0 is transmitting) # cxgbetool t5nex0 tracer 1 rx0 (watch what cxl0 is receiving) # cxgbetool t5nex0 tracer list # tcpdump -i t5nex0 <== all that cxl0 sees and puts on the wire If you were doing TSO, a tcpdump on cxl0 may have shown you ~64K "frames" with no L3/L4 checksum but this will show you the frames that were actually transmitted. /* all done */ # cxgbetool t5nex0 tracer 0 disable # cxgbetool t5nex0 tracer 1 disable # cxgbetool t5nex0 tracer list # ifconfig t5nex0 destroy
2013-07-26 22:04:11 +00:00
if (mtx_initialized(&sc->ifp_lock))
mtx_destroy(&sc->ifp_lock);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Add support for Connection Offload Policy (aka COP). COP allows fine-grained control on whether to offload a TCP connection using t4_tom, and what settings to apply to a connection selected for offload. t4_tom must still be loaded and IFCAP_TOE must still be enabled for full TCP offload to take place on an interface. The difference is that IFCAP_TOE used to be the only knob and would enable TOE for all new connections on the inteface, but now the driver will also consult the COP, if any, before offloading to the hardware TOE. A policy is a plain text file with any number of rules, one per line. Each rule has a "match" part consisting of a socket-type (L = listen, A = active open, P = passive open, D = don't care) and a pcap-filter(7) expression, and a "settings" part that specifies whether to offload the connection or not and the parameters to use if so. The general format of a rule is: [socket-type] expr => settings Example. See cxgbetool(8) for more information. [L] ip && port http => offload [L] port 443 => !offload [L] port ssh => offload [P] src net 192.168/16 && dst port ssh => offload !nagle !timestamp cong newreno [P] dst port ssh => offload !nagle ecn cong tahoe [P] dst port http => offload [A] dst port 443 => offload tls [A] dst net 192.168/16 => offload !timestamp cong highspeed The driver processes the rules for each new listen, active open, or passive open and stops at the first match. There is an implicit rule at the end of every policy that prohibits offload when no rule in the policy matches: [D] all => !offload This is a reworked and expanded version of a patch submitted by Krishnamraju Eraparaju @ Chelsio. Sponsored by: Chelsio Communications
2018-04-14 19:07:56 +00:00
if (rw_initialized(&sc->policy_lock)) {
rw_destroy(&sc->policy_lock);
#ifdef TCP_OFFLOAD
if (sc->policy != NULL)
free_offload_policy(sc->policy);
#endif
}
cxgbe(4): Add general purpose routines that offer safe access to the chip's memory windows. Convert existing users of these windows to the new routines.
2016-03-10 06:15:31 +00:00
for (i = 0; i < NUM_MEMWIN; i++) {
struct memwin *mw = &sc->memwin[i];
if (rw_initialized(&mw->mw_lock))
rw_destroy(&mw->mw_lock);
}
Remove adapters from t4_list earlier during detach. This ensures the clip task won't race with t4_destroy_clip_table. While here, make some mutex destroys unconditional since attach always initializes them. Reviewed by: np MFC after: 1 week Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D21952
2019-10-09 21:08:51 +00:00
mtx_destroy(&sc->sfl_lock);
mtx_destroy(&sc->reg_lock);
mtx_destroy(&sc->sc_lock);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
bzero(sc, sizeof(*sc));
return (0);
}
static int
cxgbe_probe(device_t dev)
{
char buf[128];
struct port_info *pi = device_get_softc(dev);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
snprintf(buf, sizeof(buf), "port %d", pi->port_id);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
device_set_desc_copy(dev, buf);
return (BUS_PROBE_DEFAULT);
}
#define T4_CAP (IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_MTU | IFCAP_HWCSUM | \
IFCAP_VLAN_HWCSUM | IFCAP_TSO | IFCAP_JUMBO_MTU | IFCAP_LRO | \
cxgbe(4): Replace T4_PKT_TIMESTAMP with something slightly less hackish.
2018-08-18 04:23:51 +00:00
IFCAP_VLAN_HWTSO | IFCAP_LINKSTATE | IFCAP_HWCSUM_IPV6 | IFCAP_HWSTATS | \
Add support for IFCAP_NOMAP to cxgbe(4). Since cxgbe(4) uses sglist instead of bus_dma, this required updates to the code that generates scatter/gather lists for packets. Also, unmapped mbufs are always sent via DMA and never as immediate data in the payload of a work request. Submitted by: gallatin (earlier version) Reviewed by: gallatin, hselasky, rrs Discussed with: np Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20616
2019-06-29 00:52:21 +00:00
IFCAP_HWRXTSTMP | IFCAP_NOMAP)
cxgbe(4): support for IPv6 TSO and LRO. Submitted by: bz (this is a modified version of that patch)
2012-06-29 19:51:06 +00:00
#define T4_CAP_ENABLE (T4_CAP)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
static int
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
cxgbe_vi_attach(device_t dev, struct vi_info *vi)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
{
struct ifnet *ifp;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct sbuf *sb;
vi->xact_addr_filt = -1;
callout_init(&vi->tick, 1);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
/* Allocate an ifnet and set it up */
Track device's NUMA domain in ifnet & alloc ifnet from NUMA local memory This commit adds new if_alloc_domain() and if_alloc_dev() methods to allocate ifnets. When called with a domain on a NUMA machine, ifalloc_domain() will record the NUMA domain in the ifnet, and it will allocate the ifnet struct from memory which is local to that NUMA node. Similarly, if_alloc_dev() is a wrapper for if_alloc_domain which uses a driver supplied device_t to call ifalloc_domain() with the appropriate domain. Note that the new if_numa_domain field fits in an alignment pad in struct ifnet, and so does not alter the size of the structure. Reviewed by: glebius, kib, markj Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D19930
2019-04-22 19:24:21 +00:00
ifp = if_alloc_dev(IFT_ETHER, dev);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (ifp == NULL) {
device_printf(dev, "Cannot allocate ifnet\n");
return (ENOMEM);
}
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
vi->ifp = ifp;
ifp->if_softc = vi;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if_initname(ifp, device_get_name(dev), device_get_unit(dev));
ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
ifp->if_init = cxgbe_init;
ifp->if_ioctl = cxgbe_ioctl;
ifp->if_transmit = cxgbe_transmit;
ifp->if_qflush = cxgbe_qflush;
cxgbe(4): implement if_get_counter.
2014-09-27 05:50:31 +00:00
ifp->if_get_counter = cxgbe_get_counter;
cxgbe(4): Implement ifnet callbacks that deal with send tags. An etid (ethoffload tid) is allocated for a send tag and it acquires a reference on the traffic class that matches the send parameters associated with the tag. Sponsored by: Chelsio Communications
2018-05-18 06:09:15 +00:00
#ifdef RATELIMIT
ifp->if_snd_tag_alloc = cxgbe_snd_tag_alloc;
ifp->if_snd_tag_modify = cxgbe_snd_tag_modify;
ifp->if_snd_tag_query = cxgbe_snd_tag_query;
ifp->if_snd_tag_free = cxgbe_snd_tag_free;
This adds the third step in getting BBR into the tree. BBR and an updated rack depend on having access to the new ratelimit api in this commit. Sponsored by: Netflix Inc. Differential Revision: https://reviews.freebsd.org/D20953
2019-08-01 14:17:31 +00:00
ifp->if_ratelimit_query = cxgbe_ratelimit_query;
cxgbe(4): Implement ifnet callbacks that deal with send tags. An etid (ethoffload tid) is allocated for a send tag and it acquires a reference on the traffic class that matches the send parameters associated with the tag. Sponsored by: Chelsio Communications
2018-05-18 06:09:15 +00:00
#endif
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
ifp->if_capabilities = T4_CAP;
cxgbe(4): Enable TXRTLMT by default when the feature is available in the kernel (options RATELIMIT) and provisioned in the driver's configuration file (nethofld > 0). Submitted by: gallatin@ Approved by: re@ (kib@)
2018-09-18 21:34:37 +00:00
ifp->if_capenable = T4_CAP_ENABLE;
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
#ifdef TCP_OFFLOAD
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (vi->nofldrxq != 0)
cxgbe/tom: List IFCAP_TOE6 as supported now that all the required pieces are in place. You still have to enable it explicitly, after loading the t4_tom KLD.
2013-01-26 01:06:27 +00:00
ifp->if_capabilities |= IFCAP_TOE;
cxgbe(4): netmap does not set IFCAP_NETMAP in an ifnet's if_capabilities any more (since r307394). Do it in the driver instead. MFC after: 1 week
2016-12-09 02:21:27 +00:00
#endif
cxgbe(4): Report IFCAP_TXRTLMT to kernels built with RATELIMIT if the firmware has provisioned resources for this feature. Sponsored by: Chelsio Communications
2018-05-24 10:55:26 +00:00
#ifdef RATELIMIT
cxgbe(4): Enable TXRTLMT by default when the feature is available in the kernel (options RATELIMIT) and provisioned in the driver's configuration file (nethofld > 0). Submitted by: gallatin@ Approved by: re@ (kib@)
2018-09-18 21:34:37 +00:00
if (is_ethoffload(vi->pi->adapter) && vi->nofldtxq != 0) {
cxgbe(4): Report IFCAP_TXRTLMT to kernels built with RATELIMIT if the firmware has provisioned resources for this feature. Sponsored by: Chelsio Communications
2018-05-24 10:55:26 +00:00
ifp->if_capabilities |= IFCAP_TXRTLMT;
cxgbe(4): Enable TXRTLMT by default when the feature is available in the kernel (options RATELIMIT) and provisioned in the driver's configuration file (nethofld > 0). Submitted by: gallatin@ Approved by: re@ (kib@)
2018-09-18 21:34:37 +00:00
ifp->if_capenable |= IFCAP_TXRTLMT;
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
#endif
cxgbe(4): support for IPv6 hardware checksumming (rx and tx).
2012-06-29 16:50:52 +00:00
ifp->if_hwassist = CSUM_TCP | CSUM_UDP | CSUM_IP | CSUM_TSO |
CSUM_UDP_IPV6 | CSUM_TCP_IPV6;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): improve the accuracy of various TSO limits reported to the kernel. Sponsored by: Chelsio Communications
2018-10-22 23:57:59 +00:00
ifp->if_hw_tsomax = IP_MAXPACKET;
ifp->if_hw_tsomaxsegcount = TX_SGL_SEGS_TSO;
#ifdef RATELIMIT
if (is_ethoffload(vi->pi->adapter) && vi->nofldtxq != 0)
ifp->if_hw_tsomaxsegcount = TX_SGL_SEGS_EO_TSO;
#endif
cxgbe(4): Report a reasonable non-zero if_hw_tsomaxsegsize to the kernel. This reverts an accidental change that snuck in with r339628. Sponsored by: Chelsio Communications
2018-10-31 18:30:17 +00:00
ifp->if_hw_tsomaxsegsize = 65536;
cxgbe(4): explicitly set various if_hw_tso* values. MFC after: 3 days
2014-09-26 22:21:02 +00:00
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
ether_ifattach(ifp, vi->hw_addr);
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
#ifdef DEV_NETMAP
cxgbe: revert r309725 After the fix contained in r341144, cxgbe does not need anymore to set the IFCAP_NETMAP flag manually. Reviewed by: np Approved by: gnn (mentor) MFC after: 3 days Differential Revision: https://reviews.freebsd.org/D17987
2018-11-28 15:29:58 +00:00
if (vi->nnmrxq != 0)
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
cxgbe_nm_attach(vi);
#endif
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
sb = sbuf_new_auto();
sbuf_printf(sb, "%d txq, %d rxq (NIC)", vi->ntxq, vi->nrxq);
cxgbe(4): Report IFCAP_TXRTLMT to kernels built with RATELIMIT if the firmware has provisioned resources for this feature. Sponsored by: Chelsio Communications
2018-05-24 10:55:26 +00:00
#if defined(TCP_OFFLOAD) || defined(RATELIMIT)
switch (ifp->if_capabilities & (IFCAP_TOE | IFCAP_TXRTLMT)) {
case IFCAP_TOE:
sbuf_printf(sb, "; %d txq (TOE)", vi->nofldtxq);
break;
case IFCAP_TOE | IFCAP_TXRTLMT:
sbuf_printf(sb, "; %d txq (TOE/ETHOFLD)", vi->nofldtxq);
break;
case IFCAP_TXRTLMT:
sbuf_printf(sb, "; %d txq (ETHOFLD)", vi->nofldtxq);
break;
}
#endif
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
#ifdef TCP_OFFLOAD
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (ifp->if_capabilities & IFCAP_TOE)
cxgbe(4): Report IFCAP_TXRTLMT to kernels built with RATELIMIT if the firmware has provisioned resources for this feature. Sponsored by: Chelsio Communications
2018-05-24 10:55:26 +00:00
sbuf_printf(sb, ", %d rxq (TOE)", vi->nofldrxq);
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
#endif
#ifdef DEV_NETMAP
if (ifp->if_capabilities & IFCAP_NETMAP)
sbuf_printf(sb, "; %d txq, %d rxq (netmap)",
vi->nnmtxq, vi->nnmrxq);
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
#endif
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
sbuf_finish(sb);
device_printf(dev, "%s\n", sbuf_data(sb));
sbuf_delete(sb);
vi_sysctls(vi);
return (0);
}
static int
cxgbe_attach(device_t dev)
{
struct port_info *pi = device_get_softc(dev);
cxgbe(4): Attach to cards with the Terminator 6 ASIC. T6 cards will come up as 't6nex' nexus devices with 'cc' ports hanging off them. The T6 firmware and configuration files will be added as soon as they are released. For now the driver will try to work with whatever firmware and configuration is on the card's flash. Sponsored by: Chelsio Communications
2016-09-16 00:08:37 +00:00
struct adapter *sc = pi->adapter;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct vi_info *vi;
int i, rc;
callout_init_mtx(&pi->tick, &pi->pi_lock, 0);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
rc = cxgbe_vi_attach(dev, &pi->vi[0]);
if (rc)
return (rc);
for_each_vi(pi, i, vi) {
if (i == 0)
continue;
cxgbe(4): Attach to cards with the Terminator 6 ASIC. T6 cards will come up as 't6nex' nexus devices with 'cc' ports hanging off them. The T6 firmware and configuration files will be added as soon as they are released. For now the driver will try to work with whatever firmware and configuration is on the card's flash. Sponsored by: Chelsio Communications
2016-09-16 00:08:37 +00:00
vi->dev = device_add_child(dev, sc->names->vi_ifnet_name, -1);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (vi->dev == NULL) {
device_printf(dev, "failed to add VI %d\n", i);
continue;
}
device_set_softc(vi->dev, vi);
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe_sysctls(pi);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
bus_generic_attach(dev);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
return (0);
}
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
static void
cxgbe_vi_detach(struct vi_info *vi)
{
struct ifnet *ifp = vi->ifp;
ether_ifdetach(ifp);
/* Let detach proceed even if these fail. */
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
#ifdef DEV_NETMAP
if (ifp->if_capabilities & IFCAP_NETMAP)
cxgbe_nm_detach(vi);
#endif
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
cxgbe_uninit_synchronized(vi);
callout_drain(&vi->tick);
vi_full_uninit(vi);
if_free(vi->ifp);
vi->ifp = NULL;
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
static int
cxgbe_detach(device_t dev)
{
struct port_info *pi = device_get_softc(dev);
struct adapter *sc = pi->adapter;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
int rc;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
/* Detach the extra VIs first. */
rc = bus_generic_detach(dev);
if (rc)
return (rc);
device_delete_children(dev);
doom_vi(sc, &pi->vi[0]);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Add support for packet-sniffing tracers to cxgbe(4). This works with all T4 and T5 based cards and is useful for analyzing TSO, LRO, TOE, and for general purpose monitoring without tapping any cxgbe or cxl ifnet directly. Tracers on the T4/T5 chips provide access to Ethernet frames exactly as they were received from or transmitted on the wire. On transmit, a tracer will capture a frame after TSO segmentation, hw VLAN tag insertion, hw L3 & L4 checksum insertion, etc. It will also capture frames generated by the TCP offload engine (TOE traffic is normally invisible to the kernel). On receive, a tracer will capture a frame before hw VLAN extraction, runt filtering, other badness filtering, before the steering/drop/L2-rewrite filters or the TOE have had a go at it, and of course before sw LRO in the driver. There are 4 tracers on a chip. A tracer can trace only in one direction (tx or rx). For now cxgbetool will set up tracers to capture the first 128B of every transmitted or received frame on a given port. This is a small subset of what the hardware can do. A pseudo ifnet with the same name as the nexus driver (t4nex0 or t5nex0) will be created for tracing. The data delivered to this ifnet is an additional copy made inside the chip. Normal delivery to cxgbe<n> or cxl<n> will be made as usual. /* watch cxl0, which is the first port hanging off t5nex0. */ # cxgbetool t5nex0 tracer 0 tx0 (watch what cxl0 is transmitting) # cxgbetool t5nex0 tracer 1 rx0 (watch what cxl0 is receiving) # cxgbetool t5nex0 tracer list # tcpdump -i t5nex0 <== all that cxl0 sees and puts on the wire If you were doing TSO, a tcpdump on cxl0 may have shown you ~64K "frames" with no L3/L4 checksum but this will show you the frames that were actually transmitted. /* all done */ # cxgbetool t5nex0 tracer 0 disable # cxgbetool t5nex0 tracer 1 disable # cxgbetool t5nex0 tracer list # ifconfig t5nex0 destroy
2013-07-26 22:04:11 +00:00
if (pi->flags & HAS_TRACEQ) {
sc->traceq = -1; /* cloner should not create ifnet */
t4_tracer_port_detach(sc);
}
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
cxgbe_vi_detach(&pi->vi[0]);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
callout_drain(&pi->tick);
cxgbe(4): Maintain one ifmedia per physical port instead of one per Virtual Interface (VI). All autonomous VIs that share a port share the same media. MFC after: 1 week Sponsored by: Chelsio Communications
2017-08-28 21:44:25 +00:00
ifmedia_removeall(&pi->media);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
end_synchronized_op(sc, 0);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
return (0);
}
static void
cxgbe_init(void *arg)
{
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct vi_info *vi = arg;
struct adapter *sc = vi->pi->adapter;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (begin_synchronized_op(sc, vi, SLEEP_OK | INTR_OK, "t4init") != 0)
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
return;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
cxgbe_init_synchronized(vi);
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
end_synchronized_op(sc, 0);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
static int
cxgbe_ioctl(struct ifnet *ifp, unsigned long cmd, caddr_t data)
{
cxgbe(4): Catch up with recent changes in the kernel -- it no longer holds non-sleepable locks around any of the driver ioctls. Sponsored by: Chelsio Communications
2018-06-14 01:27:35 +00:00
int rc = 0, mtu, flags;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct vi_info *vi = ifp->if_softc;
cxgbe(4): Maintain one ifmedia per physical port instead of one per Virtual Interface (VI). All autonomous VIs that share a port share the same media. MFC after: 1 week Sponsored by: Chelsio Communications
2017-08-28 21:44:25 +00:00
struct port_info *pi = vi->pi;
struct adapter *sc = pi->adapter;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
struct ifreq *ifr = (struct ifreq *)data;
uint32_t mask;
switch (cmd) {
case SIOCSIFMTU:
mtu = ifr->ifr_mtu;
cxgbe(4): Allow the interface MTU to be set as high as the actual hardware limit. Submitted by: jpaetzel@ Differential Revision: https://reviews.freebsd.org/D8237
2016-10-13 19:40:21 +00:00
if (mtu < ETHERMIN || mtu > MAX_MTU)
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
return (EINVAL);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
rc = begin_synchronized_op(sc, vi, SLEEP_OK | INTR_OK, "t4mtu");
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
if (rc)
return (rc);
ifp->if_mtu = mtu;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (vi->flags & VI_INIT_DONE) {
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
t4_update_fl_bufsize(ifp);
- Make note of interface MTU change if the rx queues exist, and not just when the interface is up. - Add a tunable to control the TOE's rx coalesce feature (enabled by default as it always has been). Consider the interface MTU or the coalesce size when deciding which cluster zone to use to fill the offload rx queue's free list. The tunable is: dev.{t4nex,t5nex}.<N>.toe.rx_coalesce MFC after: 1 day
2013-07-04 21:19:01 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING)
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
rc = update_mac_settings(ifp, XGMAC_MTU);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
end_synchronized_op(sc, 0);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
break;
case SIOCSIFFLAGS:
cxgbe(4): Catch up with recent changes in the kernel -- it no longer holds non-sleepable locks around any of the driver ioctls. Sponsored by: Chelsio Communications
2018-06-14 01:27:35 +00:00
rc = begin_synchronized_op(sc, vi, SLEEP_OK | INTR_OK, "t4flg");
if (rc)
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
return (rc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (ifp->if_flags & IFF_UP) {
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
flags = vi->if_flags;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if ((ifp->if_flags ^ flags) &
(IFF_PROMISC | IFF_ALLMULTI)) {
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
rc = update_mac_settings(ifp,
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
XGMAC_PROMISC | XGMAC_ALLMULTI);
}
cxgbe(4): Do not run any sleepable code in the SIOCSIFFLAGS handler when IFF_PROMISC or IFF_ALLMULTI is being flipped. bpf(4) holds its global mutex around ifpromisc in at least the bpf_dtor path. MFC after: 3 days
2014-08-04 22:32:16 +00:00
} else {
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
rc = cxgbe_init_synchronized(vi);
cxgbe(4): Do not run any sleepable code in the SIOCSIFFLAGS handler when IFF_PROMISC or IFF_ALLMULTI is being flipped. bpf(4) holds its global mutex around ifpromisc in at least the bpf_dtor path. MFC after: 3 days
2014-08-04 22:32:16 +00:00
}
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
vi->if_flags = ifp->if_flags;
cxgbe(4): Do not run any sleepable code in the SIOCSIFFLAGS handler when IFF_PROMISC or IFF_ALLMULTI is being flipped. bpf(4) holds its global mutex around ifpromisc in at least the bpf_dtor path. MFC after: 3 days
2014-08-04 22:32:16 +00:00
} else if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
rc = cxgbe_uninit_synchronized(vi);
cxgbe(4): Do not run any sleepable code in the SIOCSIFFLAGS handler when IFF_PROMISC or IFF_ALLMULTI is being flipped. bpf(4) holds its global mutex around ifpromisc in at least the bpf_dtor path. MFC after: 3 days
2014-08-04 22:32:16 +00:00
}
cxgbe(4): Catch up with recent changes in the kernel -- it no longer holds non-sleepable locks around any of the driver ioctls. Sponsored by: Chelsio Communications
2018-06-14 01:27:35 +00:00
end_synchronized_op(sc, 0);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
break;
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
case SIOCADDMULTI:
cxgbe(4): Catch up with recent changes in the kernel -- it no longer holds non-sleepable locks around any of the driver ioctls. Sponsored by: Chelsio Communications
2018-06-14 01:27:35 +00:00
case SIOCDELMULTI:
rc = begin_synchronized_op(sc, vi, SLEEP_OK | INTR_OK, "t4multi");
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (rc)
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
return (rc);
if (ifp->if_drv_flags & IFF_DRV_RUNNING)
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
rc = update_mac_settings(ifp, XGMAC_MCADDRS);
cxgbe(4): Catch up with recent changes in the kernel -- it no longer holds non-sleepable locks around any of the driver ioctls. Sponsored by: Chelsio Communications
2018-06-14 01:27:35 +00:00
end_synchronized_op(sc, 0);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
break;
case SIOCSIFCAP:
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
rc = begin_synchronized_op(sc, vi, SLEEP_OK | INTR_OK, "t4cap");
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (rc)
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
return (rc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
mask = ifr->ifr_reqcap ^ ifp->if_capenable;
if (mask & IFCAP_TXCSUM) {
ifp->if_capenable ^= IFCAP_TXCSUM;
ifp->if_hwassist ^= (CSUM_TCP | CSUM_UDP | CSUM_IP);
- Assign (don't OR) the CSUM_XXX bits to csum_flags in the rx checksum code. - Fix TSO/TSO4 mixup. - Add IFCAP_LINKSTATE to the available/enabled capabilities.
2012-06-30 02:05:09 +00:00
if (IFCAP_TSO4 & ifp->if_capenable &&
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
!(IFCAP_TXCSUM & ifp->if_capenable)) {
cxgbe(4): support for IPv6 hardware checksumming (rx and tx).
2012-06-29 16:50:52 +00:00
ifp->if_capenable &= ~IFCAP_TSO4;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if_printf(ifp,
- Assign (don't OR) the CSUM_XXX bits to csum_flags in the rx checksum code. - Fix TSO/TSO4 mixup. - Add IFCAP_LINKSTATE to the available/enabled capabilities.
2012-06-30 02:05:09 +00:00
"tso4 disabled due to -txcsum.\n");
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
}
cxgbe(4): support for IPv6 hardware checksumming (rx and tx).
2012-06-29 16:50:52 +00:00
if (mask & IFCAP_TXCSUM_IPV6) {
ifp->if_capenable ^= IFCAP_TXCSUM_IPV6;
ifp->if_hwassist ^= (CSUM_UDP_IPV6 | CSUM_TCP_IPV6);
if (IFCAP_TSO6 & ifp->if_capenable &&
!(IFCAP_TXCSUM_IPV6 & ifp->if_capenable)) {
ifp->if_capenable &= ~IFCAP_TSO6;
if_printf(ifp,
"tso6 disabled due to -txcsum6.\n");
}
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (mask & IFCAP_RXCSUM)
ifp->if_capenable ^= IFCAP_RXCSUM;
cxgbe(4): support for IPv6 hardware checksumming (rx and tx).
2012-06-29 16:50:52 +00:00
if (mask & IFCAP_RXCSUM_IPV6)
ifp->if_capenable ^= IFCAP_RXCSUM_IPV6;
/*
* Note that we leave CSUM_TSO alone (it is always set). The
* kernel takes both IFCAP_TSOx and CSUM_TSO into account before
* sending a TSO request our way, so it's sufficient to toggle
* IFCAP_TSOx only.
*/
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (mask & IFCAP_TSO4) {
cxgbe(4): support for IPv6 hardware checksumming (rx and tx).
2012-06-29 16:50:52 +00:00
if (!(IFCAP_TSO4 & ifp->if_capenable) &&
!(IFCAP_TXCSUM & ifp->if_capenable)) {
if_printf(ifp, "enable txcsum first.\n");
rc = EAGAIN;
goto fail;
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
ifp->if_capenable ^= IFCAP_TSO4;
cxgbe(4): support for IPv6 hardware checksumming (rx and tx).
2012-06-29 16:50:52 +00:00
}
if (mask & IFCAP_TSO6) {
if (!(IFCAP_TSO6 & ifp->if_capenable) &&
!(IFCAP_TXCSUM_IPV6 & ifp->if_capenable)) {
if_printf(ifp, "enable txcsum6 first.\n");
rc = EAGAIN;
goto fail;
}
ifp->if_capenable ^= IFCAP_TSO6;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
if (mask & IFCAP_LRO) {
cxgbe(4): support for IPv6 TSO and LRO. Submitted by: bz (this is a modified version of that patch)
2012-06-29 19:51:06 +00:00
#if defined(INET) || defined(INET6)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
int i;
struct sge_rxq *rxq;
ifp->if_capenable ^= IFCAP_LRO;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
for_each_rxq(vi, i, rxq) {
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (ifp->if_capenable & IFCAP_LRO)
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
rxq->iq.flags |= IQ_LRO_ENABLED;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
else
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
rxq->iq.flags &= ~IQ_LRO_ENABLED;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
#endif
}
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
#ifdef TCP_OFFLOAD
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
if (mask & IFCAP_TOE) {
int enable = (ifp->if_capenable ^ mask) & IFCAP_TOE;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
rc = toe_capability(vi, enable);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
if (rc != 0)
goto fail;
ifp->if_capenable ^= mask;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
#endif
if (mask & IFCAP_VLAN_HWTAGGING) {
ifp->if_capenable ^= IFCAP_VLAN_HWTAGGING;
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING)
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
rc = update_mac_settings(ifp, XGMAC_VLANEX);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
if (mask & IFCAP_VLAN_MTU) {
ifp->if_capenable ^= IFCAP_VLAN_MTU;
/* Need to find out how to disable auto-mtu-inflation */
}
if (mask & IFCAP_VLAN_HWTSO)
ifp->if_capenable ^= IFCAP_VLAN_HWTSO;
if (mask & IFCAP_VLAN_HWCSUM)
ifp->if_capenable ^= IFCAP_VLAN_HWCSUM;
cxgbe(4): Report IFCAP_TXRTLMT to kernels built with RATELIMIT if the firmware has provisioned resources for this feature. Sponsored by: Chelsio Communications
2018-05-24 10:55:26 +00:00
#ifdef RATELIMIT
if (mask & IFCAP_TXRTLMT)
ifp->if_capenable ^= IFCAP_TXRTLMT;
#endif
cxgbe(4): Replace T4_PKT_TIMESTAMP with something slightly less hackish.
2018-08-18 04:23:51 +00:00
if (mask & IFCAP_HWRXTSTMP) {
int i;
struct sge_rxq *rxq;
ifp->if_capenable ^= IFCAP_HWRXTSTMP;
for_each_rxq(vi, i, rxq) {
if (ifp->if_capenable & IFCAP_HWRXTSTMP)
rxq->iq.flags |= IQ_RX_TIMESTAMP;
else
rxq->iq.flags &= ~IQ_RX_TIMESTAMP;
}
}
Add support for IFCAP_NOMAP to cxgbe(4). Since cxgbe(4) uses sglist instead of bus_dma, this required updates to the code that generates scatter/gather lists for packets. Also, unmapped mbufs are always sent via DMA and never as immediate data in the payload of a work request. Submitted by: gallatin (earlier version) Reviewed by: gallatin, hselasky, rrs Discussed with: np Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20616
2019-06-29 00:52:21 +00:00
if (mask & IFCAP_NOMAP)
ifp->if_capenable ^= IFCAP_NOMAP;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
#ifdef VLAN_CAPABILITIES
VLAN_CAPABILITIES(ifp);
#endif
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
fail:
end_synchronized_op(sc, 0);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
break;
case SIOCSIFMEDIA:
case SIOCGIFMEDIA:
cxgbe(4): Support SIOGIFXMEDIA so that ifconfig displays correct media for 25Gbps and 100Gbps ports. This should have been part of r305713, which is when the driver first started reporting extended media types.
2016-09-24 13:23:47 +00:00
case SIOCGIFXMEDIA:
cxgbe(4): Maintain one ifmedia per physical port instead of one per Virtual Interface (VI). All autonomous VIs that share a port share the same media. MFC after: 1 week Sponsored by: Chelsio Communications
2017-08-28 21:44:25 +00:00
ifmedia_ioctl(ifp, ifr, &pi->media, cmd);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
break;
cxgbe(4): add support for the SIOCGI2C ioctl.
2014-09-12 21:56:57 +00:00
case SIOCGI2C: {
struct ifi2creq i2c;
Use an accessor function to access ifr_data. This fixes 32-bit compat (no ioctl command defintions are required as struct ifreq is the same size). This is believed to be sufficent to fully support ifconfig on 32-bit systems. Reviewed by: kib Obtained from: CheriBSD MFC after: 1 week Relnotes: yes Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D14900
2018-03-30 18:50:13 +00:00
rc = copyin(ifr_data_get_ptr(ifr), &i2c, sizeof(i2c));
cxgbe(4): add support for the SIOCGI2C ioctl.
2014-09-12 21:56:57 +00:00
if (rc != 0)
break;
if (i2c.dev_addr != 0xA0 && i2c.dev_addr != 0xA2) {
rc = EPERM;
break;
}
if (i2c.len > sizeof(i2c.data)) {
rc = EINVAL;
break;
}
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
rc = begin_synchronized_op(sc, vi, SLEEP_OK | INTR_OK, "t4i2c");
cxgbe(4): add support for the SIOCGI2C ioctl.
2014-09-12 21:56:57 +00:00
if (rc)
return (rc);
cxgbe(4): Maintain one ifmedia per physical port instead of one per Virtual Interface (VI). All autonomous VIs that share a port share the same media. MFC after: 1 week Sponsored by: Chelsio Communications
2017-08-28 21:44:25 +00:00
rc = -t4_i2c_rd(sc, sc->mbox, pi->port_id, i2c.dev_addr,
cxgbe(4): add support for the SIOCGI2C ioctl.
2014-09-12 21:56:57 +00:00
i2c.offset, i2c.len, &i2c.data[0]);
end_synchronized_op(sc, 0);
if (rc == 0)
Use an accessor function to access ifr_data. This fixes 32-bit compat (no ioctl command defintions are required as struct ifreq is the same size). This is believed to be sufficent to fully support ifconfig on 32-bit systems. Reviewed by: kib Obtained from: CheriBSD MFC after: 1 week Relnotes: yes Sponsored by: DARPA, AFRL Differential Revision: https://reviews.freebsd.org/D14900
2018-03-30 18:50:13 +00:00
rc = copyout(&i2c, ifr_data_get_ptr(ifr), sizeof(i2c));
cxgbe(4): add support for the SIOCGI2C ioctl.
2014-09-12 21:56:57 +00:00
break;
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
default:
rc = ether_ioctl(ifp, cmd, data);
}
return (rc);
}
static int
cxgbe_transmit(struct ifnet *ifp, struct mbuf *m)
{
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct vi_info *vi = ifp->if_softc;
struct port_info *pi = vi->pi;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
struct adapter *sc = pi->adapter;
cxgbe(4): major tx rework. a) Front load as much work as possible in if_transmit, before any driver lock or software queue has to get involved. b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This is specifically for the tx multiqueue model where one of the if_transmit producer threads becomes the consumer and other producers carry on as usual. mp_ring is implemented as standalone code and it should be possible to use it in any driver with tx multiqueue. It also has: - the ability to enqueue/dequeue multiple items. This might become significant if packet batching is ever implemented. - an abdication mechanism to allow a thread to give up writing tx descriptors and have another if_transmit thread take over. A thread that's writing tx descriptors can end up doing so for an unbounded time period if a) there are other if_transmit threads continuously feeding the sofware queue, and b) the chip keeps up with whatever the thread is throwing at it. - accurate statistics about interesting events even when the stats come at the expense of additional branches/conditional code. The NIC txq lock is uncontested on the fast path at this point. I've left it there for synchronization with the control events (interface up/down, modload/unload). c) Add support for "type 1" coalescing work request in the normal NIC tx path. This work request is optimized for frames with a single item in the DMA gather list. These are very common when forwarding packets. Note that netmap tx in cxgbe already uses these "type 1" work requests. d) Do not request automatic cidx updates every 32 descriptors. Instead, request updates via bits in individual work requests (still every 32 descriptors approximately). Also, request an automatic final update when the queue idles after activity. This means NIC tx reclaim is still performed lazily but it will catch up quickly as soon as the queue idles. This seems to be the best middle ground and I'll probably do something similar for netmap tx as well. e) Implement a faster tx path for WRQs (used by TOE tx and control queues, _not_ by the normal NIC tx). Allow work requests to be written directly to the hardware descriptor ring if room is available. I will convert t4_tom and iw_cxgbe modules to this faster style gradually. MFC after: 2 months
2014-12-31 23:19:16 +00:00
struct sge_txq *txq;
void *items[1];
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
int rc;
M_ASSERTPKTHDR(m);
cxgbe(4): major tx rework. a) Front load as much work as possible in if_transmit, before any driver lock or software queue has to get involved. b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This is specifically for the tx multiqueue model where one of the if_transmit producer threads becomes the consumer and other producers carry on as usual. mp_ring is implemented as standalone code and it should be possible to use it in any driver with tx multiqueue. It also has: - the ability to enqueue/dequeue multiple items. This might become significant if packet batching is ever implemented. - an abdication mechanism to allow a thread to give up writing tx descriptors and have another if_transmit thread take over. A thread that's writing tx descriptors can end up doing so for an unbounded time period if a) there are other if_transmit threads continuously feeding the sofware queue, and b) the chip keeps up with whatever the thread is throwing at it. - accurate statistics about interesting events even when the stats come at the expense of additional branches/conditional code. The NIC txq lock is uncontested on the fast path at this point. I've left it there for synchronization with the control events (interface up/down, modload/unload). c) Add support for "type 1" coalescing work request in the normal NIC tx path. This work request is optimized for frames with a single item in the DMA gather list. These are very common when forwarding packets. Note that netmap tx in cxgbe already uses these "type 1" work requests. d) Do not request automatic cidx updates every 32 descriptors. Instead, request updates via bits in individual work requests (still every 32 descriptors approximately). Also, request an automatic final update when the queue idles after activity. This means NIC tx reclaim is still performed lazily but it will catch up quickly as soon as the queue idles. This seems to be the best middle ground and I'll probably do something similar for netmap tx as well. e) Implement a faster tx path for WRQs (used by TOE tx and control queues, _not_ by the normal NIC tx). Allow work requests to be written directly to the hardware descriptor ring if room is available. I will convert t4_tom and iw_cxgbe modules to this faster style gradually. MFC after: 2 months
2014-12-31 23:19:16 +00:00
MPASS(m->m_nextpkt == NULL); /* not quite ready for this yet */
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
if (__predict_false(pi->link_cfg.link_ok == false)) {
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
m_freem(m);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
return (ENETDOWN);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
rc = parse_pkt(sc, &m);
cxgbe(4): major tx rework. a) Front load as much work as possible in if_transmit, before any driver lock or software queue has to get involved. b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This is specifically for the tx multiqueue model where one of the if_transmit producer threads becomes the consumer and other producers carry on as usual. mp_ring is implemented as standalone code and it should be possible to use it in any driver with tx multiqueue. It also has: - the ability to enqueue/dequeue multiple items. This might become significant if packet batching is ever implemented. - an abdication mechanism to allow a thread to give up writing tx descriptors and have another if_transmit thread take over. A thread that's writing tx descriptors can end up doing so for an unbounded time period if a) there are other if_transmit threads continuously feeding the sofware queue, and b) the chip keeps up with whatever the thread is throwing at it. - accurate statistics about interesting events even when the stats come at the expense of additional branches/conditional code. The NIC txq lock is uncontested on the fast path at this point. I've left it there for synchronization with the control events (interface up/down, modload/unload). c) Add support for "type 1" coalescing work request in the normal NIC tx path. This work request is optimized for frames with a single item in the DMA gather list. These are very common when forwarding packets. Note that netmap tx in cxgbe already uses these "type 1" work requests. d) Do not request automatic cidx updates every 32 descriptors. Instead, request updates via bits in individual work requests (still every 32 descriptors approximately). Also, request an automatic final update when the queue idles after activity. This means NIC tx reclaim is still performed lazily but it will catch up quickly as soon as the queue idles. This seems to be the best middle ground and I'll probably do something similar for netmap tx as well. e) Implement a faster tx path for WRQs (used by TOE tx and control queues, _not_ by the normal NIC tx). Allow work requests to be written directly to the hardware descriptor ring if room is available. I will convert t4_tom and iw_cxgbe modules to this faster style gradually. MFC after: 2 months
2014-12-31 23:19:16 +00:00
if (__predict_false(rc != 0)) {
MPASS(m == NULL); /* was freed already */
atomic_add_int(&pi->tx_parse_error, 1); /* rare, atomic is ok */
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
return (rc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
cxgbe(4): Data path for rate-limited tx. This is hardware support for the SO_MAX_PACING_RATE sockopt (see setsockopt(2)), which is available in kernels built with "options RATELIMIT". Relnotes: Yes Sponsored by: Chelsio Communications
2018-05-24 10:18:14 +00:00
#ifdef RATELIMIT
Restructure mbuf send tags to provide stronger guarantees. - Perform ifp mismatch checks (to determine if a send tag is allocated for a different ifp than the one the packet is being output on), in ip_output() and ip6_output(). This avoids sending packets with send tags to ifnet drivers that don't support send tags. Since we are now checking for ifp mismatches before invoking if_output, we can now try to allocate a new tag before invoking if_output sending the original packet on the new tag if allocation succeeds. To avoid code duplication for the fragment and unfragmented cases, add ip_output_send() and ip6_output_send() as wrappers around if_output and nd6_output_ifp, respectively. All of the logic for setting send tags and dealing with send tag-related errors is done in these wrapper functions. For pseudo interfaces that wrap other network interfaces (vlan and lagg), wrapper send tags are now allocated so that ip*_output see the wrapper ifp as the ifp in the send tag. The if_transmit routines rewrite the send tags after performing an ifp mismatch check. If an ifp mismatch is detected, the transmit routines fail with EAGAIN. - To provide clearer life cycle management of send tags, especially in the presence of vlan and lagg wrapper tags, add a reference count to send tags managed via m_snd_tag_ref() and m_snd_tag_rele(). Provide a helper function (m_snd_tag_init()) for use by drivers supporting send tags. m_snd_tag_init() takes care of the if_ref on the ifp meaning that code alloating send tags via if_snd_tag_alloc no longer has to manage that manually. Similarly, m_snd_tag_rele drops the refcount on the ifp after invoking if_snd_tag_free when the last reference to a send tag is dropped. This also closes use after free races if there are pending packets in driver tx rings after the socket is closed (e.g. from tcpdrop). In order for m_free to work reliably, add a new CSUM_SND_TAG flag in csum_flags to indicate 'snd_tag' is set (rather than 'rcvif'). Drivers now also check this flag instead of checking snd_tag against NULL. This avoids false positive matches when a forwarded packet has a non-NULL rcvif that was treated as a send tag. - cxgbe was relying on snd_tag_free being called when the inp was detached so that it could kick the firmware to flush any pending work on the flow. This is because the driver doesn't require ACK messages from the firmware for every request, but instead does a kind of manual interrupt coalescing by only setting a flag to request a completion on a subset of requests. If all of the in-flight requests don't have the flag when the tag is detached from the inp, the flow might never return the credits. The current snd_tag_free command issues a flush command to force the credits to return. However, the credit return is what also frees the mbufs, and since those mbufs now hold references on the tag, this meant that snd_tag_free would never be called. To fix, explicitly drop the mbuf's reference on the snd tag when the mbuf is queued in the firmware work queue. This means that once the inp's reference on the tag goes away and all in-flight mbufs have been queued to the firmware, tag's refcount will drop to zero and snd_tag_free will kick in and send the flush request. Note that we need to avoid doing this in the middle of ethofld_tx(), so the driver grabs a temporary reference on the tag around that loop to defer the free to the end of the function in case it sends the last mbuf to the queue after the inp has dropped its reference on the tag. - mlx5 preallocates send tags and was using the ifp pointer even when the send tag wasn't in use. Explicitly use the ifp from other data structures instead. - Sprinkle some assertions in various places to assert that received packets don't have a send tag, and that other places that overwrite rcvif (e.g. 802.11 transmit) don't clobber a send tag pointer. Reviewed by: gallatin, hselasky, rgrimes, ae Sponsored by: Netflix Differential Revision: https://reviews.freebsd.org/D20117
2019-05-24 22:30:40 +00:00
if (m->m_pkthdr.csum_flags & CSUM_SND_TAG) {
MPASS(m->m_pkthdr.snd_tag->ifp == ifp);
cxgbe(4): Data path for rate-limited tx. This is hardware support for the SO_MAX_PACING_RATE sockopt (see setsockopt(2)), which is available in kernels built with "options RATELIMIT". Relnotes: Yes Sponsored by: Chelsio Communications
2018-05-24 10:18:14 +00:00
return (ethofld_transmit(ifp, m));
}
#endif
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): major tx rework. a) Front load as much work as possible in if_transmit, before any driver lock or software queue has to get involved. b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This is specifically for the tx multiqueue model where one of the if_transmit producer threads becomes the consumer and other producers carry on as usual. mp_ring is implemented as standalone code and it should be possible to use it in any driver with tx multiqueue. It also has: - the ability to enqueue/dequeue multiple items. This might become significant if packet batching is ever implemented. - an abdication mechanism to allow a thread to give up writing tx descriptors and have another if_transmit thread take over. A thread that's writing tx descriptors can end up doing so for an unbounded time period if a) there are other if_transmit threads continuously feeding the sofware queue, and b) the chip keeps up with whatever the thread is throwing at it. - accurate statistics about interesting events even when the stats come at the expense of additional branches/conditional code. The NIC txq lock is uncontested on the fast path at this point. I've left it there for synchronization with the control events (interface up/down, modload/unload). c) Add support for "type 1" coalescing work request in the normal NIC tx path. This work request is optimized for frames with a single item in the DMA gather list. These are very common when forwarding packets. Note that netmap tx in cxgbe already uses these "type 1" work requests. d) Do not request automatic cidx updates every 32 descriptors. Instead, request updates via bits in individual work requests (still every 32 descriptors approximately). Also, request an automatic final update when the queue idles after activity. This means NIC tx reclaim is still performed lazily but it will catch up quickly as soon as the queue idles. This seems to be the best middle ground and I'll probably do something similar for netmap tx as well. e) Implement a faster tx path for WRQs (used by TOE tx and control queues, _not_ by the normal NIC tx). Allow work requests to be written directly to the hardware descriptor ring if room is available. I will convert t4_tom and iw_cxgbe modules to this faster style gradually. MFC after: 2 months
2014-12-31 23:19:16 +00:00
/* Select a txq. */
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
txq = &sc->sge.txq[vi->first_txq];
cxgbe(4): major tx rework. a) Front load as much work as possible in if_transmit, before any driver lock or software queue has to get involved. b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This is specifically for the tx multiqueue model where one of the if_transmit producer threads becomes the consumer and other producers carry on as usual. mp_ring is implemented as standalone code and it should be possible to use it in any driver with tx multiqueue. It also has: - the ability to enqueue/dequeue multiple items. This might become significant if packet batching is ever implemented. - an abdication mechanism to allow a thread to give up writing tx descriptors and have another if_transmit thread take over. A thread that's writing tx descriptors can end up doing so for an unbounded time period if a) there are other if_transmit threads continuously feeding the sofware queue, and b) the chip keeps up with whatever the thread is throwing at it. - accurate statistics about interesting events even when the stats come at the expense of additional branches/conditional code. The NIC txq lock is uncontested on the fast path at this point. I've left it there for synchronization with the control events (interface up/down, modload/unload). c) Add support for "type 1" coalescing work request in the normal NIC tx path. This work request is optimized for frames with a single item in the DMA gather list. These are very common when forwarding packets. Note that netmap tx in cxgbe already uses these "type 1" work requests. d) Do not request automatic cidx updates every 32 descriptors. Instead, request updates via bits in individual work requests (still every 32 descriptors approximately). Also, request an automatic final update when the queue idles after activity. This means NIC tx reclaim is still performed lazily but it will catch up quickly as soon as the queue idles. This seems to be the best middle ground and I'll probably do something similar for netmap tx as well. e) Implement a faster tx path for WRQs (used by TOE tx and control queues, _not_ by the normal NIC tx). Allow work requests to be written directly to the hardware descriptor ring if room is available. I will convert t4_tom and iw_cxgbe modules to this faster style gradually. MFC after: 2 months
2014-12-31 23:19:16 +00:00
if (M_HASHTYPE_GET(m) != M_HASHTYPE_NONE)
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
txq += ((m->m_pkthdr.flowid % (vi->ntxq - vi->rsrv_noflowq)) +
vi->rsrv_noflowq);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): major tx rework. a) Front load as much work as possible in if_transmit, before any driver lock or software queue has to get involved. b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This is specifically for the tx multiqueue model where one of the if_transmit producer threads becomes the consumer and other producers carry on as usual. mp_ring is implemented as standalone code and it should be possible to use it in any driver with tx multiqueue. It also has: - the ability to enqueue/dequeue multiple items. This might become significant if packet batching is ever implemented. - an abdication mechanism to allow a thread to give up writing tx descriptors and have another if_transmit thread take over. A thread that's writing tx descriptors can end up doing so for an unbounded time period if a) there are other if_transmit threads continuously feeding the sofware queue, and b) the chip keeps up with whatever the thread is throwing at it. - accurate statistics about interesting events even when the stats come at the expense of additional branches/conditional code. The NIC txq lock is uncontested on the fast path at this point. I've left it there for synchronization with the control events (interface up/down, modload/unload). c) Add support for "type 1" coalescing work request in the normal NIC tx path. This work request is optimized for frames with a single item in the DMA gather list. These are very common when forwarding packets. Note that netmap tx in cxgbe already uses these "type 1" work requests. d) Do not request automatic cidx updates every 32 descriptors. Instead, request updates via bits in individual work requests (still every 32 descriptors approximately). Also, request an automatic final update when the queue idles after activity. This means NIC tx reclaim is still performed lazily but it will catch up quickly as soon as the queue idles. This seems to be the best middle ground and I'll probably do something similar for netmap tx as well. e) Implement a faster tx path for WRQs (used by TOE tx and control queues, _not_ by the normal NIC tx). Allow work requests to be written directly to the hardware descriptor ring if room is available. I will convert t4_tom and iw_cxgbe modules to this faster style gradually. MFC after: 2 months
2014-12-31 23:19:16 +00:00
items[0] = m;
rc = mp_ring_enqueue(txq->r, items, 1, 4096);
if (__predict_false(rc != 0))
m_freem(m);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
return (rc);
}
static void
cxgbe_qflush(struct ifnet *ifp)
{
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct vi_info *vi = ifp->if_softc;
Fix a couple of bad races that can occur when a cxgbe interface is taken down. The ingress queue lock was unused and has been removed as part of these changes. - An in-flight egress update from the SGE must be handled before the queue that requested it is destroyed. Wait for the update to arrive. - Interrupt handlers must stop processing rx events for a queue before the queue is destroyed. Events that have not yet been processed should be ignored once the queue disappears. MFC after: 1 week
2011-04-15 03:09:27 +00:00
struct sge_txq *txq;
int i;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
/* queues do not exist if !VI_INIT_DONE. */
if (vi->flags & VI_INIT_DONE) {
for_each_txq(vi, i, txq) {
Fix a couple of bad races that can occur when a cxgbe interface is taken down. The ingress queue lock was unused and has been removed as part of these changes. - An in-flight egress update from the SGE must be handled before the queue that requested it is destroyed. Wait for the update to arrive. - Interrupt handlers must stop processing rx events for a queue before the queue is destroyed. Events that have not yet been processed should be ignored once the queue disappears. MFC after: 1 week
2011-04-15 03:09:27 +00:00
TXQ_LOCK(txq);
cxgbe(4): Do not assume that if_qflush is always followed by inteface-down. MFC after: 3 days Sponsored by: Chelsio Communications
2017-05-09 18:33:41 +00:00
txq->eq.flags |= EQ_QFLUSH;
Fix a couple of bad races that can occur when a cxgbe interface is taken down. The ingress queue lock was unused and has been removed as part of these changes. - An in-flight egress update from the SGE must be handled before the queue that requested it is destroyed. Wait for the update to arrive. - Interrupt handlers must stop processing rx events for a queue before the queue is destroyed. Events that have not yet been processed should be ignored once the queue disappears. MFC after: 1 week
2011-04-15 03:09:27 +00:00
TXQ_UNLOCK(txq);
cxgbe(4): major tx rework. a) Front load as much work as possible in if_transmit, before any driver lock or software queue has to get involved. b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This is specifically for the tx multiqueue model where one of the if_transmit producer threads becomes the consumer and other producers carry on as usual. mp_ring is implemented as standalone code and it should be possible to use it in any driver with tx multiqueue. It also has: - the ability to enqueue/dequeue multiple items. This might become significant if packet batching is ever implemented. - an abdication mechanism to allow a thread to give up writing tx descriptors and have another if_transmit thread take over. A thread that's writing tx descriptors can end up doing so for an unbounded time period if a) there are other if_transmit threads continuously feeding the sofware queue, and b) the chip keeps up with whatever the thread is throwing at it. - accurate statistics about interesting events even when the stats come at the expense of additional branches/conditional code. The NIC txq lock is uncontested on the fast path at this point. I've left it there for synchronization with the control events (interface up/down, modload/unload). c) Add support for "type 1" coalescing work request in the normal NIC tx path. This work request is optimized for frames with a single item in the DMA gather list. These are very common when forwarding packets. Note that netmap tx in cxgbe already uses these "type 1" work requests. d) Do not request automatic cidx updates every 32 descriptors. Instead, request updates via bits in individual work requests (still every 32 descriptors approximately). Also, request an automatic final update when the queue idles after activity. This means NIC tx reclaim is still performed lazily but it will catch up quickly as soon as the queue idles. This seems to be the best middle ground and I'll probably do something similar for netmap tx as well. e) Implement a faster tx path for WRQs (used by TOE tx and control queues, _not_ by the normal NIC tx). Allow work requests to be written directly to the hardware descriptor ring if room is available. I will convert t4_tom and iw_cxgbe modules to this faster style gradually. MFC after: 2 months
2014-12-31 23:19:16 +00:00
while (!mp_ring_is_idle(txq->r)) {
mp_ring_check_drainage(txq->r, 0);
pause("qflush", 1);
}
cxgbe(4): Do not assume that if_qflush is always followed by inteface-down. MFC after: 3 days Sponsored by: Chelsio Communications
2017-05-09 18:33:41 +00:00
TXQ_LOCK(txq);
txq->eq.flags &= ~EQ_QFLUSH;
TXQ_UNLOCK(txq);
Fix a couple of bad races that can occur when a cxgbe interface is taken down. The ingress queue lock was unused and has been removed as part of these changes. - An in-flight egress update from the SGE must be handled before the queue that requested it is destroyed. Wait for the update to arrive. - Interrupt handlers must stop processing rx events for a queue before the queue is destroyed. Events that have not yet been processed should be ignored once the queue disappears. MFC after: 1 week
2011-04-15 03:09:27 +00:00
}
}
if_qflush(ifp);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
cxgbe(4): implement if_get_counter.
2014-09-27 05:50:31 +00:00
static uint64_t
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
vi_get_counter(struct ifnet *ifp, ift_counter c)
{
struct vi_info *vi = ifp->if_softc;
struct fw_vi_stats_vf *s = &vi->stats;
vi_refresh_stats(vi->pi->adapter, vi);
switch (c) {
case IFCOUNTER_IPACKETS:
return (s->rx_bcast_frames + s->rx_mcast_frames +
s->rx_ucast_frames);
case IFCOUNTER_IERRORS:
return (s->rx_err_frames);
case IFCOUNTER_OPACKETS:
return (s->tx_bcast_frames + s->tx_mcast_frames +
s->tx_ucast_frames + s->tx_offload_frames);
case IFCOUNTER_OERRORS:
return (s->tx_drop_frames);
case IFCOUNTER_IBYTES:
return (s->rx_bcast_bytes + s->rx_mcast_bytes +
s->rx_ucast_bytes);
case IFCOUNTER_OBYTES:
return (s->tx_bcast_bytes + s->tx_mcast_bytes +
s->tx_ucast_bytes + s->tx_offload_bytes);
case IFCOUNTER_IMCASTS:
return (s->rx_mcast_frames);
case IFCOUNTER_OMCASTS:
return (s->tx_mcast_frames);
case IFCOUNTER_OQDROPS: {
uint64_t drops;
drops = 0;
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
if (vi->flags & VI_INIT_DONE) {
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
int i;
struct sge_txq *txq;
for_each_txq(vi, i, txq)
drops += counter_u64_fetch(txq->r->drops);
}
return (drops);
}
default:
return (if_get_counter_default(ifp, c));
}
}
uint64_t
cxgbe(4): implement if_get_counter.
2014-09-27 05:50:31 +00:00
cxgbe_get_counter(struct ifnet *ifp, ift_counter c)
{
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct vi_info *vi = ifp->if_softc;
struct port_info *pi = vi->pi;
cxgbe(4): implement if_get_counter.
2014-09-27 05:50:31 +00:00
struct adapter *sc = pi->adapter;
struct port_stats *s = &pi->stats;
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
if (pi->nvi > 1 || sc->flags & IS_VF)
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
return (vi_get_counter(ifp, c));
cxgbe(4): implement if_get_counter.
2014-09-27 05:50:31 +00:00
cxgbe_refresh_stats(sc, pi);
switch (c) {
case IFCOUNTER_IPACKETS:
cxgbe(4): Tidy up PAUSE frame accounting. Figure out if the chip is counting PAUSE frames in the "normal" stats and take them out if it is. This fixes a bug in the tx stats because the default hardware behavior is different for Tx and Rx but the driver was treating both the same way. The result was that OPACKETS, OBYTES, and OMCASTS were under-reported (if tx_pause > 0) before this change. Note that the mac_stats sysctl still gives you the raw value of these statistics straight from the device registers.
2016-03-17 01:15:16 +00:00
return (s->rx_frames);
cxgbe(4): implement if_get_counter.
2014-09-27 05:50:31 +00:00
case IFCOUNTER_IERRORS:
return (s->rx_jabber + s->rx_runt + s->rx_too_long +
s->rx_fcs_err + s->rx_len_err);
case IFCOUNTER_OPACKETS:
cxgbe(4): Tidy up PAUSE frame accounting. Figure out if the chip is counting PAUSE frames in the "normal" stats and take them out if it is. This fixes a bug in the tx stats because the default hardware behavior is different for Tx and Rx but the driver was treating both the same way. The result was that OPACKETS, OBYTES, and OMCASTS were under-reported (if tx_pause > 0) before this change. Note that the mac_stats sysctl still gives you the raw value of these statistics straight from the device registers.
2016-03-17 01:15:16 +00:00
return (s->tx_frames);
cxgbe(4): implement if_get_counter.
2014-09-27 05:50:31 +00:00
case IFCOUNTER_OERRORS:
return (s->tx_error_frames);
case IFCOUNTER_IBYTES:
cxgbe(4): Tidy up PAUSE frame accounting. Figure out if the chip is counting PAUSE frames in the "normal" stats and take them out if it is. This fixes a bug in the tx stats because the default hardware behavior is different for Tx and Rx but the driver was treating both the same way. The result was that OPACKETS, OBYTES, and OMCASTS were under-reported (if tx_pause > 0) before this change. Note that the mac_stats sysctl still gives you the raw value of these statistics straight from the device registers.
2016-03-17 01:15:16 +00:00
return (s->rx_octets);
cxgbe(4): implement if_get_counter.
2014-09-27 05:50:31 +00:00
case IFCOUNTER_OBYTES:
cxgbe(4): Tidy up PAUSE frame accounting. Figure out if the chip is counting PAUSE frames in the "normal" stats and take them out if it is. This fixes a bug in the tx stats because the default hardware behavior is different for Tx and Rx but the driver was treating both the same way. The result was that OPACKETS, OBYTES, and OMCASTS were under-reported (if tx_pause > 0) before this change. Note that the mac_stats sysctl still gives you the raw value of these statistics straight from the device registers.
2016-03-17 01:15:16 +00:00
return (s->tx_octets);
cxgbe(4): implement if_get_counter.
2014-09-27 05:50:31 +00:00
case IFCOUNTER_IMCASTS:
cxgbe(4): Tidy up PAUSE frame accounting. Figure out if the chip is counting PAUSE frames in the "normal" stats and take them out if it is. This fixes a bug in the tx stats because the default hardware behavior is different for Tx and Rx but the driver was treating both the same way. The result was that OPACKETS, OBYTES, and OMCASTS were under-reported (if tx_pause > 0) before this change. Note that the mac_stats sysctl still gives you the raw value of these statistics straight from the device registers.
2016-03-17 01:15:16 +00:00
return (s->rx_mcast_frames);
cxgbe(4): implement if_get_counter.
2014-09-27 05:50:31 +00:00
case IFCOUNTER_OMCASTS:
cxgbe(4): Tidy up PAUSE frame accounting. Figure out if the chip is counting PAUSE frames in the "normal" stats and take them out if it is. This fixes a bug in the tx stats because the default hardware behavior is different for Tx and Rx but the driver was treating both the same way. The result was that OPACKETS, OBYTES, and OMCASTS were under-reported (if tx_pause > 0) before this change. Note that the mac_stats sysctl still gives you the raw value of these statistics straight from the device registers.
2016-03-17 01:15:16 +00:00
return (s->tx_mcast_frames);
cxgbe(4): implement if_get_counter.
2014-09-27 05:50:31 +00:00
case IFCOUNTER_IQDROPS:
return (s->rx_ovflow0 + s->rx_ovflow1 + s->rx_ovflow2 +
s->rx_ovflow3 + s->rx_trunc0 + s->rx_trunc1 + s->rx_trunc2 +
s->rx_trunc3 + pi->tnl_cong_drops);
case IFCOUNTER_OQDROPS: {
uint64_t drops;
drops = s->tx_drop;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (vi->flags & VI_INIT_DONE) {
cxgbe(4): implement if_get_counter.
2014-09-27 05:50:31 +00:00
int i;
struct sge_txq *txq;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
for_each_txq(vi, i, txq)
cxgbe(4): major tx rework. a) Front load as much work as possible in if_transmit, before any driver lock or software queue has to get involved. b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This is specifically for the tx multiqueue model where one of the if_transmit producer threads becomes the consumer and other producers carry on as usual. mp_ring is implemented as standalone code and it should be possible to use it in any driver with tx multiqueue. It also has: - the ability to enqueue/dequeue multiple items. This might become significant if packet batching is ever implemented. - an abdication mechanism to allow a thread to give up writing tx descriptors and have another if_transmit thread take over. A thread that's writing tx descriptors can end up doing so for an unbounded time period if a) there are other if_transmit threads continuously feeding the sofware queue, and b) the chip keeps up with whatever the thread is throwing at it. - accurate statistics about interesting events even when the stats come at the expense of additional branches/conditional code. The NIC txq lock is uncontested on the fast path at this point. I've left it there for synchronization with the control events (interface up/down, modload/unload). c) Add support for "type 1" coalescing work request in the normal NIC tx path. This work request is optimized for frames with a single item in the DMA gather list. These are very common when forwarding packets. Note that netmap tx in cxgbe already uses these "type 1" work requests. d) Do not request automatic cidx updates every 32 descriptors. Instead, request updates via bits in individual work requests (still every 32 descriptors approximately). Also, request an automatic final update when the queue idles after activity. This means NIC tx reclaim is still performed lazily but it will catch up quickly as soon as the queue idles. This seems to be the best middle ground and I'll probably do something similar for netmap tx as well. e) Implement a faster tx path for WRQs (used by TOE tx and control queues, _not_ by the normal NIC tx). Allow work requests to be written directly to the hardware descriptor ring if room is available. I will convert t4_tom and iw_cxgbe modules to this faster style gradually. MFC after: 2 months
2014-12-31 23:19:16 +00:00
drops += counter_u64_fetch(txq->r->drops);
cxgbe(4): implement if_get_counter.
2014-09-27 05:50:31 +00:00
}
return (drops);
}
default:
return (if_get_counter_default(ifp, c));
}
}
cxgbe(4): Implement ifm_change callback. Sponsored by: Chelsio Communications
2018-05-31 02:10:50 +00:00
/*
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
* The kernel picks a media from the list we had provided but we still validate
* the requeste.
cxgbe(4): Implement ifm_change callback. Sponsored by: Chelsio Communications
2018-05-31 02:10:50 +00:00
*/
cxgbev(4): Updates to the VF driver to cope with recent ifmedia and ctrlq changes in the base driver. MFC after: 1 week Sponsored by: Chelsio Communications
2018-08-23 00:58:10 +00:00
int
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe_media_change(struct ifnet *ifp)
{
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct vi_info *vi = ifp->if_softc;
cxgbe(4): Implement ifm_change callback. Sponsored by: Chelsio Communications
2018-05-31 02:10:50 +00:00
struct port_info *pi = vi->pi;
struct ifmedia *ifm = &pi->media;
struct link_config *lc = &pi->link_cfg;
struct adapter *sc = pi->adapter;
int rc;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Implement ifm_change callback. Sponsored by: Chelsio Communications
2018-05-31 02:10:50 +00:00
rc = begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK, "t4mec");
if (rc != 0)
return (rc);
PORT_LOCK(pi);
if (IFM_SUBTYPE(ifm->ifm_media) == IFM_AUTO) {
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
/* ifconfig .. media autoselect */
if (!(lc->supported & FW_PORT_CAP32_ANEG)) {
rc = ENOTSUP; /* AN not supported by transceiver */
goto done;
}
cxgbe(4): Implement ifm_change callback. Sponsored by: Chelsio Communications
2018-05-31 02:10:50 +00:00
lc->requested_aneg = AUTONEG_ENABLE;
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
lc->requested_speed = 0;
lc->requested_fc |= PAUSE_AUTONEG;
cxgbe(4): Implement ifm_change callback. Sponsored by: Chelsio Communications
2018-05-31 02:10:50 +00:00
} else {
lc->requested_aneg = AUTONEG_DISABLE;
lc->requested_speed =
ifmedia_baudrate(ifm->ifm_media) / 1000000;
lc->requested_fc = 0;
if (IFM_OPTIONS(ifm->ifm_media) & IFM_ETH_RXPAUSE)
lc->requested_fc |= PAUSE_RX;
if (IFM_OPTIONS(ifm->ifm_media) & IFM_ETH_TXPAUSE)
lc->requested_fc |= PAUSE_TX;
}
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
if (pi->up_vis > 0) {
fixup_link_config(pi);
rc = apply_link_config(pi);
}
done:
cxgbe(4): Implement ifm_change callback. Sponsored by: Chelsio Communications
2018-05-31 02:10:50 +00:00
PORT_UNLOCK(pi);
end_synchronized_op(sc, 0);
return (rc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
/*
* Base media word (without ETHER, pause, link active, etc.) for the port at the
* given speed.
*/
static int
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
port_mword(struct port_info *pi, uint32_t speed)
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
{
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
MPASS(speed & M_FW_PORT_CAP32_SPEED);
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
MPASS(powerof2(speed));
switch(pi->port_type) {
case FW_PORT_TYPE_BT_SGMII:
case FW_PORT_TYPE_BT_XFI:
case FW_PORT_TYPE_BT_XAUI:
/* BaseT */
switch (speed) {
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
case FW_PORT_CAP32_SPEED_100M:
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_100_T);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
case FW_PORT_CAP32_SPEED_1G:
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_1000_T);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
case FW_PORT_CAP32_SPEED_10G:
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_10G_T);
}
break;
case FW_PORT_TYPE_KX4:
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
if (speed == FW_PORT_CAP32_SPEED_10G)
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_10G_KX4);
break;
case FW_PORT_TYPE_CX4:
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
if (speed == FW_PORT_CAP32_SPEED_10G)
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_10G_CX4);
break;
case FW_PORT_TYPE_KX:
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
if (speed == FW_PORT_CAP32_SPEED_1G)
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_1000_KX);
break;
case FW_PORT_TYPE_KR:
case FW_PORT_TYPE_BP_AP:
case FW_PORT_TYPE_BP4_AP:
case FW_PORT_TYPE_BP40_BA:
case FW_PORT_TYPE_KR4_100G:
case FW_PORT_TYPE_KR_SFP28:
case FW_PORT_TYPE_KR_XLAUI:
switch (speed) {
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
case FW_PORT_CAP32_SPEED_1G:
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_1000_KX);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
case FW_PORT_CAP32_SPEED_10G:
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_10G_KR);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
case FW_PORT_CAP32_SPEED_25G:
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_25G_KR);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
case FW_PORT_CAP32_SPEED_40G:
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_40G_KR4);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
case FW_PORT_CAP32_SPEED_50G:
return (IFM_50G_KR2);
case FW_PORT_CAP32_SPEED_100G:
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_100G_KR4);
}
break;
case FW_PORT_TYPE_FIBER_XFI:
case FW_PORT_TYPE_FIBER_XAUI:
case FW_PORT_TYPE_SFP:
case FW_PORT_TYPE_QSFP_10G:
case FW_PORT_TYPE_QSA:
case FW_PORT_TYPE_QSFP:
case FW_PORT_TYPE_CR4_QSFP:
case FW_PORT_TYPE_CR_QSFP:
case FW_PORT_TYPE_CR2_QSFP:
case FW_PORT_TYPE_SFP28:
/* Pluggable transceiver */
switch (pi->mod_type) {
case FW_PORT_MOD_TYPE_LR:
switch (speed) {
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
case FW_PORT_CAP32_SPEED_1G:
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_1000_LX);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
case FW_PORT_CAP32_SPEED_10G:
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_10G_LR);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
case FW_PORT_CAP32_SPEED_25G:
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_25G_LR);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
case FW_PORT_CAP32_SPEED_40G:
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_40G_LR4);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
case FW_PORT_CAP32_SPEED_50G:
return (IFM_50G_LR2);
case FW_PORT_CAP32_SPEED_100G:
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_100G_LR4);
}
break;
case FW_PORT_MOD_TYPE_SR:
switch (speed) {
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
case FW_PORT_CAP32_SPEED_1G:
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_1000_SX);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
case FW_PORT_CAP32_SPEED_10G:
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_10G_SR);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
case FW_PORT_CAP32_SPEED_25G:
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_25G_SR);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
case FW_PORT_CAP32_SPEED_40G:
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_40G_SR4);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
case FW_PORT_CAP32_SPEED_50G:
return (IFM_50G_SR2);
case FW_PORT_CAP32_SPEED_100G:
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_100G_SR4);
}
break;
case FW_PORT_MOD_TYPE_ER:
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
if (speed == FW_PORT_CAP32_SPEED_10G)
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_10G_ER);
break;
case FW_PORT_MOD_TYPE_TWINAX_PASSIVE:
case FW_PORT_MOD_TYPE_TWINAX_ACTIVE:
switch (speed) {
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
case FW_PORT_CAP32_SPEED_1G:
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_1000_CX);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
case FW_PORT_CAP32_SPEED_10G:
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_10G_TWINAX);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
case FW_PORT_CAP32_SPEED_25G:
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_25G_CR);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
case FW_PORT_CAP32_SPEED_40G:
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_40G_CR4);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
case FW_PORT_CAP32_SPEED_50G:
return (IFM_50G_CR2);
case FW_PORT_CAP32_SPEED_100G:
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_100G_CR4);
}
break;
case FW_PORT_MOD_TYPE_LRM:
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
if (speed == FW_PORT_CAP32_SPEED_10G)
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (IFM_10G_LRM);
break;
case FW_PORT_MOD_TYPE_NA:
MPASS(0); /* Not pluggable? */
/* fall throough */
case FW_PORT_MOD_TYPE_ERROR:
case FW_PORT_MOD_TYPE_UNKNOWN:
case FW_PORT_MOD_TYPE_NOTSUPPORTED:
break;
case FW_PORT_MOD_TYPE_NONE:
return (IFM_NONE);
}
break;
case FW_PORT_TYPE_NONE:
return (IFM_NONE);
}
return (IFM_UNKNOWN);
}
cxgbev(4): Updates to the VF driver to cope with recent ifmedia and ctrlq changes in the base driver. MFC after: 1 week Sponsored by: Chelsio Communications
2018-08-23 00:58:10 +00:00
void
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe_media_status(struct ifnet *ifp, struct ifmediareq *ifmr)
{
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct vi_info *vi = ifp->if_softc;
struct port_info *pi = vi->pi;
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
struct adapter *sc = pi->adapter;
cxgbe(4): Various link/media related improvements. - Deal with changes to port_type, and not just port_mod when a transceiver is changed. This fixes hot swapping of transceivers of different types (QSFP+ or QSA or QSFP28 in a QSFP28 port, SFP+ or SFP28 in a SFP28 port, etc.). - Always refresh media information for ifconfig if the port is down. The firmware does not generate tranceiver-change interrupts unless at least one VI is enabled on the physical port. Before this change ifconfig diplayed potentially stale information for ports that were administratively down. - Always recalculate and reapply L1 config on a transceiver change. - Display PAUSE settings in ifconfig. The driver sysctls for this continue to work as well. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-17 00:42:13 +00:00
struct link_config *lc = &pi->link_cfg;
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
if (begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK, "t4med") != 0)
return;
PORT_LOCK(pi);
if (pi->up_vis == 0) {
/*
* If all the interfaces are administratively down the firmware
* does not report transceiver changes. Refresh port info here
* so that ifconfig displays accurate ifmedia at all times.
* This is the only reason we have a synchronized op in this
* function. Just PORT_LOCK would have been enough otherwise.
*/
t4_update_port_info(pi);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
build_medialist(pi);
cxgbe(4): Various link/media related improvements. - Deal with changes to port_type, and not just port_mod when a transceiver is changed. This fixes hot swapping of transceivers of different types (QSFP+ or QSA or QSFP28 in a QSFP28 port, SFP+ or SFP28 in a SFP28 port, etc.). - Always refresh media information for ifconfig if the port is down. The firmware does not generate tranceiver-change interrupts unless at least one VI is enabled on the physical port. Before this change ifconfig diplayed potentially stale information for ports that were administratively down. - Always recalculate and reapply L1 config on a transceiver change. - Display PAUSE settings in ifconfig. The driver sysctls for this continue to work as well. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-17 00:42:13 +00:00
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
/* ifm_status */
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
ifmr->ifm_status = IFM_AVALID;
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
if (lc->link_ok == false)
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
goto done;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
ifmr->ifm_status |= IFM_ACTIVE;
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
/* ifm_active */
cxgbe(4): Include full duplex mediaopt in media that can be reported as active. Always report full duplex in active media. Sponsored by: Chelsio Communications
2018-06-01 16:46:29 +00:00
ifmr->ifm_active = IFM_ETHER | IFM_FDX;
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
ifmr->ifm_active &= ~(IFM_ETH_TXPAUSE | IFM_ETH_RXPAUSE);
cxgbe(4): Various link/media related improvements. - Deal with changes to port_type, and not just port_mod when a transceiver is changed. This fixes hot swapping of transceivers of different types (QSFP+ or QSA or QSFP28 in a QSFP28 port, SFP+ or SFP28 in a SFP28 port, etc.). - Always refresh media information for ifconfig if the port is down. The firmware does not generate tranceiver-change interrupts unless at least one VI is enabled on the physical port. Before this change ifconfig diplayed potentially stale information for ports that were administratively down. - Always recalculate and reapply L1 config on a transceiver change. - Display PAUSE settings in ifconfig. The driver sysctls for this continue to work as well. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-17 00:42:13 +00:00
if (lc->fc & PAUSE_RX)
ifmr->ifm_active |= IFM_ETH_RXPAUSE;
if (lc->fc & PAUSE_TX)
ifmr->ifm_active |= IFM_ETH_TXPAUSE;
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
ifmr->ifm_active |= port_mword(pi, speed_to_fwcap(lc->speed));
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
done:
PORT_UNLOCK(pi);
end_synchronized_op(sc, 0);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
static int
vcxgbe_probe(device_t dev)
{
char buf[128];
struct vi_info *vi = device_get_softc(dev);
snprintf(buf, sizeof(buf), "port %d vi %td", vi->pi->port_id,
vi - vi->pi->vi);
device_set_desc_copy(dev, buf);
return (BUS_PROBE_DEFAULT);
}
static int
cxgbe(4): Do not access the mailbox without appropriate locks while creating hardware VIs. This fixes a bad race on systems with hw.cxgbe.num_vis > 1. Reported by: olivier@ MFC after: 1 week Sponsored by: Chelsio Communications
2017-08-28 22:41:15 +00:00
alloc_extra_vi(struct adapter *sc, struct port_info *pi, struct vi_info *vi)
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
{
int func, index, rc;
cxgbe(4): Do not access the mailbox without appropriate locks while creating hardware VIs. This fixes a bad race on systems with hw.cxgbe.num_vis > 1. Reported by: olivier@ MFC after: 1 week Sponsored by: Chelsio Communications
2017-08-28 22:41:15 +00:00
uint32_t param, val;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
cxgbe(4): Do not access the mailbox without appropriate locks while creating hardware VIs. This fixes a bad race on systems with hw.cxgbe.num_vis > 1. Reported by: olivier@ MFC after: 1 week Sponsored by: Chelsio Communications
2017-08-28 22:41:15 +00:00
ASSERT_SYNCHRONIZED_OP(sc);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
index = vi - pi->vi;
cxgbe(4): vi_mac_funcs should include the base Ethernet function. It is already used in the driver as if it does. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-28 07:50:54 +00:00
MPASS(index > 0); /* This function deals with _extra_ VIs only */
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
KASSERT(index < nitems(vi_mac_funcs),
("%s: VI %s doesn't have a MAC func", __func__,
cxgbe(4): Do not access the mailbox without appropriate locks while creating hardware VIs. This fixes a bad race on systems with hw.cxgbe.num_vis > 1. Reported by: olivier@ MFC after: 1 week Sponsored by: Chelsio Communications
2017-08-28 22:41:15 +00:00
device_get_nameunit(vi->dev)));
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
func = vi_mac_funcs[index];
rc = t4_alloc_vi_func(sc, sc->mbox, pi->tx_chan, sc->pf, 0, 1,
cxgbe(4): Treat the viid as an opaque identifier. Recent firmwares prefer to use a different format for viid internally and this change allows them to do so. MFC after: 1 week Sponsored by: Chelsio Communications
2019-03-20 17:27:11 +00:00
vi->hw_addr, &vi->rss_size, &vi->vfvld, &vi->vin, func, 0);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (rc < 0) {
cxgbe(4): Do not access the mailbox without appropriate locks while creating hardware VIs. This fixes a bad race on systems with hw.cxgbe.num_vis > 1. Reported by: olivier@ MFC after: 1 week Sponsored by: Chelsio Communications
2017-08-28 22:41:15 +00:00
device_printf(vi->dev, "failed to allocate virtual interface %d"
"for port %d: %d\n", index, pi->port_id, -rc);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
return (-rc);
}
vi->viid = rc;
cxgbe(4): Do not access the mailbox without appropriate locks while creating hardware VIs. This fixes a bad race on systems with hw.cxgbe.num_vis > 1. Reported by: olivier@ MFC after: 1 week Sponsored by: Chelsio Communications
2017-08-28 22:41:15 +00:00
if (vi->rss_size == 1) {
/*
* This VI didn't get a slice of the RSS table. Reduce the
* number of VIs being created (hw.cxgbe.num_vis) or modify the
* configuration file (nvi, rssnvi for this PF) if this is a
* problem.
*/
device_printf(vi->dev, "RSS table not available.\n");
vi->rss_base = 0xffff;
return (0);
}
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
param = V_FW_PARAMS_MNEM(FW_PARAMS_MNEM_DEV) |
V_FW_PARAMS_PARAM_X(FW_PARAMS_PARAM_DEV_RSSINFO) |
V_FW_PARAMS_PARAM_YZ(vi->viid);
rc = t4_query_params(sc, sc->mbox, sc->pf, 0, 1, &param, &val);
if (rc)
vi->rss_base = 0xffff;
else {
cxgbe(4): Do not access the mailbox without appropriate locks while creating hardware VIs. This fixes a bad race on systems with hw.cxgbe.num_vis > 1. Reported by: olivier@ MFC after: 1 week Sponsored by: Chelsio Communications
2017-08-28 22:41:15 +00:00
MPASS((val >> 16) == vi->rss_size);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
vi->rss_base = val & 0xffff;
}
cxgbe(4): Do not access the mailbox without appropriate locks while creating hardware VIs. This fixes a bad race on systems with hw.cxgbe.num_vis > 1. Reported by: olivier@ MFC after: 1 week Sponsored by: Chelsio Communications
2017-08-28 22:41:15 +00:00
return (0);
}
static int
vcxgbe_attach(device_t dev)
{
struct vi_info *vi;
struct port_info *pi;
struct adapter *sc;
int rc;
vi = device_get_softc(dev);
pi = vi->pi;
sc = pi->adapter;
rc = begin_synchronized_op(sc, vi, SLEEP_OK | INTR_OK, "t4via");
if (rc)
return (rc);
rc = alloc_extra_vi(sc, pi, vi);
end_synchronized_op(sc, 0);
if (rc)
return (rc);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
rc = cxgbe_vi_attach(dev, vi);
if (rc) {
t4_free_vi(sc, sc->mbox, sc->pf, 0, vi->viid);
return (rc);
}
return (0);
}
static int
vcxgbe_detach(device_t dev)
{
struct vi_info *vi;
struct adapter *sc;
vi = device_get_softc(dev);
sc = vi->pi->adapter;
doom_vi(sc, vi);
cxgbe_vi_detach(vi);
t4_free_vi(sc, sc->mbox, sc->pf, 0, vi->viid);
end_synchronized_op(sc, 0);
return (0);
}
cxgbe(4): Delay the panic due to a fatal error by 30s. This lets information logged by the interrupt handler reach the system log before the system goes down.
2019-02-09 01:49:53 +00:00
static struct callout fatal_callout;
static void
delayed_panic(void *arg)
{
struct adapter *sc = arg;
panic("%s: panic on fatal error", device_get_nameunit(sc->dev));
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
void
cxgbe(4): Improved error reporting and diagnostics. "slow" interrupt handler: - Expand the list of INT_CAUSE registers known to the driver. - Add decode information for many more bits but decouple it from the rest of intr_info so that it is entirely optional. - Call t4_fatal_err exactly once, and from the top level PL intr handler. t4_fatal_err: - Use t4_shutdown_adapter from the common code to stop the adapter. - Stop servicing slow interrupts after the first fatal one. Driver/firmware interaction: - CH_DUMP_MBOX: note whether the mailbox being dumped is a command or a reply or something else. - Log the raw value of pcie_fw for some errors. - Use correct log levels (debug vs. error). Sponsored by: Chelsio Communications
2019-02-01 20:42:49 +00:00
t4_fatal_err(struct adapter *sc, bool fw_error)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
{
cxgbe(4): Improved error reporting and diagnostics. "slow" interrupt handler: - Expand the list of INT_CAUSE registers known to the driver. - Add decode information for many more bits but decouple it from the rest of intr_info so that it is entirely optional. - Call t4_fatal_err exactly once, and from the top level PL intr handler. t4_fatal_err: - Use t4_shutdown_adapter from the common code to stop the adapter. - Stop servicing slow interrupts after the first fatal one. Driver/firmware interaction: - CH_DUMP_MBOX: note whether the mailbox being dumped is a command or a reply or something else. - Log the raw value of pcie_fw for some errors. - Use correct log levels (debug vs. error). Sponsored by: Chelsio Communications
2019-02-01 20:42:49 +00:00
t4_shutdown_adapter(sc);
log(LOG_ALERT, "%s: encountered fatal error, adapter stopped.\n",
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
device_get_nameunit(sc->dev));
cxgbe(4): Improved error reporting and diagnostics. "slow" interrupt handler: - Expand the list of INT_CAUSE registers known to the driver. - Add decode information for many more bits but decouple it from the rest of intr_info so that it is entirely optional. - Call t4_fatal_err exactly once, and from the top level PL intr handler. t4_fatal_err: - Use t4_shutdown_adapter from the common code to stop the adapter. - Stop servicing slow interrupts after the first fatal one. Driver/firmware interaction: - CH_DUMP_MBOX: note whether the mailbox being dumped is a command or a reply or something else. - Log the raw value of pcie_fw for some errors. - Use correct log levels (debug vs. error). Sponsored by: Chelsio Communications
2019-02-01 20:42:49 +00:00
if (fw_error) {
ASSERT_SYNCHRONIZED_OP(sc);
sc->flags |= ADAP_ERR;
} else {
ADAPTER_LOCK(sc);
sc->flags |= ADAP_ERR;
ADAPTER_UNLOCK(sc);
}
cxgbe(4): Delay the panic due to a fatal error by 30s. This lets information logged by the interrupt handler reach the system log before the system goes down.
2019-02-09 01:49:53 +00:00
if (t4_panic_on_fatal_err) {
log(LOG_ALERT, "%s: panic on fatal error after 30s",
device_get_nameunit(sc->dev));
callout_reset(&fatal_callout, hz * 30, delayed_panic, sc);
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
void
t4_add_adapter(struct adapter *sc)
{
sx_xlock(&t4_list_lock);
SLIST_INSERT_HEAD(&t4_list, sc, link);
sx_xunlock(&t4_list_lock);
}
int
t4_map_bars_0_and_4(struct adapter *sc)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
{
sc->regs_rid = PCIR_BAR(0);
sc->regs_res = bus_alloc_resource_any(sc->dev, SYS_RES_MEMORY,
&sc->regs_rid, RF_ACTIVE);
if (sc->regs_res == NULL) {
device_printf(sc->dev, "cannot map registers.\n");
return (ENXIO);
}
sc->bt = rman_get_bustag(sc->regs_res);
sc->bh = rman_get_bushandle(sc->regs_res);
sc->mmio_len = rman_get_size(sc->regs_res);
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
setbit(&sc->doorbells, DOORBELL_KDB);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
sc->msix_rid = PCIR_BAR(4);
sc->msix_res = bus_alloc_resource_any(sc->dev, SYS_RES_MEMORY,
&sc->msix_rid, RF_ACTIVE);
if (sc->msix_res == NULL) {
device_printf(sc->dev, "cannot map MSI-X BAR.\n");
return (ENXIO);
}
return (0);
}
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
int
t4_map_bar_2(struct adapter *sc)
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
{
/*
* T4: only iWARP driver uses the userspace doorbells. There is no need
* to map it if RDMA is disabled.
*/
if (is_t4(sc) && sc->rdmacaps == 0)
return (0);
sc->udbs_rid = PCIR_BAR(2);
sc->udbs_res = bus_alloc_resource_any(sc->dev, SYS_RES_MEMORY,
&sc->udbs_rid, RF_ACTIVE);
if (sc->udbs_res == NULL) {
device_printf(sc->dev, "cannot map doorbell BAR.\n");
return (ENXIO);
}
sc->udbs_base = rman_get_virtual(sc->udbs_res);
cxgbe(4): Deal with the slightly different SGE_STAT_CFG in T6. Sponsored by: Chelsio Communications
2016-09-11 17:57:53 +00:00
if (chip_id(sc) >= CHELSIO_T5) {
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
setbit(&sc->doorbells, DOORBELL_UDB);
#if defined(__i386__) || defined(__amd64__)
if (t5_write_combine) {
cxgbe(4): Deal with the slightly different SGE_STAT_CFG in T6. Sponsored by: Chelsio Communications
2016-09-11 17:57:53 +00:00
int rc, mode;
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
/*
* Enable write combining on BAR2. This is the
* userspace doorbell BAR and is split into 128B
* (UDBS_SEG_SIZE) doorbell regions, each associated
* with an egress queue. The first 64B has the doorbell
* and the second 64B can be used to submit a tx work
* request with an implicit doorbell.
*/
rc = pmap_change_attr((vm_offset_t)sc->udbs_base,
rman_get_size(sc->udbs_res), PAT_WRITE_COMBINING);
if (rc == 0) {
clrbit(&sc->doorbells, DOORBELL_UDB);
Cosmetic change (s/wrwc/wcwr/;s/WRWC/WCWR/). MFC after: 3 days.
2013-04-11 22:49:29 +00:00
setbit(&sc->doorbells, DOORBELL_WCWR);
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
setbit(&sc->doorbells, DOORBELL_UDBWC);
} else {
device_printf(sc->dev,
"couldn't enable write combining: %d\n",
rc);
}
cxgbe(4): Deal with the slightly different SGE_STAT_CFG in T6. Sponsored by: Chelsio Communications
2016-09-11 17:57:53 +00:00
mode = is_t5(sc) ? V_STATMODE(0) : V_T6_STATMODE(0);
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
t4_write_reg(sc, A_SGE_STAT_CFG,
cxgbe(4): Deal with the slightly different SGE_STAT_CFG in T6. Sponsored by: Chelsio Communications
2016-09-11 17:57:53 +00:00
V_STATSOURCE_T5(7) | mode);
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
}
#endif
}
cxgbe(4): Disable write-combined doorbells by default. This had been the default behavior but was changed accidentally as part of the recent iw_cxgbe+OFED overhaul. Fix another bug in that change while here: the global knob affects all the adapters in the system and should be left alone by per-adapter code. MFC after: 3 days Sponsored by: Chelsio Communications
2018-05-10 06:33:54 +00:00
sc->iwt.wc_en = isset(&sc->doorbells, DOORBELL_UDBWC) ? 1 : 0;
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
return (0);
}
cxgbe(4): Add general purpose routines that offer safe access to the chip's memory windows. Convert existing users of these windows to the new routines.
2016-03-10 06:15:31 +00:00
struct memwin_init {
uint32_t base;
uint32_t aperture;
};
static const struct memwin_init t4_memwin[NUM_MEMWIN] = {
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
{ MEMWIN0_BASE, MEMWIN0_APERTURE },
{ MEMWIN1_BASE, MEMWIN1_APERTURE },
{ MEMWIN2_BASE_T4, MEMWIN2_APERTURE_T4 }
};
cxgbe(4): Add general purpose routines that offer safe access to the chip's memory windows. Convert existing users of these windows to the new routines.
2016-03-10 06:15:31 +00:00
static const struct memwin_init t5_memwin[NUM_MEMWIN] = {
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
{ MEMWIN0_BASE, MEMWIN0_APERTURE },
{ MEMWIN1_BASE, MEMWIN1_APERTURE },
{ MEMWIN2_BASE_T5, MEMWIN2_APERTURE_T5 },
};
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
static void
setup_memwin(struct adapter *sc)
{
cxgbe(4): Add general purpose routines that offer safe access to the chip's memory windows. Convert existing users of these windows to the new routines.
2016-03-10 06:15:31 +00:00
const struct memwin_init *mw_init;
struct memwin *mw;
int i;
Allow cxgbe(4) running within a VM to attach to its devices that have been exported via PCI passthrough. - Do not check for a specific physical function (PF) before claiming a device. Different PFs have different device-ids so this check is redundant anyway. - Obtain the PF# from the WHOAMI register instead of pci_get_function(). - Setup the memory windows using the real BAR0 address, not what the VM says it is. Obtained from: Chelsio Communications
2012-06-26 00:34:34 +00:00
uint32_t bar0;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
if (is_t4(sc)) {
/*
* Read low 32b of bar0 indirectly via the hardware backdoor
* mechanism. Works from within PCI passthrough environments
* too, where rman_get_start() can return a different value. We
* need to program the T4 memory window decoders with the actual
* addresses that will be coming across the PCIe link.
*/
bar0 = t4_hw_pci_read_cfg4(sc, PCIR_BAR(0));
bar0 &= (uint32_t) PCIM_BAR_MEM_BASE;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Add general purpose routines that offer safe access to the chip's memory windows. Convert existing users of these windows to the new routines.
2016-03-10 06:15:31 +00:00
mw_init = &t4_memwin[0];
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
} else {
cxgbe(4): Add general purpose routines that offer safe access to the chip's memory windows. Convert existing users of these windows to the new routines.
2016-03-10 06:15:31 +00:00
/* T5+ use the relative offset inside the PCIe BAR */
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
bar0 = 0;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Add general purpose routines that offer safe access to the chip's memory windows. Convert existing users of these windows to the new routines.
2016-03-10 06:15:31 +00:00
mw_init = &t5_memwin[0];
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Add general purpose routines that offer safe access to the chip's memory windows. Convert existing users of these windows to the new routines.
2016-03-10 06:15:31 +00:00
for (i = 0, mw = &sc->memwin[0]; i < NUM_MEMWIN; i++, mw_init++, mw++) {
rw_init(&mw->mw_lock, "memory window access");
mw->mw_base = mw_init->base;
mw->mw_aperture = mw_init->aperture;
mw->mw_curpos = 0;
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
t4_write_reg(sc,
PCIE_MEM_ACCESS_REG(A_PCIE_MEM_ACCESS_BASE_WIN, i),
cxgbe(4): Add general purpose routines that offer safe access to the chip's memory windows. Convert existing users of these windows to the new routines.
2016-03-10 06:15:31 +00:00
(mw->mw_base + bar0) | V_BIR(0) |
V_WINDOW(ilog2(mw->mw_aperture) - 10));
rw_wlock(&mw->mw_lock);
position_memwin(sc, i, 0);
rw_wunlock(&mw->mw_lock);
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
}
Allow cxgbe(4) running within a VM to attach to its devices that have been exported via PCI passthrough. - Do not check for a specific physical function (PF) before claiming a device. Different PFs have different device-ids so this check is redundant anyway. - Obtain the PF# from the WHOAMI register instead of pci_get_function(). - Setup the memory windows using the real BAR0 address, not what the VM says it is. Obtained from: Chelsio Communications
2012-06-26 00:34:34 +00:00
/* flush */
t4_read_reg(sc, PCIE_MEM_ACCESS_REG(A_PCIE_MEM_ACCESS_BASE_WIN, 2));
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
cxgbe(4): Add general purpose routines that offer safe access to the chip's memory windows. Convert existing users of these windows to the new routines.
2016-03-10 06:15:31 +00:00
/*
* Positions the memory window at the given address in the card's address space.
* There are some alignment requirements and the actual position may be at an
* address prior to the requested address. mw->mw_curpos always has the actual
* position of the window.
*/
static void
position_memwin(struct adapter *sc, int idx, uint32_t addr)
{
struct memwin *mw;
uint32_t pf;
uint32_t reg;
MPASS(idx >= 0 && idx < NUM_MEMWIN);
mw = &sc->memwin[idx];
rw_assert(&mw->mw_lock, RA_WLOCKED);
if (is_t4(sc)) {
pf = 0;
mw->mw_curpos = addr & ~0xf; /* start must be 16B aligned */
} else {
pf = V_PFNUM(sc->pf);
mw->mw_curpos = addr & ~0x7f; /* start must be 128B aligned */
}
reg = PCIE_MEM_ACCESS_REG(A_PCIE_MEM_ACCESS_OFFSET, idx);
t4_write_reg(sc, reg, mw->mw_curpos | pf);
t4_read_reg(sc, reg); /* flush */
}
cxgbe: Implement tcp_info handler for connections handled by t4_tom. The TCB is read using a memory window right now. A better alternate to get self-consistent, uncached information would be to use a GET_TCB request but waiting for a reply from hw while holding non-sleepable locks is quite inconvenient. Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D14817
2018-04-03 01:22:15 +00:00
int
cxgbe(4): Add general purpose routines that offer safe access to the chip's memory windows. Convert existing users of these windows to the new routines.
2016-03-10 06:15:31 +00:00
rw_via_memwin(struct adapter *sc, int idx, uint32_t addr, uint32_t *val,
int len, int rw)
{
struct memwin *mw;
uint32_t mw_end, v;
MPASS(idx >= 0 && idx < NUM_MEMWIN);
/* Memory can only be accessed in naturally aligned 4 byte units */
if (addr & 3 || len & 3 || len <= 0)
return (EINVAL);
mw = &sc->memwin[idx];
while (len > 0) {
rw_rlock(&mw->mw_lock);
mw_end = mw->mw_curpos + mw->mw_aperture;
cxgbe(4): Fix bug in r296603. The memory window needs to be repositioned if the start address isn't in the window already. One of the bounds check used the end address instead.
2016-03-10 20:36:32 +00:00
if (addr >= mw_end || addr < mw->mw_curpos) {
cxgbe(4): Add general purpose routines that offer safe access to the chip's memory windows. Convert existing users of these windows to the new routines.
2016-03-10 06:15:31 +00:00
/* Will need to reposition the window */
if (!rw_try_upgrade(&mw->mw_lock)) {
rw_runlock(&mw->mw_lock);
rw_wlock(&mw->mw_lock);
}
rw_assert(&mw->mw_lock, RA_WLOCKED);
position_memwin(sc, idx, addr);
rw_downgrade(&mw->mw_lock);
mw_end = mw->mw_curpos + mw->mw_aperture;
}
rw_assert(&mw->mw_lock, RA_RLOCKED);
while (addr < mw_end && len > 0) {
if (rw == 0) {
v = t4_read_reg(sc, mw->mw_base + addr -
mw->mw_curpos);
*val++ = le32toh(v);
} else {
v = *val++;
t4_write_reg(sc, mw->mw_base + addr -
Cleanup unnecessary semicolons from the kernel. Found with devel/coccinelle.
2016-04-10 23:07:00 +00:00
mw->mw_curpos, htole32(v));
cxgbe(4): Add general purpose routines that offer safe access to the chip's memory windows. Convert existing users of these windows to the new routines.
2016-03-10 06:15:31 +00:00
}
addr += 4;
len -= 4;
}
rw_runlock(&mw->mw_lock);
}
return (0);
}
cxgbe(4): Break up alloc_tid_tabs and move the atid routines to the base NIC driver. The atid services will be used by new features (hashfilters and inline TLS) that do not involve TOE. Sponsored by: Chelsio Communications
2018-04-26 19:00:35 +00:00
int
alloc_atid_tab(struct tid_info *t, int flags)
{
int i;
MPASS(t->natids > 0);
MPASS(t->atid_tab == NULL);
t->atid_tab = malloc(t->natids * sizeof(*t->atid_tab), M_CXGBE,
M_ZERO | flags);
if (t->atid_tab == NULL)
return (ENOMEM);
mtx_init(&t->atid_lock, "atid lock", NULL, MTX_DEF);
t->afree = t->atid_tab;
t->atids_in_use = 0;
for (i = 1; i < t->natids; i++)
t->atid_tab[i - 1].next = &t->atid_tab[i];
t->atid_tab[t->natids - 1].next = NULL;
return (0);
}
void
free_atid_tab(struct tid_info *t)
{
KASSERT(t->atids_in_use == 0,
("%s: %d atids still in use.", __func__, t->atids_in_use));
if (mtx_initialized(&t->atid_lock))
mtx_destroy(&t->atid_lock);
free(t->atid_tab, M_CXGBE);
t->atid_tab = NULL;
}
int
alloc_atid(struct adapter *sc, void *ctx)
{
struct tid_info *t = &sc->tids;
int atid = -1;
mtx_lock(&t->atid_lock);
if (t->afree) {
union aopen_entry *p = t->afree;
atid = p - t->atid_tab;
cxgbe(4): Convert ACT_OPEN_RPL to a shared CPL. Reserve 3b in the 14b atid to identify the owner and use it to dispatch the CPL. This allows all CPLs that use an atid to be used as shared CPLs, although ACT_OPEN_RPL is the only one being converted in this revision. Sponsored by: Chelsio Communications
2018-04-30 21:47:30 +00:00
MPASS(atid <= M_TID_TID);
cxgbe(4): Break up alloc_tid_tabs and move the atid routines to the base NIC driver. The atid services will be used by new features (hashfilters and inline TLS) that do not involve TOE. Sponsored by: Chelsio Communications
2018-04-26 19:00:35 +00:00
t->afree = p->next;
p->data = ctx;
t->atids_in_use++;
}
mtx_unlock(&t->atid_lock);
return (atid);
}
void *
lookup_atid(struct adapter *sc, int atid)
{
struct tid_info *t = &sc->tids;
return (t->atid_tab[atid].data);
}
void
free_atid(struct adapter *sc, int atid)
{
struct tid_info *t = &sc->tids;
union aopen_entry *p = &t->atid_tab[atid];
mtx_lock(&t->atid_lock);
p->next = t->afree;
t->afree = p;
t->atids_in_use--;
mtx_unlock(&t->atid_lock);
}
cxgbe(4): Move release_tid to the base NIC driver for future consumers. Sponsored by: Chelsio Communications.
2018-04-26 22:04:21 +00:00
static void
queue_tid_release(struct adapter *sc, int tid)
{
CXGBE_UNIMPLEMENTED("deferred tid release");
}
void
release_tid(struct adapter *sc, int tid, struct sge_wrq *ctrlq)
{
struct wrqe *wr;
struct cpl_tid_release *req;
wr = alloc_wrqe(sizeof(*req), ctrlq);
if (wr == NULL) {
queue_tid_release(sc, tid); /* defer */
return;
}
req = wrtod(wr);
INIT_TP_WR_MIT_CPL(req, CPL_TID_RELEASE, tid);
t4_wrq_tx(sc, wr);
}
cxgbe(4): Allow the addr/len pair that is being validated in validate_mem_range to span multiple memory types. Update validate_mt_off_len to use validate_mem_range.
2016-03-10 02:43:10 +00:00
static int
t4_range_cmp(const void *a, const void *b)
{
return ((const struct t4_range *)a)->start -
((const struct t4_range *)b)->start;
}
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
/*
cxgbe(4): Allow the addr/len pair that is being validated in validate_mem_range to span multiple memory types. Update validate_mt_off_len to use validate_mem_range.
2016-03-10 02:43:10 +00:00
* Verify that the memory range specified by the addr/len pair is valid within
* the card's address space.
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
*/
static int
Treat the memory lengths for CHELSIO_T4_GET_MEM as unsigned. Previously attempts to read the MC region were failing since the length was greater than 2^31. Reviewed by: np MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D17857
2018-11-06 22:33:36 +00:00
validate_mem_range(struct adapter *sc, uint32_t addr, uint32_t len)
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
{
cxgbe(4): Allow the addr/len pair that is being validated in validate_mem_range to span multiple memory types. Update validate_mt_off_len to use validate_mem_range.
2016-03-10 02:43:10 +00:00
struct t4_range mem_ranges[4], *r, *next;
uint32_t em, addr_len;
int i, n, remaining;
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
/* Memory can only be accessed in naturally aligned 4 byte units */
Treat the memory lengths for CHELSIO_T4_GET_MEM as unsigned. Previously attempts to read the MC region were failing since the length was greater than 2^31. Reviewed by: np MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D17857
2018-11-06 22:33:36 +00:00
if (addr & 3 || len & 3 || len == 0)
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
return (EINVAL);
/* Enabled memories */
em = t4_read_reg(sc, A_MA_TARGET_MEM_ENABLE);
cxgbe(4): Allow the addr/len pair that is being validated in validate_mem_range to span multiple memory types. Update validate_mt_off_len to use validate_mem_range.
2016-03-10 02:43:10 +00:00
r = &mem_ranges[0];
n = 0;
bzero(r, sizeof(mem_ranges));
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
if (em & F_EDRAM0_ENABLE) {
addr_len = t4_read_reg(sc, A_MA_EDRAM0_BAR);
cxgbe(4): Allow the addr/len pair that is being validated in validate_mem_range to span multiple memory types. Update validate_mt_off_len to use validate_mem_range.
2016-03-10 02:43:10 +00:00
r->size = G_EDRAM0_SIZE(addr_len) << 20;
if (r->size > 0) {
r->start = G_EDRAM0_BASE(addr_len) << 20;
if (addr >= r->start &&
addr + len <= r->start + r->size)
return (0);
r++;
n++;
}
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
}
if (em & F_EDRAM1_ENABLE) {
addr_len = t4_read_reg(sc, A_MA_EDRAM1_BAR);
cxgbe(4): Allow the addr/len pair that is being validated in validate_mem_range to span multiple memory types. Update validate_mt_off_len to use validate_mem_range.
2016-03-10 02:43:10 +00:00
r->size = G_EDRAM1_SIZE(addr_len) << 20;
if (r->size > 0) {
r->start = G_EDRAM1_BASE(addr_len) << 20;
if (addr >= r->start &&
addr + len <= r->start + r->size)
return (0);
r++;
n++;
}
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
}
if (em & F_EXT_MEM_ENABLE) {
addr_len = t4_read_reg(sc, A_MA_EXT_MEMORY_BAR);
cxgbe(4): Allow the addr/len pair that is being validated in validate_mem_range to span multiple memory types. Update validate_mt_off_len to use validate_mem_range.
2016-03-10 02:43:10 +00:00
r->size = G_EXT_MEM_SIZE(addr_len) << 20;
if (r->size > 0) {
r->start = G_EXT_MEM_BASE(addr_len) << 20;
if (addr >= r->start &&
addr + len <= r->start + r->size)
return (0);
r++;
n++;
}
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
}
cxgbe(4): Allow the addr/len pair that is being validated in validate_mem_range to span multiple memory types. Update validate_mt_off_len to use validate_mem_range.
2016-03-10 02:43:10 +00:00
if (is_t5(sc) && em & F_EXT_MEM1_ENABLE) {
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
addr_len = t4_read_reg(sc, A_MA_EXT_MEMORY1_BAR);
cxgbe(4): Allow the addr/len pair that is being validated in validate_mem_range to span multiple memory types. Update validate_mt_off_len to use validate_mem_range.
2016-03-10 02:43:10 +00:00
r->size = G_EXT_MEM1_SIZE(addr_len) << 20;
if (r->size > 0) {
r->start = G_EXT_MEM1_BASE(addr_len) << 20;
if (addr >= r->start &&
addr + len <= r->start + r->size)
return (0);
r++;
n++;
}
}
MPASS(n <= nitems(mem_ranges));
if (n > 1) {
/* Sort and merge the ranges. */
qsort(mem_ranges, n, sizeof(struct t4_range), t4_range_cmp);
/* Start from index 0 and examine the next n - 1 entries. */
r = &mem_ranges[0];
for (remaining = n - 1; remaining > 0; remaining--, r++) {
MPASS(r->size > 0); /* r is a valid entry. */
next = r + 1;
MPASS(next->size > 0); /* and so is the next one. */
while (r->start + r->size >= next->start) {
/* Merge the next one into the current entry. */
r->size = max(r->start + r->size,
next->start + next->size) - r->start;
n--; /* One fewer entry in total. */
if (--remaining == 0)
goto done; /* short circuit */
next++;
}
if (next != r + 1) {
/*
* Some entries were merged into r and next
* points to the first valid entry that couldn't
* be merged.
*/
MPASS(next->size > 0); /* must be valid */
memcpy(r + 1, next, remaining * sizeof(*r));
#ifdef INVARIANTS
/*
* This so that the foo->size assertion in the
* next iteration of the loop do the right
* thing for entries that were pulled up and are
* no longer valid.
*/
MPASS(n < nitems(mem_ranges));
bzero(&mem_ranges[n], (nitems(mem_ranges) - n) *
sizeof(struct t4_range));
#endif
}
}
done:
/* Done merging the ranges. */
MPASS(n > 0);
r = &mem_ranges[0];
for (i = 0; i < n; i++, r++) {
if (addr >= r->start &&
addr + len <= r->start + r->size)
return (0);
}
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
}
return (EFAULT);
}
cxgbe(4): Update T4 and T5 firmwares to 1.9.12.0
2013-10-14 21:25:07 +00:00
static int
fwmtype_to_hwmtype(int mtype)
{
switch (mtype) {
case FW_MEMTYPE_EDC0:
return (MEM_EDC0);
case FW_MEMTYPE_EDC1:
return (MEM_EDC1);
case FW_MEMTYPE_EXTMEM:
return (MEM_MC0);
case FW_MEMTYPE_EXTMEM1:
return (MEM_MC1);
default:
panic("%s: cannot translate fw mtype %d.", __func__, mtype);
}
}
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
/*
* Verify that the memory range specified by the memtype/offset/len pair is
* valid and lies entirely within the memtype specified. The global address of
* the start of the range is returned in addr.
*/
static int
Treat the memory lengths for CHELSIO_T4_GET_MEM as unsigned. Previously attempts to read the MC region were failing since the length was greater than 2^31. Reviewed by: np MFC after: 2 months Differential Revision: https://reviews.freebsd.org/D17857
2018-11-06 22:33:36 +00:00
validate_mt_off_len(struct adapter *sc, int mtype, uint32_t off, uint32_t len,
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
uint32_t *addr)
{
cxgbe(4): Allow the addr/len pair that is being validated in validate_mem_range to span multiple memory types. Update validate_mt_off_len to use validate_mem_range.
2016-03-10 02:43:10 +00:00
uint32_t em, addr_len, maddr;
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
/* Memory can only be accessed in naturally aligned 4 byte units */
if (off & 3 || len & 3 || len == 0)
return (EINVAL);
em = t4_read_reg(sc, A_MA_TARGET_MEM_ENABLE);
cxgbe(4): Update T4 and T5 firmwares to 1.9.12.0
2013-10-14 21:25:07 +00:00
switch (fwmtype_to_hwmtype(mtype)) {
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
case MEM_EDC0:
if (!(em & F_EDRAM0_ENABLE))
return (EINVAL);
addr_len = t4_read_reg(sc, A_MA_EDRAM0_BAR);
maddr = G_EDRAM0_BASE(addr_len) << 20;
break;
case MEM_EDC1:
if (!(em & F_EDRAM1_ENABLE))
return (EINVAL);
addr_len = t4_read_reg(sc, A_MA_EDRAM1_BAR);
maddr = G_EDRAM1_BASE(addr_len) << 20;
break;
case MEM_MC:
if (!(em & F_EXT_MEM_ENABLE))
return (EINVAL);
addr_len = t4_read_reg(sc, A_MA_EXT_MEMORY_BAR);
maddr = G_EXT_MEM_BASE(addr_len) << 20;
break;
case MEM_MC1:
cxgbe(4): Allow the addr/len pair that is being validated in validate_mem_range to span multiple memory types. Update validate_mt_off_len to use validate_mem_range.
2016-03-10 02:43:10 +00:00
if (!is_t5(sc) || !(em & F_EXT_MEM1_ENABLE))
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
return (EINVAL);
addr_len = t4_read_reg(sc, A_MA_EXT_MEMORY1_BAR);
maddr = G_EXT_MEM1_BASE(addr_len) << 20;
break;
default:
return (EINVAL);
}
cxgbe(4): Allow the addr/len pair that is being validated in validate_mem_range to span multiple memory types. Update validate_mt_off_len to use validate_mem_range.
2016-03-10 02:43:10 +00:00
*addr = maddr + off; /* global address */
return (validate_mem_range(sc, *addr, len));
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
}
cxgbe(4): Improvements to the code that deals with the firmware's log. - Query the location of the log very early during attach. Refresh the location later after establishing contact with the firmware. - Save the log's location as a flat address in devlog_params. - Use a memory window instead of backdoor access to the EDC/MC to read the log.
2016-03-10 23:17:26 +00:00
static int
fixup_devlog_params(struct adapter *sc)
{
struct devlog_params *dparams = &sc->params.devlog;
int rc;
rc = validate_mt_off_len(sc, dparams->memtype, dparams->start,
dparams->size, &dparams->addr);
return (rc);
}
cxgbe(4): Do not forward interrupts to queues with freelists. This leaves the firmware event queue (fwq) as the only queue that can take interrupts for others. This simplifies cfg_itype_and_nqueues and queue allocation in the driver at the cost of a little (never?) used configuration. It also allows service_iq to be split into two specialized variants in the future. MFC after: 2 months Sponsored by: Chelsio Communications
2017-12-22 19:10:19 +00:00
static void
update_nirq(struct intrs_and_queues *iaq, int nports)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
{
cxgbe(4): Do not forward interrupts to queues with freelists. This leaves the firmware event queue (fwq) as the only queue that can take interrupts for others. This simplifies cfg_itype_and_nqueues and queue allocation in the driver at the cost of a little (never?) used configuration. It also allows service_iq to be split into two specialized variants in the future. MFC after: 2 months Sponsored by: Chelsio Communications
2017-12-22 19:10:19 +00:00
int extra = T4_EXTRA_INTR;
iaq->nirq = extra;
iaq->nirq += nports * (iaq->nrxq + iaq->nofldrxq);
iaq->nirq += nports * (iaq->num_vis - 1) *
max(iaq->nrxq_vi, iaq->nnmrxq_vi);
iaq->nirq += nports * (iaq->num_vis - 1) * iaq->nofldrxq_vi;
}
/*
* Adjust requirements to fit the number of interrupts available.
*/
static void
calculate_iaq(struct adapter *sc, struct intrs_and_queues *iaq, int itype,
int navail)
{
int old_nirq;
const int nports = sc->params.nports;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
MPASS(nports > 0);
cxgbe(4): Do not forward interrupts to queues with freelists. This leaves the firmware event queue (fwq) as the only queue that can take interrupts for others. This simplifies cfg_itype_and_nqueues and queue allocation in the driver at the cost of a little (never?) used configuration. It also allows service_iq to be split into two specialized variants in the future. MFC after: 2 months Sponsored by: Chelsio Communications
2017-12-22 19:10:19 +00:00
MPASS(navail > 0);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
bzero(iaq, sizeof(*iaq));
cxgbe(4): Do not forward interrupts to queues with freelists. This leaves the firmware event queue (fwq) as the only queue that can take interrupts for others. This simplifies cfg_itype_and_nqueues and queue allocation in the driver at the cost of a little (never?) used configuration. It also allows service_iq to be split into two specialized variants in the future. MFC after: 2 months Sponsored by: Chelsio Communications
2017-12-22 19:10:19 +00:00
iaq->intr_type = itype;
cxgbe(4): Sanitize t4_num_vis during MOD_LOAD like all other t4_* tunables. Add num_vis to the intrs_and_queues structure as it affects the number of interrupts requested and queues created. In future cfg_itype_and_nqueues might lower it incrementally instead of going straight to 1 when enough interrupts aren't available. Sponsored by: Chelsio Communications
2017-11-16 01:33:53 +00:00
iaq->num_vis = t4_num_vis;
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
iaq->ntxq = t4_ntxq;
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
iaq->ntxq_vi = t4_ntxq_vi;
cxgbe(4): Do not forward interrupts to queues with freelists. This leaves the firmware event queue (fwq) as the only queue that can take interrupts for others. This simplifies cfg_itype_and_nqueues and queue allocation in the driver at the cost of a little (never?) used configuration. It also allows service_iq to be split into two specialized variants in the future. MFC after: 2 months Sponsored by: Chelsio Communications
2017-12-22 19:10:19 +00:00
iaq->nrxq = t4_nrxq;
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
iaq->nrxq_vi = t4_nrxq_vi;
cxgbe(4): Allocate offload Tx queues when a card has resources provisioned for NIC_ETHOFLD and the kernel has option RATELIMIT. It is possible to use the chip's offload queues for normal NIC Tx and not just TOE Tx. The difference is that these queues support out of order processing of work requests and have a per-"flowid" mechanism for tracking credits between the driver and hardware. This allows Tx for any number of flows bound to different rate limits to be submitted to a single Tx queue and the work requests for slow flows won't cause HOL blocking for the rest. Sponsored by: Chelsio Communications
2018-05-17 01:42:18 +00:00
#if defined(TCP_OFFLOAD) || defined(RATELIMIT)
if (is_offload(sc) || is_ethoffload(sc)) {
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
iaq->nofldtxq = t4_nofldtxq;
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
iaq->nofldtxq_vi = t4_nofldtxq_vi;
cxgbe(4): Allocate offload Tx queues when a card has resources provisioned for NIC_ETHOFLD and the kernel has option RATELIMIT. It is possible to use the chip's offload queues for normal NIC Tx and not just TOE Tx. The difference is that these queues support out of order processing of work requests and have a per-"flowid" mechanism for tracking credits between the driver and hardware. This allows Tx for any number of flows bound to different rate limits to be submitted to a single Tx queue and the work requests for slow flows won't cause HOL blocking for the rest. Sponsored by: Chelsio Communications
2018-05-17 01:42:18 +00:00
}
#endif
#ifdef TCP_OFFLOAD
if (is_offload(sc)) {
cxgbe(4): Do not forward interrupts to queues with freelists. This leaves the firmware event queue (fwq) as the only queue that can take interrupts for others. This simplifies cfg_itype_and_nqueues and queue allocation in the driver at the cost of a little (never?) used configuration. It also allows service_iq to be split into two specialized variants in the future. MFC after: 2 months Sponsored by: Chelsio Communications
2017-12-22 19:10:19 +00:00
iaq->nofldrxq = t4_nofldrxq;
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
iaq->nofldrxq_vi = t4_nofldrxq_vi;
Do not allocate extra vectors when adapter is not TOE capable (or toecaps have been disallowed by the user). + one very minor unrelated cleanup in t4_sge.c
2012-06-22 22:59:42 +00:00
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
#endif
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
#ifdef DEV_NETMAP
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
iaq->nnmtxq_vi = t4_nnmtxq_vi;
iaq->nnmrxq_vi = t4_nnmrxq_vi;
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
#endif
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Do not forward interrupts to queues with freelists. This leaves the firmware event queue (fwq) as the only queue that can take interrupts for others. This simplifies cfg_itype_and_nqueues and queue allocation in the driver at the cost of a little (never?) used configuration. It also allows service_iq to be split into two specialized variants in the future. MFC after: 2 months Sponsored by: Chelsio Communications
2017-12-22 19:10:19 +00:00
update_nirq(iaq, nports);
if (iaq->nirq <= navail &&
(itype != INTR_MSI || powerof2(iaq->nirq))) {
/*
* This is the normal case -- there are enough interrupts for
* everything.
*/
goto done;
}
/*
* If extra VIs have been configured try reducing their count and see if
* that works.
*/
while (iaq->num_vis > 1) {
iaq->num_vis--;
update_nirq(iaq, nports);
if (iaq->nirq <= navail &&
(itype != INTR_MSI || powerof2(iaq->nirq))) {
device_printf(sc->dev, "virtual interfaces per port "
"reduced to %d from %d. nrxq=%u, nofldrxq=%u, "
"nrxq_vi=%u nofldrxq_vi=%u, nnmrxq_vi=%u. "
"itype %d, navail %u, nirq %d.\n",
iaq->num_vis, t4_num_vis, iaq->nrxq, iaq->nofldrxq,
iaq->nrxq_vi, iaq->nofldrxq_vi, iaq->nnmrxq_vi,
itype, navail, iaq->nirq);
goto done;
}
}
/*
* Extra VIs will not be created. Log a message if they were requested.
*/
MPASS(iaq->num_vis == 1);
iaq->ntxq_vi = iaq->nrxq_vi = 0;
iaq->nofldtxq_vi = iaq->nofldrxq_vi = 0;
iaq->nnmtxq_vi = iaq->nnmrxq_vi = 0;
if (iaq->num_vis != t4_num_vis) {
device_printf(sc->dev, "extra virtual interfaces disabled. "
"nrxq=%u, nofldrxq=%u, nrxq_vi=%u nofldrxq_vi=%u, "
"nnmrxq_vi=%u. itype %d, navail %u, nirq %d.\n",
iaq->nrxq, iaq->nofldrxq, iaq->nrxq_vi, iaq->nofldrxq_vi,
iaq->nnmrxq_vi, itype, navail, iaq->nirq);
}
/*
* Keep reducing the number of NIC rx queues to the next lower power of
* 2 (for even RSS distribution) and halving the TOE rx queues and see
* if that works.
*/
do {
if (iaq->nrxq > 1) {
do {
iaq->nrxq--;
} while (!powerof2(iaq->nrxq));
}
if (iaq->nofldrxq > 1)
iaq->nofldrxq >>= 1;
old_nirq = iaq->nirq;
update_nirq(iaq, nports);
if (iaq->nirq <= navail &&
(itype != INTR_MSI || powerof2(iaq->nirq))) {
device_printf(sc->dev, "running with reduced number of "
"rx queues because of shortage of interrupts. "
"nrxq=%u, nofldrxq=%u. "
"itype %d, navail %u, nirq %d.\n", iaq->nrxq,
iaq->nofldrxq, itype, navail, iaq->nirq);
goto done;
}
} while (old_nirq != iaq->nirq);
/* One interrupt for everything. Ugh. */
device_printf(sc->dev, "running with minimal number of queues. "
"itype %d, navail %u.\n", itype, navail);
iaq->nirq = 1;
MPASS(iaq->nrxq == 1);
iaq->ntxq = 1;
if (iaq->nofldrxq > 1)
iaq->nofldtxq = 1;
done:
MPASS(iaq->num_vis > 0);
if (iaq->num_vis > 1) {
MPASS(iaq->nrxq_vi > 0);
MPASS(iaq->ntxq_vi > 0);
}
MPASS(iaq->nirq > 0);
MPASS(iaq->nrxq > 0);
MPASS(iaq->ntxq > 0);
if (itype == INTR_MSI) {
MPASS(powerof2(iaq->nirq));
}
}
static int
cfg_itype_and_nqueues(struct adapter *sc, struct intrs_and_queues *iaq)
{
int rc, itype, navail, nalloc;
Do not over-allocate MSI interrupts for the case where each ingress queue has its own interrupt. If the exact number that we need is not a power of 2 and we're using MSI, then switch to interrupt multiplexing. While here, replace the magic numbers with something more readable. MFC after: 3 days
2011-03-24 01:03:01 +00:00
for (itype = INTR_MSIX; itype; itype >>= 1) {
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
if ((itype & t4_intr_types) == 0)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
continue; /* not allowed */
Do not over-allocate MSI interrupts for the case where each ingress queue has its own interrupt. If the exact number that we need is not a power of 2 and we're using MSI, then switch to interrupt multiplexing. While here, replace the magic numbers with something more readable. MFC after: 3 days
2011-03-24 01:03:01 +00:00
if (itype == INTR_MSIX)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
navail = pci_msix_count(sc->dev);
Do not over-allocate MSI interrupts for the case where each ingress queue has its own interrupt. If the exact number that we need is not a power of 2 and we're using MSI, then switch to interrupt multiplexing. While here, replace the magic numbers with something more readable. MFC after: 3 days
2011-03-24 01:03:01 +00:00
else if (itype == INTR_MSI)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
navail = pci_msi_count(sc->dev);
else
navail = 1;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
restart:
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (navail == 0)
continue;
cxgbe(4): Do not forward interrupts to queues with freelists. This leaves the firmware event queue (fwq) as the only queue that can take interrupts for others. This simplifies cfg_itype_and_nqueues and queue allocation in the driver at the cost of a little (never?) used configuration. It also allows service_iq to be split into two specialized variants in the future. MFC after: 2 months Sponsored by: Chelsio Communications
2017-12-22 19:10:19 +00:00
calculate_iaq(sc, iaq, itype, navail);
nalloc = iaq->nirq;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
rc = 0;
Do not over-allocate MSI interrupts for the case where each ingress queue has its own interrupt. If the exact number that we need is not a power of 2 and we're using MSI, then switch to interrupt multiplexing. While here, replace the magic numbers with something more readable. MFC after: 3 days
2011-03-24 01:03:01 +00:00
if (itype == INTR_MSIX)
cxgbe(4): Do not forward interrupts to queues with freelists. This leaves the firmware event queue (fwq) as the only queue that can take interrupts for others. This simplifies cfg_itype_and_nqueues and queue allocation in the driver at the cost of a little (never?) used configuration. It also allows service_iq to be split into two specialized variants in the future. MFC after: 2 months Sponsored by: Chelsio Communications
2017-12-22 19:10:19 +00:00
rc = pci_alloc_msix(sc->dev, &nalloc);
Do not over-allocate MSI interrupts for the case where each ingress queue has its own interrupt. If the exact number that we need is not a power of 2 and we're using MSI, then switch to interrupt multiplexing. While here, replace the magic numbers with something more readable. MFC after: 3 days
2011-03-24 01:03:01 +00:00
else if (itype == INTR_MSI)
cxgbe(4): Do not forward interrupts to queues with freelists. This leaves the firmware event queue (fwq) as the only queue that can take interrupts for others. This simplifies cfg_itype_and_nqueues and queue allocation in the driver at the cost of a little (never?) used configuration. It also allows service_iq to be split into two specialized variants in the future. MFC after: 2 months Sponsored by: Chelsio Communications
2017-12-22 19:10:19 +00:00
rc = pci_alloc_msi(sc->dev, &nalloc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Do not forward interrupts to queues with freelists. This leaves the firmware event queue (fwq) as the only queue that can take interrupts for others. This simplifies cfg_itype_and_nqueues and queue allocation in the driver at the cost of a little (never?) used configuration. It also allows service_iq to be split into two specialized variants in the future. MFC after: 2 months Sponsored by: Chelsio Communications
2017-12-22 19:10:19 +00:00
if (rc == 0 && nalloc > 0) {
if (nalloc == iaq->nirq)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
return (0);
/*
* Didn't get the number requested. Use whatever number
cxgbe(4): Do not forward interrupts to queues with freelists. This leaves the firmware event queue (fwq) as the only queue that can take interrupts for others. This simplifies cfg_itype_and_nqueues and queue allocation in the driver at the cost of a little (never?) used configuration. It also allows service_iq to be split into two specialized variants in the future. MFC after: 2 months Sponsored by: Chelsio Communications
2017-12-22 19:10:19 +00:00
* the kernel is willing to allocate.
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
*/
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
device_printf(sc->dev, "fewer vectors than requested, "
"type=%d, req=%d, rcvd=%d; will downshift req.\n",
cxgbe(4): Do not forward interrupts to queues with freelists. This leaves the firmware event queue (fwq) as the only queue that can take interrupts for others. This simplifies cfg_itype_and_nqueues and queue allocation in the driver at the cost of a little (never?) used configuration. It also allows service_iq to be split into two specialized variants in the future. MFC after: 2 months Sponsored by: Chelsio Communications
2017-12-22 19:10:19 +00:00
itype, iaq->nirq, nalloc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
pci_release_msi(sc->dev);
cxgbe(4): Do not forward interrupts to queues with freelists. This leaves the firmware event queue (fwq) as the only queue that can take interrupts for others. This simplifies cfg_itype_and_nqueues and queue allocation in the driver at the cost of a little (never?) used configuration. It also allows service_iq to be split into two specialized variants in the future. MFC after: 2 months Sponsored by: Chelsio Communications
2017-12-22 19:10:19 +00:00
navail = nalloc;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
goto restart;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
device_printf(sc->dev,
"failed to allocate vectors:%d, type=%d, req=%d, rcvd=%d\n",
cxgbe(4): Do not forward interrupts to queues with freelists. This leaves the firmware event queue (fwq) as the only queue that can take interrupts for others. This simplifies cfg_itype_and_nqueues and queue allocation in the driver at the cost of a little (never?) used configuration. It also allows service_iq to be split into two specialized variants in the future. MFC after: 2 months Sponsored by: Chelsio Communications
2017-12-22 19:10:19 +00:00
itype, rc, iaq->nirq, nalloc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
device_printf(sc->dev,
"failed to find a usable interrupt type. "
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
"allowed=%d, msi-x=%d, msi=%d, intx=1", t4_intr_types,
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
pci_msix_count(sc->dev), pci_msi_count(sc->dev));
return (ENXIO);
}
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
#define FW_VERSION(chip) ( \
- Include the T5 firmware with the driver. - Update the T4 firmware to the latest. - Minor reorganization and updates to the version macros, etc. Obtained from: Chelsio MFC after: 1 day
2013-07-03 23:52:15 +00:00
V_FW_HDR_FW_VER_MAJOR(chip##FW_VERSION_MAJOR) | \
V_FW_HDR_FW_VER_MINOR(chip##FW_VERSION_MINOR) | \
V_FW_HDR_FW_VER_MICRO(chip##FW_VERSION_MICRO) | \
V_FW_HDR_FW_VER_BUILD(chip##FW_VERSION_BUILD))
#define FW_INTFVER(chip, intf) (chip##FW_HDR_INTFVER_##intf)
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
cxgbe(4): Use a truncated firmware header for version checks. All the version numbers are towards the begining of the header. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 17:58:06 +00:00
/* Just enough of fw_hdr to cover all version info. */
struct fw_h {
__u8 ver;
__u8 chip;
__be16 len512;
__be32 fw_ver;
__be32 tp_microcode_ver;
__u8 intfver_nic;
__u8 intfver_vnic;
__u8 intfver_ofld;
__u8 intfver_ri;
__u8 intfver_iscsipdu;
__u8 intfver_iscsi;
__u8 intfver_fcoepdu;
__u8 intfver_fcoe;
};
/* Spot check a couple of fields. */
CTASSERT(offsetof(struct fw_h, fw_ver) == offsetof(struct fw_hdr, fw_ver));
CTASSERT(offsetof(struct fw_h, intfver_nic) == offsetof(struct fw_hdr, intfver_nic));
CTASSERT(offsetof(struct fw_h, intfver_fcoe) == offsetof(struct fw_hdr, intfver_fcoe));
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
struct fw_info {
uint8_t chip;
char *kld_name;
char *fw_mod_name;
cxgbe(4): Use a truncated firmware header for version checks. All the version numbers are towards the begining of the header. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 17:58:06 +00:00
struct fw_h fw_h;
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
} fw_info[] = {
{
.chip = CHELSIO_T4,
.kld_name = "t4fw_cfg",
.fw_mod_name = "t4fw",
cxgbe(4): Use a truncated firmware header for version checks. All the version numbers are towards the begining of the header. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 17:58:06 +00:00
.fw_h = {
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
.chip = FW_HDR_CHIP_T4,
cxgbe(4): Remove homemade version of htobe32 from the driver. It was needed only for ia64 where it was implemented as a call to bswapXX, which was always a real function. htobeXX with a constant argument is calculated at compile-time everywhere else. MFC after: 1 week Sponsored by: Chelsio Communications
2018-06-12 06:46:03 +00:00
.fw_ver = htobe32(FW_VERSION(T4)),
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
.intfver_nic = FW_INTFVER(T4, NIC),
.intfver_vnic = FW_INTFVER(T4, VNIC),
.intfver_ofld = FW_INTFVER(T4, OFLD),
.intfver_ri = FW_INTFVER(T4, RI),
.intfver_iscsipdu = FW_INTFVER(T4, ISCSIPDU),
.intfver_iscsi = FW_INTFVER(T4, ISCSI),
.intfver_fcoepdu = FW_INTFVER(T4, FCOEPDU),
.intfver_fcoe = FW_INTFVER(T4, FCOE),
},
}, {
.chip = CHELSIO_T5,
.kld_name = "t5fw_cfg",
.fw_mod_name = "t5fw",
cxgbe(4): Use a truncated firmware header for version checks. All the version numbers are towards the begining of the header. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 17:58:06 +00:00
.fw_h = {
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
.chip = FW_HDR_CHIP_T5,
cxgbe(4): Remove homemade version of htobe32 from the driver. It was needed only for ia64 where it was implemented as a call to bswapXX, which was always a real function. htobeXX with a constant argument is calculated at compile-time everywhere else. MFC after: 1 week Sponsored by: Chelsio Communications
2018-06-12 06:46:03 +00:00
.fw_ver = htobe32(FW_VERSION(T5)),
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
.intfver_nic = FW_INTFVER(T5, NIC),
.intfver_vnic = FW_INTFVER(T5, VNIC),
.intfver_ofld = FW_INTFVER(T5, OFLD),
.intfver_ri = FW_INTFVER(T5, RI),
.intfver_iscsipdu = FW_INTFVER(T5, ISCSIPDU),
.intfver_iscsi = FW_INTFVER(T5, ISCSI),
.intfver_fcoepdu = FW_INTFVER(T5, FCOEPDU),
.intfver_fcoe = FW_INTFVER(T5, FCOE),
},
cxgbe(4): Attach to cards with the Terminator 6 ASIC. T6 cards will come up as 't6nex' nexus devices with 'cc' ports hanging off them. The T6 firmware and configuration files will be added as soon as they are released. For now the driver will try to work with whatever firmware and configuration is on the card's flash. Sponsored by: Chelsio Communications
2016-09-16 00:08:37 +00:00
}, {
.chip = CHELSIO_T6,
.kld_name = "t6fw_cfg",
.fw_mod_name = "t6fw",
cxgbe(4): Use a truncated firmware header for version checks. All the version numbers are towards the begining of the header. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 17:58:06 +00:00
.fw_h = {
cxgbe(4): Attach to cards with the Terminator 6 ASIC. T6 cards will come up as 't6nex' nexus devices with 'cc' ports hanging off them. The T6 firmware and configuration files will be added as soon as they are released. For now the driver will try to work with whatever firmware and configuration is on the card's flash. Sponsored by: Chelsio Communications
2016-09-16 00:08:37 +00:00
.chip = FW_HDR_CHIP_T6,
cxgbe(4): Remove homemade version of htobe32 from the driver. It was needed only for ia64 where it was implemented as a call to bswapXX, which was always a real function. htobeXX with a constant argument is calculated at compile-time everywhere else. MFC after: 1 week Sponsored by: Chelsio Communications
2018-06-12 06:46:03 +00:00
.fw_ver = htobe32(FW_VERSION(T6)),
cxgbe(4): Attach to cards with the Terminator 6 ASIC. T6 cards will come up as 't6nex' nexus devices with 'cc' ports hanging off them. The T6 firmware and configuration files will be added as soon as they are released. For now the driver will try to work with whatever firmware and configuration is on the card's flash. Sponsored by: Chelsio Communications
2016-09-16 00:08:37 +00:00
.intfver_nic = FW_INTFVER(T6, NIC),
.intfver_vnic = FW_INTFVER(T6, VNIC),
.intfver_ofld = FW_INTFVER(T6, OFLD),
.intfver_ri = FW_INTFVER(T6, RI),
.intfver_iscsipdu = FW_INTFVER(T6, ISCSIPDU),
.intfver_iscsi = FW_INTFVER(T6, ISCSI),
.intfver_fcoepdu = FW_INTFVER(T6, FCOEPDU),
.intfver_fcoe = FW_INTFVER(T6, FCOE),
},
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
}
};
static struct fw_info *
find_fw_info(int chip)
{
int i;
for (i = 0; i < nitems(fw_info); i++) {
if (fw_info[i].chip == chip)
return (&fw_info[i]);
}
return (NULL);
}
cxgbe(4): Consider all the API versions of the interfaces exported by the firmware (instead of just the main firmware version) when evaluating firmware compatibility. Document the new "hw.cxgbe.fw_install" knob being introduced here. This should fix kern/173584 too. Setting hw.cxgbe.fw_install=2 will mostly do what was requested in the PR but it's a bit more intelligent in that it won't reinstall the same firmware repeatedly if the knob is left set. PR: kern/173584 MFC after: 5 days
2013-02-26 20:35:54 +00:00
/*
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
* Is the given firmware API compatible with the one the driver was compiled
* with?
cxgbe(4): Consider all the API versions of the interfaces exported by the firmware (instead of just the main firmware version) when evaluating firmware compatibility. Document the new "hw.cxgbe.fw_install" knob being introduced here. This should fix kern/173584 too. Setting hw.cxgbe.fw_install=2 will mostly do what was requested in the PR but it's a bit more intelligent in that it won't reinstall the same firmware repeatedly if the knob is left set. PR: kern/173584 MFC after: 5 days
2013-02-26 20:35:54 +00:00
*/
static int
cxgbe(4): Use a truncated firmware header for version checks. All the version numbers are towards the begining of the header. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 17:58:06 +00:00
fw_compatible(const struct fw_h *hdr1, const struct fw_h *hdr2)
cxgbe(4): Consider all the API versions of the interfaces exported by the firmware (instead of just the main firmware version) when evaluating firmware compatibility. Document the new "hw.cxgbe.fw_install" knob being introduced here. This should fix kern/173584 too. Setting hw.cxgbe.fw_install=2 will mostly do what was requested in the PR but it's a bit more intelligent in that it won't reinstall the same firmware repeatedly if the knob is left set. PR: kern/173584 MFC after: 5 days
2013-02-26 20:35:54 +00:00
{
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
/* short circuit if it's the exact same firmware version */
if (hdr1->chip == hdr2->chip && hdr1->fw_ver == hdr2->fw_ver)
cxgbe(4): Consider all the API versions of the interfaces exported by the firmware (instead of just the main firmware version) when evaluating firmware compatibility. Document the new "hw.cxgbe.fw_install" knob being introduced here. This should fix kern/173584 too. Setting hw.cxgbe.fw_install=2 will mostly do what was requested in the PR but it's a bit more intelligent in that it won't reinstall the same firmware repeatedly if the knob is left set. PR: kern/173584 MFC after: 5 days
2013-02-26 20:35:54 +00:00
return (1);
/*
* XXX: Is this too conservative? Perhaps I should limit this to the
* features that are supported in the driver.
*/
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
#define SAME_INTF(x) (hdr1->intfver_##x == hdr2->intfver_##x)
if (hdr1->chip == hdr2->chip && SAME_INTF(nic) && SAME_INTF(vnic) &&
SAME_INTF(ofld) && SAME_INTF(ri) && SAME_INTF(iscsipdu) &&
SAME_INTF(iscsi) && SAME_INTF(fcoepdu) && SAME_INTF(fcoe))
cxgbe(4): Consider all the API versions of the interfaces exported by the firmware (instead of just the main firmware version) when evaluating firmware compatibility. Document the new "hw.cxgbe.fw_install" knob being introduced here. This should fix kern/173584 too. Setting hw.cxgbe.fw_install=2 will mostly do what was requested in the PR but it's a bit more intelligent in that it won't reinstall the same firmware repeatedly if the knob is left set. PR: kern/173584 MFC after: 5 days
2013-02-26 20:35:54 +00:00
return (1);
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
#undef SAME_INTF
cxgbe(4): Consider all the API versions of the interfaces exported by the firmware (instead of just the main firmware version) when evaluating firmware compatibility. Document the new "hw.cxgbe.fw_install" knob being introduced here. This should fix kern/173584 too. Setting hw.cxgbe.fw_install=2 will mostly do what was requested in the PR but it's a bit more intelligent in that it won't reinstall the same firmware repeatedly if the knob is left set. PR: kern/173584 MFC after: 5 days
2013-02-26 20:35:54 +00:00
return (0);
}
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
static int
load_fw_module(struct adapter *sc, const struct firmware **dcfg,
const struct firmware **fw)
{
struct fw_info *fw_info;
*dcfg = NULL;
if (fw != NULL)
*fw = NULL;
fw_info = find_fw_info(chip_id(sc));
if (fw_info == NULL) {
device_printf(sc->dev,
"unable to look up firmware information for chip %d.\n",
chip_id(sc));
return (EINVAL);
}
*dcfg = firmware_get(fw_info->kld_name);
if (*dcfg != NULL) {
if (fw != NULL)
*fw = firmware_get(fw_info->fw_mod_name);
return (0);
}
return (ENOENT);
}
static void
unload_fw_module(struct adapter *sc, const struct firmware *dcfg,
const struct firmware *fw)
{
if (fw != NULL)
firmware_put(fw, FIRMWARE_UNLOAD);
if (dcfg != NULL)
firmware_put(dcfg, FIRMWARE_UNLOAD);
}
- Explain clearly why a different firmware is being installed (if/when it is being installed). Improve other error messages while here. - Select special FPGA specific configuration profile when appropriate. MFC after: 3 days
2013-04-11 19:39:40 +00:00
/*
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
* Return values:
* 0 means no firmware install attempted.
* ERESTART means a firmware install was attempted and was successful.
* +ve errno means a firmware install was attempted but failed.
- Explain clearly why a different firmware is being installed (if/when it is being installed). Improve other error messages while here. - Select special FPGA specific configuration profile when appropriate. MFC after: 3 days
2013-04-11 19:39:40 +00:00
*/
static int
cxgbe(4): Use a truncated firmware header for version checks. All the version numbers are towards the begining of the header. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 17:58:06 +00:00
install_kld_firmware(struct adapter *sc, struct fw_h *card_fw,
const struct fw_h *drv_fw, const char *reason, int *already)
- Explain clearly why a different firmware is being installed (if/when it is being installed). Improve other error messages while here. - Select special FPGA specific configuration profile when appropriate. MFC after: 3 days
2013-04-11 19:39:40 +00:00
{
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
const struct firmware *cfg, *fw;
const uint32_t c = be32toh(card_fw->fw_ver);
cxgbe(4): Allow negative values in hw.cxgbe.fw_install and take them to mean that the driver should taste the firmware in the KLD and use that firmware's version for all its fw_install checks. The driver gets firmware version information from compiled-in values by default and this change allows custom (or older/newer) firmware modules to be used with the stock driver. There is no change in default behavior. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 18:42:16 +00:00
uint32_t d, k;
int rc, fw_install;
struct fw_h bundled_fw;
bool load_attempted;
cfg = fw = NULL;
load_attempted = false;
fw_install = t4_fw_install < 0 ? -t4_fw_install : t4_fw_install;
- Explain clearly why a different firmware is being installed (if/when it is being installed). Improve other error messages while here. - Select special FPGA specific configuration profile when appropriate. MFC after: 3 days
2013-04-11 19:39:40 +00:00
cxgbe(4): Allow negative values in hw.cxgbe.fw_install and take them to mean that the driver should taste the firmware in the KLD and use that firmware's version for all its fw_install checks. The driver gets firmware version information from compiled-in values by default and this change allows custom (or older/newer) firmware modules to be used with the stock driver. There is no change in default behavior. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 18:42:16 +00:00
memcpy(&bundled_fw, drv_fw, sizeof(bundled_fw));
if (t4_fw_install < 0) {
rc = load_fw_module(sc, &cfg, &fw);
if (rc != 0 || fw == NULL) {
device_printf(sc->dev,
"failed to load firmware module: %d. cfg %p, fw %p;"
" will use compiled-in firmware version for"
"hw.cxgbe.fw_install checks.\n",
rc, cfg, fw);
} else {
memcpy(&bundled_fw, fw->data, sizeof(bundled_fw));
}
load_attempted = true;
}
d = be32toh(bundled_fw.fw_ver);
cxgbe(4): Make sure bundled_fw is always initialized before use. This fixes a bug that prevented the driver from auto-flashing the firmware when it didn't see one on the card. This feature was introduced in r321390 and this bug was introduced in r343269. Reported by: gallatin@ MFC after: 1 week Sponsored by: Chelsio Communications
2019-04-22 17:00:30 +00:00
if (reason != NULL)
goto install;
if ((sc->flags & FW_OK) == 0) {
if (c == 0xffffffff) {
reason = "missing";
goto install;
}
rc = 0;
goto done;
}
cxgbe(4): Allow negative values in hw.cxgbe.fw_install and take them to mean that the driver should taste the firmware in the KLD and use that firmware's version for all its fw_install checks. The driver gets firmware version information from compiled-in values by default and this change allows custom (or older/newer) firmware modules to be used with the stock driver. There is no change in default behavior. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 18:42:16 +00:00
if (!fw_compatible(card_fw, &bundled_fw)) {
- Explain clearly why a different firmware is being installed (if/when it is being installed). Improve other error messages while here. - Select special FPGA specific configuration profile when appropriate. MFC after: 3 days
2013-04-11 19:39:40 +00:00
reason = "incompatible or unusable";
goto install;
}
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
if (d > c) {
- Explain clearly why a different firmware is being installed (if/when it is being installed). Improve other error messages while here. - Select special FPGA specific configuration profile when appropriate. MFC after: 3 days
2013-04-11 19:39:40 +00:00
reason = "older than the version bundled with this driver";
goto install;
}
cxgbe(4): Allow negative values in hw.cxgbe.fw_install and take them to mean that the driver should taste the firmware in the KLD and use that firmware's version for all its fw_install checks. The driver gets firmware version information from compiled-in values by default and this change allows custom (or older/newer) firmware modules to be used with the stock driver. There is no change in default behavior. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 18:42:16 +00:00
if (fw_install == 2 && d != c) {
- Explain clearly why a different firmware is being installed (if/when it is being installed). Improve other error messages while here. - Select special FPGA specific configuration profile when appropriate. MFC after: 3 days
2013-04-11 19:39:40 +00:00
reason = "different than the version bundled with this driver";
goto install;
}
cxgbe(4): Allow negative values in hw.cxgbe.fw_install and take them to mean that the driver should taste the firmware in the KLD and use that firmware's version for all its fw_install checks. The driver gets firmware version information from compiled-in values by default and this change allows custom (or older/newer) firmware modules to be used with the stock driver. There is no change in default behavior. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 18:42:16 +00:00
/* No reason to do anything to the firmware already on the card. */
rc = 0;
goto done;
- Explain clearly why a different firmware is being installed (if/when it is being installed). Improve other error messages while here. - Select special FPGA specific configuration profile when appropriate. MFC after: 3 days
2013-04-11 19:39:40 +00:00
install:
cxgbe(4): Allow negative values in hw.cxgbe.fw_install and take them to mean that the driver should taste the firmware in the KLD and use that firmware's version for all its fw_install checks. The driver gets firmware version information from compiled-in values by default and this change allows custom (or older/newer) firmware modules to be used with the stock driver. There is no change in default behavior. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 18:42:16 +00:00
rc = 0;
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
if ((*already)++)
cxgbe(4): Allow negative values in hw.cxgbe.fw_install and take them to mean that the driver should taste the firmware in the KLD and use that firmware's version for all its fw_install checks. The driver gets firmware version information from compiled-in values by default and this change allows custom (or older/newer) firmware modules to be used with the stock driver. There is no change in default behavior. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 18:42:16 +00:00
goto done;
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
cxgbe(4): Allow negative values in hw.cxgbe.fw_install and take them to mean that the driver should taste the firmware in the KLD and use that firmware's version for all its fw_install checks. The driver gets firmware version information from compiled-in values by default and this change allows custom (or older/newer) firmware modules to be used with the stock driver. There is no change in default behavior. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 18:42:16 +00:00
if (fw_install == 0) {
cxgbe(4): Never install a firmware if hw.cxgbe.fw_install is 0. MFC after: 1 week
2013-06-05 20:57:52 +00:00
device_printf(sc->dev, "firmware on card (%u.%u.%u.%u) is %s, "
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
"but the driver is prohibited from installing a firmware "
"on the card.\n",
cxgbe(4): Never install a firmware if hw.cxgbe.fw_install is 0. MFC after: 1 week
2013-06-05 20:57:52 +00:00
G_FW_HDR_FW_VER_MAJOR(c), G_FW_HDR_FW_VER_MINOR(c),
G_FW_HDR_FW_VER_MICRO(c), G_FW_HDR_FW_VER_BUILD(c), reason);
cxgbe(4): Allow negative values in hw.cxgbe.fw_install and take them to mean that the driver should taste the firmware in the KLD and use that firmware's version for all its fw_install checks. The driver gets firmware version information from compiled-in values by default and this change allows custom (or older/newer) firmware modules to be used with the stock driver. There is no change in default behavior. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 18:42:16 +00:00
goto done;
cxgbe(4): Never install a firmware if hw.cxgbe.fw_install is 0. MFC after: 1 week
2013-06-05 20:57:52 +00:00
}
cxgbe(4): Allow negative values in hw.cxgbe.fw_install and take them to mean that the driver should taste the firmware in the KLD and use that firmware's version for all its fw_install checks. The driver gets firmware version information from compiled-in values by default and this change allows custom (or older/newer) firmware modules to be used with the stock driver. There is no change in default behavior. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 18:42:16 +00:00
/*
* We'll attempt to install a firmware. Load the module first (if it
* hasn't been loaded already).
*/
if (!load_attempted) {
rc = load_fw_module(sc, &cfg, &fw);
if (rc != 0 || fw == NULL) {
device_printf(sc->dev,
"failed to load firmware module: %d. cfg %p, fw %p\n",
rc, cfg, fw);
/* carry on */
}
}
if (fw == NULL) {
device_printf(sc->dev, "firmware on card (%u.%u.%u.%u) is %s, "
"but the driver cannot take corrective action because it "
"is unable to load the firmware module.\n",
G_FW_HDR_FW_VER_MAJOR(c), G_FW_HDR_FW_VER_MINOR(c),
G_FW_HDR_FW_VER_MICRO(c), G_FW_HDR_FW_VER_BUILD(c), reason);
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
rc = sc->flags & FW_OK ? 0 : ENOENT;
goto done;
}
k = be32toh(((const struct fw_hdr *)fw->data)->fw_ver);
if (k != d) {
cxgbe(4): Allow negative values in hw.cxgbe.fw_install and take them to mean that the driver should taste the firmware in the KLD and use that firmware's version for all its fw_install checks. The driver gets firmware version information from compiled-in values by default and this change allows custom (or older/newer) firmware modules to be used with the stock driver. There is no change in default behavior. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 18:42:16 +00:00
MPASS(t4_fw_install > 0);
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
device_printf(sc->dev,
"firmware in KLD (%u.%u.%u.%u) is not what the driver was "
cxgbe(4): Allow negative values in hw.cxgbe.fw_install and take them to mean that the driver should taste the firmware in the KLD and use that firmware's version for all its fw_install checks. The driver gets firmware version information from compiled-in values by default and this change allows custom (or older/newer) firmware modules to be used with the stock driver. There is no change in default behavior. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 18:42:16 +00:00
"expecting (%u.%u.%u.%u) and will not be used.\n",
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
G_FW_HDR_FW_VER_MAJOR(k), G_FW_HDR_FW_VER_MINOR(k),
cxgbe(4): Allow negative values in hw.cxgbe.fw_install and take them to mean that the driver should taste the firmware in the KLD and use that firmware's version for all its fw_install checks. The driver gets firmware version information from compiled-in values by default and this change allows custom (or older/newer) firmware modules to be used with the stock driver. There is no change in default behavior. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 18:42:16 +00:00
G_FW_HDR_FW_VER_MICRO(k), G_FW_HDR_FW_VER_BUILD(k),
G_FW_HDR_FW_VER_MAJOR(d), G_FW_HDR_FW_VER_MINOR(d),
G_FW_HDR_FW_VER_MICRO(d), G_FW_HDR_FW_VER_BUILD(d));
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
rc = sc->flags & FW_OK ? 0 : EINVAL;
goto done;
}
- Explain clearly why a different firmware is being installed (if/when it is being installed). Improve other error messages while here. - Select special FPGA specific configuration profile when appropriate. MFC after: 3 days
2013-04-11 19:39:40 +00:00
cxgbe(4): Allow negative values in hw.cxgbe.fw_install and take them to mean that the driver should taste the firmware in the KLD and use that firmware's version for all its fw_install checks. The driver gets firmware version information from compiled-in values by default and this change allows custom (or older/newer) firmware modules to be used with the stock driver. There is no change in default behavior. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 18:42:16 +00:00
device_printf(sc->dev, "firmware on card (%u.%u.%u.%u) is %s, "
"installing firmware %u.%u.%u.%u on card.\n",
G_FW_HDR_FW_VER_MAJOR(c), G_FW_HDR_FW_VER_MINOR(c),
G_FW_HDR_FW_VER_MICRO(c), G_FW_HDR_FW_VER_BUILD(c), reason,
G_FW_HDR_FW_VER_MAJOR(d), G_FW_HDR_FW_VER_MINOR(d),
G_FW_HDR_FW_VER_MICRO(d), G_FW_HDR_FW_VER_BUILD(d));
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
rc = -t4_fw_upgrade(sc, sc->mbox, fw->data, fw->datasize, 0);
if (rc != 0) {
device_printf(sc->dev, "failed to install firmware: %d\n", rc);
} else {
/* Installed successfully, update the cached header too. */
rc = ERESTART;
memcpy(card_fw, fw->data, sizeof(*card_fw));
}
done:
unload_fw_module(sc, cfg, fw);
return (rc);
- Explain clearly why a different firmware is being installed (if/when it is being installed). Improve other error messages while here. - Select special FPGA specific configuration profile when appropriate. MFC after: 3 days
2013-04-11 19:39:40 +00:00
}
cxgbe(4): Install the firmware bundled with the driver to the card if it doesn't seem to have one. This lets the driver recover automatically from incomplete firmware upgrades (panic, reboot, power loss, etc. in the middle of an upgrade). MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-23 18:10:47 +00:00
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
/*
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
* Establish contact with the firmware and attempt to become the master driver.
*
* A firmware will be installed to the card if needed (if the driver is allowed
* to do so).
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
*/
static int
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
contact_firmware(struct adapter *sc)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
{
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
int rc, already = 0;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
enum dev_state state;
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
struct fw_info *fw_info;
struct fw_hdr *card_fw; /* fw on the card */
cxgbe(4): Use a truncated firmware header for version checks. All the version numbers are towards the begining of the header. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 17:58:06 +00:00
const struct fw_h *drv_fw;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
fw_info = find_fw_info(chip_id(sc));
if (fw_info == NULL) {
device_printf(sc->dev,
"unable to look up firmware information for chip %d.\n",
chip_id(sc));
return (EINVAL);
}
cxgbe(4): Use a truncated firmware header for version checks. All the version numbers are towards the begining of the header. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 17:58:06 +00:00
drv_fw = &fw_info->fw_h;
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
cxgbe(4): Consider all the API versions of the interfaces exported by the firmware (instead of just the main firmware version) when evaluating firmware compatibility. Document the new "hw.cxgbe.fw_install" knob being introduced here. This should fix kern/173584 too. Setting hw.cxgbe.fw_install=2 will mostly do what was requested in the PR but it's a bit more intelligent in that it won't reinstall the same firmware repeatedly if the knob is left set. PR: kern/173584 MFC after: 5 days
2013-02-26 20:35:54 +00:00
/* Read the header of the firmware on the card */
card_fw = malloc(sizeof(*card_fw), M_CXGBE, M_ZERO | M_WAITOK);
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
restart:
rc = -t4_get_fw_hdr(sc, card_fw);
if (rc != 0) {
cxgbe(4): Consider all the API versions of the interfaces exported by the firmware (instead of just the main firmware version) when evaluating firmware compatibility. Document the new "hw.cxgbe.fw_install" knob being introduced here. This should fix kern/173584 too. Setting hw.cxgbe.fw_install=2 will mostly do what was requested in the PR but it's a bit more intelligent in that it won't reinstall the same firmware repeatedly if the knob is left set. PR: kern/173584 MFC after: 5 days
2013-02-26 20:35:54 +00:00
device_printf(sc->dev,
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
"unable to read firmware header from card's flash: %d\n",
rc);
goto done;
cxgbe(4): Consider all the API versions of the interfaces exported by the firmware (instead of just the main firmware version) when evaluating firmware compatibility. Document the new "hw.cxgbe.fw_install" knob being introduced here. This should fix kern/173584 too. Setting hw.cxgbe.fw_install=2 will mostly do what was requested in the PR but it's a bit more intelligent in that it won't reinstall the same firmware repeatedly if the knob is left set. PR: kern/173584 MFC after: 5 days
2013-02-26 20:35:54 +00:00
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Use a truncated firmware header for version checks. All the version numbers are towards the begining of the header. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 17:58:06 +00:00
rc = install_kld_firmware(sc, (struct fw_h *)card_fw, drv_fw, NULL,
&already);
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
if (rc == ERESTART)
goto restart;
if (rc != 0)
goto done;
cxgbe(4): Install the firmware bundled with the driver to the card if it doesn't seem to have one. This lets the driver recover automatically from incomplete firmware upgrades (panic, reboot, power loss, etc. in the middle of an upgrade). MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-23 18:10:47 +00:00
rc = t4_fw_hello(sc, sc->mbox, sc->mbox, MASTER_MAY, &state);
if (rc < 0 || state == DEV_STATE_ERR) {
rc = -rc;
device_printf(sc->dev,
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
"failed to connect to the firmware: %d, %d. "
"PCIE_FW 0x%08x\n", rc, state, t4_read_reg(sc, A_PCIE_FW));
#if 0
cxgbe(4): Use a truncated firmware header for version checks. All the version numbers are towards the begining of the header. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 17:58:06 +00:00
if (install_kld_firmware(sc, (struct fw_h *)card_fw, drv_fw,
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
"not responding properly to HELLO", &already) == ERESTART)
goto restart;
#endif
cxgbe(4): Install the firmware bundled with the driver to the card if it doesn't seem to have one. This lets the driver recover automatically from incomplete firmware upgrades (panic, reboot, power loss, etc. in the middle of an upgrade). MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-23 18:10:47 +00:00
goto done;
}
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
MPASS(be32toh(card_fw->flags) & FW_HDR_FLAGS_RESET_HALT);
sc->flags |= FW_OK; /* The firmware responded to the FW_HELLO. */
if (rc == sc->pf) {
cxgbe(4): Install the firmware bundled with the driver to the card if it doesn't seem to have one. This lets the driver recover automatically from incomplete firmware upgrades (panic, reboot, power loss, etc. in the middle of an upgrade). MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-23 18:10:47 +00:00
sc->flags |= MASTER_PF;
cxgbe(4): Use a truncated firmware header for version checks. All the version numbers are towards the begining of the header. MFC after: 1 week Sponsored by: Chelsio Communications
2019-01-21 17:58:06 +00:00
rc = install_kld_firmware(sc, (struct fw_h *)card_fw, drv_fw,
NULL, &already);
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
if (rc == ERESTART)
rc = 0;
else if (rc != 0)
goto done;
} else if (state == DEV_STATE_UNINIT) {
cxgbe(4): Install the firmware bundled with the driver to the card if it doesn't seem to have one. This lets the driver recover automatically from incomplete firmware upgrades (panic, reboot, power loss, etc. in the middle of an upgrade). MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-23 18:10:47 +00:00
/*
* We didn't get to be the master so we definitely won't be
* configuring the chip. It's a bug if someone else hasn't
* configured it already.
*/
device_printf(sc->dev, "couldn't be master(%d), "
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
"device not already initialized either(%d). "
"PCIE_FW 0x%08x\n", rc, state, t4_read_reg(sc, A_PCIE_FW));
cxgbe(4): Install the firmware bundled with the driver to the card if it doesn't seem to have one. This lets the driver recover automatically from incomplete firmware upgrades (panic, reboot, power loss, etc. in the middle of an upgrade). MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-23 18:10:47 +00:00
rc = EPROTO;
goto done;
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
} else {
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
/*
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
* Some other PF is the master and has configured the chip.
* This is allowed but untested.
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
*/
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
device_printf(sc->dev, "PF%d is master, device state %d. "
"PCIE_FW 0x%08x\n", rc, state, t4_read_reg(sc, A_PCIE_FW));
snprintf(sc->cfg_file, sizeof(sc->cfg_file), "pf%d", rc);
sc->cfcsum = 0;
rc = 0;
}
done:
if (rc != 0 && sc->flags & FW_OK) {
t4_fw_bye(sc, sc->mbox);
sc->flags &= ~FW_OK;
}
free(card_fw, M_CXGBE);
return (rc);
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
static int
copy_cfg_file_to_card(struct adapter *sc, char *cfg_file,
uint32_t mtype, uint32_t moff)
{
struct fw_info *fw_info;
const struct firmware *dcfg, *rcfg = NULL;
const uint32_t *cfdata;
uint32_t cflen, addr;
int rc;
load_fw_module(sc, &dcfg, NULL);
/* Card specific interpretation of "default". */
if (strncmp(cfg_file, DEFAULT_CF, sizeof(t4_cfg_file)) == 0) {
if (pci_get_device(sc->dev) == 0x440a)
snprintf(cfg_file, sizeof(t4_cfg_file), UWIRE_CF);
if (is_fpga(sc))
snprintf(cfg_file, sizeof(t4_cfg_file), FPGA_CF);
}
if (strncmp(cfg_file, DEFAULT_CF, sizeof(t4_cfg_file)) == 0) {
if (dcfg == NULL) {
cxgbe(4): Consider all the API versions of the interfaces exported by the firmware (instead of just the main firmware version) when evaluating firmware compatibility. Document the new "hw.cxgbe.fw_install" knob being introduced here. This should fix kern/173584 too. Setting hw.cxgbe.fw_install=2 will mostly do what was requested in the PR but it's a bit more intelligent in that it won't reinstall the same firmware repeatedly if the knob is left set. PR: kern/173584 MFC after: 5 days
2013-02-26 20:35:54 +00:00
device_printf(sc->dev,
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
"KLD with default config is not available.\n");
rc = ENOENT;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
goto done;
Upgrade the firmware on the card automatically if a better version is available. Downgrade only for a major version mismatch. MFC after: 1 week
2011-03-05 03:12:50 +00:00
}
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
cfdata = dcfg->data;
cflen = dcfg->datasize & ~3;
} else {
char s[32];
Upgrade the firmware on the card automatically if a better version is available. Downgrade only for a major version mismatch. MFC after: 1 week
2011-03-05 03:12:50 +00:00
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
fw_info = find_fw_info(chip_id(sc));
if (fw_info == NULL) {
device_printf(sc->dev,
"unable to look up firmware information for chip %d.\n",
chip_id(sc));
rc = EINVAL;
goto done;
}
snprintf(s, sizeof(s), "%s_%s", fw_info->kld_name, cfg_file);
cxgbe(4): Consider all the API versions of the interfaces exported by the firmware (instead of just the main firmware version) when evaluating firmware compatibility. Document the new "hw.cxgbe.fw_install" knob being introduced here. This should fix kern/173584 too. Setting hw.cxgbe.fw_install=2 will mostly do what was requested in the PR but it's a bit more intelligent in that it won't reinstall the same firmware repeatedly if the knob is left set. PR: kern/173584 MFC after: 5 days
2013-02-26 20:35:54 +00:00
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
rcfg = firmware_get(s);
if (rcfg == NULL) {
device_printf(sc->dev,
"unable to load module \"%s\" for configuration "
"profile \"%s\".\n", s, cfg_file);
rc = ENOENT;
goto done;
}
cfdata = rcfg->data;
cflen = rcfg->datasize & ~3;
}
cxgbe(4): Consider all the API versions of the interfaces exported by the firmware (instead of just the main firmware version) when evaluating firmware compatibility. Document the new "hw.cxgbe.fw_install" knob being introduced here. This should fix kern/173584 too. Setting hw.cxgbe.fw_install=2 will mostly do what was requested in the PR but it's a bit more intelligent in that it won't reinstall the same firmware repeatedly if the knob is left set. PR: kern/173584 MFC after: 5 days
2013-02-26 20:35:54 +00:00
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
if (cflen > FLASH_CFG_MAX_SIZE) {
device_printf(sc->dev,
"config file too long (%d, max allowed is %d).\n",
cflen, FLASH_CFG_MAX_SIZE);
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
rc = EINVAL;
cxgbe(4): Consider all the API versions of the interfaces exported by the firmware (instead of just the main firmware version) when evaluating firmware compatibility. Document the new "hw.cxgbe.fw_install" knob being introduced here. This should fix kern/173584 too. Setting hw.cxgbe.fw_install=2 will mostly do what was requested in the PR but it's a bit more intelligent in that it won't reinstall the same firmware repeatedly if the knob is left set. PR: kern/173584 MFC after: 5 days
2013-02-26 20:35:54 +00:00
goto done;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
rc = validate_mt_off_len(sc, mtype, moff, cflen, &addr);
if (rc != 0) {
device_printf(sc->dev,
"%s: addr (%d/0x%x) or len %d is not valid: %d.\n",
__func__, mtype, moff, cflen, rc);
rc = EINVAL;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
goto done;
}
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
write_via_memwin(sc, 2, addr, cfdata, cflen);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
done:
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
if (rcfg != NULL)
firmware_put(rcfg, FIRMWARE_UNLOAD);
unload_fw_module(sc, dcfg, NULL);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
return (rc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
struct caps_allowed {
uint16_t nbmcaps;
uint16_t linkcaps;
uint16_t switchcaps;
uint16_t niccaps;
uint16_t toecaps;
uint16_t rdmacaps;
uint16_t cryptocaps;
uint16_t iscsicaps;
uint16_t fcoecaps;
};
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
#define FW_PARAM_DEV(param) \
(V_FW_PARAMS_MNEM(FW_PARAMS_MNEM_DEV) | \
V_FW_PARAMS_PARAM_X(FW_PARAMS_PARAM_DEV_##param))
#define FW_PARAM_PFVF(param) \
(V_FW_PARAMS_MNEM(FW_PARAMS_MNEM_PFVF) | \
V_FW_PARAMS_PARAM_X(FW_PARAMS_PARAM_PFVF_##param))
/*
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
* Provide a configuration profile to the firmware and have it initialize the
* chip accordingly. This may involve uploading a configuration file to the
* card.
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
*/
Firmware device log. # sysctl dev.t4nex.0.devlog MFC after: mdf's sysctl+sbuf changes are MFC'd
2011-05-31 23:49:13 +00:00
static int
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
apply_cfg_and_initialize(struct adapter *sc, char *cfg_file,
const struct caps_allowed *caps_allowed)
Firmware device log. # sysctl dev.t4nex.0.devlog MFC after: mdf's sysctl+sbuf changes are MFC'd
2011-05-31 23:49:13 +00:00
{
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
int rc;
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
struct fw_caps_config_cmd caps;
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
uint32_t mtype, moff, finicsum, cfcsum, param, val;
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
rc = -t4_fw_reset(sc, sc->mbox, F_PIORSTMODE | F_PIORST);
if (rc != 0) {
device_printf(sc->dev, "firmware reset failed: %d.\n", rc);
return (rc);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
bzero(&caps, sizeof(caps));
caps.op_to_write = htobe32(V_FW_CMD_OP(FW_CAPS_CONFIG_CMD) |
F_FW_CMD_REQUEST | F_FW_CMD_READ);
if (strncmp(cfg_file, BUILTIN_CF, sizeof(t4_cfg_file)) == 0) {
mtype = 0;
moff = 0;
caps.cfvalid_to_len16 = htobe32(FW_LEN16(caps));
} else if (strncmp(cfg_file, FLASH_CF, sizeof(t4_cfg_file)) == 0) {
mtype = FW_MEMTYPE_FLASH;
moff = t4_flash_cfg_addr(sc);
caps.cfvalid_to_len16 = htobe32(F_FW_CAPS_CONFIG_CMD_CFVALID |
V_FW_CAPS_CONFIG_CMD_MEMTYPE_CF(mtype) |
V_FW_CAPS_CONFIG_CMD_MEMADDR64K_CF(moff >> 16) |
FW_LEN16(caps));
} else {
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
/*
* Ask the firmware where it wants us to upload the config file.
*/
param = FW_PARAM_DEV(CF);
rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 1, &param, &val);
if (rc != 0) {
/* No support for config file? Shouldn't happen. */
device_printf(sc->dev,
"failed to query config file location: %d.\n", rc);
goto done;
}
mtype = G_FW_PARAMS_PARAM_Y(val);
moff = G_FW_PARAMS_PARAM_Z(val) << 16;
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
caps.cfvalid_to_len16 = htobe32(F_FW_CAPS_CONFIG_CMD_CFVALID |
V_FW_CAPS_CONFIG_CMD_MEMTYPE_CF(mtype) |
V_FW_CAPS_CONFIG_CMD_MEMADDR64K_CF(moff >> 16) |
FW_LEN16(caps));
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
rc = copy_cfg_file_to_card(sc, cfg_file, mtype, moff);
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
if (rc != 0) {
device_printf(sc->dev,
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
"failed to upload config file to card: %d.\n", rc);
goto done;
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
rc = -t4_wr_mbox(sc, sc->mbox, &caps, sizeof(caps), &caps);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (rc != 0) {
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
device_printf(sc->dev, "failed to pre-process config file: %d "
"(mtype %d, moff 0x%x).\n", rc, mtype, moff);
goto done;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
finicsum = be32toh(caps.finicsum);
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
cfcsum = be32toh(caps.cfcsum); /* actual */
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
if (finicsum != cfcsum) {
device_printf(sc->dev,
"WARNING: config file checksum mismatch: %08x %08x\n",
finicsum, cfcsum);
}
sc->cfcsum = cfcsum;
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
snprintf(sc->cfg_file, sizeof(sc->cfg_file), "%s", cfg_file);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
/*
* Let the firmware know what features will (not) be used so it can tune
* things accordingly.
*/
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
#define LIMIT_CAPS(x) do { \
caps.x##caps &= htobe16(caps_allowed->x##caps); \
} while (0)
LIMIT_CAPS(nbm);
LIMIT_CAPS(link);
LIMIT_CAPS(switch);
LIMIT_CAPS(nic);
LIMIT_CAPS(toe);
LIMIT_CAPS(rdma);
LIMIT_CAPS(crypto);
LIMIT_CAPS(iscsi);
LIMIT_CAPS(fcoe);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
#undef LIMIT_CAPS
cxgbe(4): Add support for hash filters. These filters reside in the card's memory instead of its TCAM and can be configured via a new "hashfilter" subcommand in cxgbetool. Hash and normal TCAM filters can be used together. The hardware does an exact-match of packet fields for hash filters, unlike the masked match performed for TCAM filters. Any T5/T6 card with memory can support at least half a million hash filters. The sample config file with the driver configures 512K of these, it is possible to double this to 1 million+ in some cases. The chip does an exact-match of fields of incoming datagrams with hash filters and performs the action configured for the filter if it matches. The fields to match are specified in a "filter mask" in the firmware config file. The filter mask always includes the 5-tuple (sip, dip, sport, dport, ipproto). It can, optionally, also include any subset of the filter mode (see filterMode and filterMask in the firmware config file). For example: filterMode = fragmentation, mpshittype, protocol, vlan, port, fcoe filterMask = protocol, port, vlan Exact values of the 5-tuple, the physical port, and VLAN tag would have to be provided while setting up a hash filter with the chip configuration above. Hash filters support all actions supported by TCAM filters. A packet that hits a hash filter can be dropped, let through (with optional steering to a specific queue or RSS region), switched out of another port (with optional L2 rewrite of DMAC, SMAC, VLAN tag), or get NAT'ed. (Support for some of these will show up in the driver in a follow-up commit very shortly). Sponsored by: Chelsio Communications
2018-05-09 04:09:49 +00:00
if (caps.niccaps & htobe16(FW_CAPS_CONFIG_NIC_HASHFILTER)) {
/*
* TOE and hashfilters are mutually exclusive. It is a config
* file or firmware bug if both are reported as available. Try
* to cope with the situation in non-debug builds by disabling
* TOE.
*/
MPASS(caps.toecaps == 0);
caps.toecaps = 0;
caps.rdmacaps = 0;
caps.iscsicaps = 0;
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
caps.op_to_write = htobe32(V_FW_CMD_OP(FW_CAPS_CONFIG_CMD) |
F_FW_CMD_REQUEST | F_FW_CMD_WRITE);
caps.cfvalid_to_len16 = htobe32(FW_LEN16(caps));
rc = -t4_wr_mbox(sc, sc->mbox, &caps, sizeof(caps), NULL);
if (rc != 0) {
device_printf(sc->dev,
"failed to process config file: %d.\n", rc);
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
goto done;
}
t4_tweak_chip_settings(sc);
cxgbe(4): Request high priority filter support explicitly, as required by recent firmwares. MFC after: 1 week Sponsored by: Chelsio Communications
2019-02-28 05:45:14 +00:00
set_params__pre_init(sc);
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
/* get basic stuff going */
rc = -t4_fw_initialize(sc, sc->mbox);
if (rc != 0) {
device_printf(sc->dev, "fw_initialize failed: %d.\n", rc);
goto done;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
done:
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
return (rc);
}
/*
* Partition chip resources for use between various PFs, VFs, etc.
*/
static int
partition_resources(struct adapter *sc)
{
char cfg_file[sizeof(t4_cfg_file)];
struct caps_allowed caps_allowed;
int rc;
bool fallback;
/* Only the master driver gets to configure the chip resources. */
MPASS(sc->flags & MASTER_PF);
#define COPY_CAPS(x) do { \
caps_allowed.x##caps = t4_##x##caps_allowed; \
} while (0)
bzero(&caps_allowed, sizeof(caps_allowed));
COPY_CAPS(nbm);
COPY_CAPS(link);
COPY_CAPS(switch);
COPY_CAPS(nic);
COPY_CAPS(toe);
COPY_CAPS(rdma);
COPY_CAPS(crypto);
COPY_CAPS(iscsi);
COPY_CAPS(fcoe);
fallback = sc->debug_flags & DF_DISABLE_CFG_RETRY ? false : true;
snprintf(cfg_file, sizeof(cfg_file), "%s", t4_cfg_file);
retry:
rc = apply_cfg_and_initialize(sc, cfg_file, &caps_allowed);
if (rc != 0 && fallback) {
device_printf(sc->dev,
"failed (%d) to configure card with \"%s\" profile, "
"will fall back to a basic configuration and retry.\n",
rc, cfg_file);
snprintf(cfg_file, sizeof(cfg_file), "%s", BUILTIN_CF);
bzero(&caps_allowed, sizeof(caps_allowed));
COPY_CAPS(switch);
cxgbe(4): Use the simplest configuration possible when falling back from the default configuration. MFC after: 1 week Sponsored by: Chelsio Communications
2019-07-09 19:32:31 +00:00
caps_allowed.niccaps = FW_CAPS_CONFIG_NIC;
cxgbe(4): Fall back to a basic configuration in case of any error during card initialization. This is an expanded version of r333682. Break up prep_firmware into simpler routines while here. Load the firmware/config KLD only if needed. MFC after: 1 month Sponsored by: Chelsio Communications
2018-12-06 06:18:21 +00:00
fallback = false;
goto retry;
}
#undef COPY_CAPS
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
return (rc);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
/*
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
* Retrieve parameters that are needed (or nice to have) very early.
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
*/
static int
get_params__pre_init(struct adapter *sc)
{
int rc;
uint32_t param[2], val[2];
cxgbe(4): Provide more details about the card in the sysctl MIB. dev.t5nex.0.%desc: Chelsio T580-CR dev.t5nex.0.hw_revision: 1 dev.t5nex.0.sn: PT13140042 dev.t5nex.0.pn: 110117150A0 dev.t5nex.0.ec: 0000000000000000 dev.t5nex.0.na: 0007432AF490 dev.t5nex.0.vpd_version: 3 dev.t5nex.0.scfg_version: 53255 dev.t5nex.0.bs_version: 1.1.0.0 dev.t5nex.0.er_version: 1.0.0.68 dev.t5nex.0.tp_version: 0.1.4.9 dev.t5nex.0.firmware_version: 1.16.2.0 Sponsored by: Chelsio Communications
2016-08-27 00:13:41 +00:00
t4_get_version_info(sc);
snprintf(sc->fw_version, sizeof(sc->fw_version), "%u.%u.%u.%u",
G_FW_HDR_FW_VER_MAJOR(sc->params.fw_vers),
G_FW_HDR_FW_VER_MINOR(sc->params.fw_vers),
G_FW_HDR_FW_VER_MICRO(sc->params.fw_vers),
G_FW_HDR_FW_VER_BUILD(sc->params.fw_vers));
snprintf(sc->bs_version, sizeof(sc->bs_version), "%u.%u.%u.%u",
G_FW_HDR_FW_VER_MAJOR(sc->params.bs_vers),
G_FW_HDR_FW_VER_MINOR(sc->params.bs_vers),
G_FW_HDR_FW_VER_MICRO(sc->params.bs_vers),
G_FW_HDR_FW_VER_BUILD(sc->params.bs_vers));
snprintf(sc->tp_version, sizeof(sc->tp_version), "%u.%u.%u.%u",
G_FW_HDR_FW_VER_MAJOR(sc->params.tp_vers),
G_FW_HDR_FW_VER_MINOR(sc->params.tp_vers),
G_FW_HDR_FW_VER_MICRO(sc->params.tp_vers),
G_FW_HDR_FW_VER_BUILD(sc->params.tp_vers));
snprintf(sc->er_version, sizeof(sc->er_version), "%u.%u.%u.%u",
G_FW_HDR_FW_VER_MAJOR(sc->params.er_vers),
G_FW_HDR_FW_VER_MINOR(sc->params.er_vers),
G_FW_HDR_FW_VER_MICRO(sc->params.er_vers),
G_FW_HDR_FW_VER_BUILD(sc->params.er_vers));
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
param[0] = FW_PARAM_DEV(PORTVEC);
param[1] = FW_PARAM_DEV(CCLK);
rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 2, param, val);
if (rc != 0) {
device_printf(sc->dev,
"failed to query parameters (pre_init): %d.\n", rc);
return (rc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
sc->params.portvec = val[0];
Use native FreeBSD facilities everywhere except the shared code in common/ MFC after: 1 week
2012-09-13 09:10:10 +00:00
sc->params.nports = bitcount32(val[0]);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
sc->params.vpd.cclk = val[1];
/* Read device log parameters. */
cxgbe(4): Improvements to the code that deals with the firmware's log. - Query the location of the log very early during attach. Refresh the location later after establishing contact with the firmware. - Save the log's location as a flat address in devlog_params. - Use a memory window instead of backdoor access to the EDC/MC to read the log.
2016-03-10 23:17:26 +00:00
rc = -t4_init_devlog_params(sc, 1);
if (rc == 0)
fixup_devlog_params(sc);
else {
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
device_printf(sc->dev,
"failed to get devlog parameters: %d.\n", rc);
rc = 0; /* devlog isn't critical for device operation */
}
return (rc);
}
cxgbe(4): Request high priority filter support explicitly, as required by recent firmwares. MFC after: 1 week Sponsored by: Chelsio Communications
2019-02-28 05:45:14 +00:00
/*
* Any params that need to be set before FW_INITIALIZE.
*/
static int
set_params__pre_init(struct adapter *sc)
{
int rc = 0;
uint32_t param, val;
if (chip_id(sc) >= CHELSIO_T6) {
param = FW_PARAM_DEV(HPFILTER_REGION_SUPPORT);
val = 1;
rc = -t4_set_params(sc, sc->mbox, sc->pf, 0, 1, &param, &val);
/* firmwares < 1.20.1.0 do not have this param. */
if (rc == FW_EINVAL && sc->params.fw_vers <
(V_FW_HDR_FW_VER_MAJOR(1) | V_FW_HDR_FW_VER_MINOR(20) |
V_FW_HDR_FW_VER_MICRO(1) | V_FW_HDR_FW_VER_BUILD(0))) {
rc = 0;
}
if (rc != 0) {
device_printf(sc->dev,
"failed to enable high priority filters :%d.\n",
rc);
}
}
cxgbe(4): Treat the viid as an opaque identifier. Recent firmwares prefer to use a different format for viid internally and this change allows them to do so. MFC after: 1 week Sponsored by: Chelsio Communications
2019-03-20 17:27:11 +00:00
/* Enable opaque VIIDs with firmwares that support it. */
param = FW_PARAM_DEV(OPAQUE_VIID_SMT_EXTN);
val = 1;
rc = -t4_set_params(sc, sc->mbox, sc->pf, 0, 1, &param, &val);
if (rc == 0 && val == 1)
sc->params.viid_smt_extn_support = true;
else
sc->params.viid_smt_extn_support = false;
cxgbe(4): Request high priority filter support explicitly, as required by recent firmwares. MFC after: 1 week Sponsored by: Chelsio Communications
2019-02-28 05:45:14 +00:00
return (rc);
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
/*
* Retrieve various parameters that are of interest to the driver. The device
* has been initialized by the firmware at this point.
*/
static int
get_params__post_init(struct adapter *sc)
{
int rc;
uint32_t param[7], val[7];
struct fw_caps_config_cmd caps;
param[0] = FW_PARAM_PFVF(IQFLINT_START);
param[1] = FW_PARAM_PFVF(EQ_START);
param[2] = FW_PARAM_PFVF(FILTER_START);
param[3] = FW_PARAM_PFVF(FILTER_END);
cxgbe(4): Updates to the hardware L2 table management code. - Add full support for IPv6 addresses. - Read the size of the L2 table during attach. Do not assume that PCIe physical function 4 of the card has all of the table to itself. - Use FNV instead of Jenkins to hash L3 addresses and drop the private copy of jhash.h from the driver. MFC after: 1 week
2013-01-14 20:36:22 +00:00
param[4] = FW_PARAM_PFVF(L2T_START);
param[5] = FW_PARAM_PFVF(L2T_END);
cxgbe(4): Add core Vdd to the sysctl MIB. Sponsored by: Chelsio Communications
2017-11-17 23:22:39 +00:00
param[6] = V_FW_PARAMS_MNEM(FW_PARAMS_MNEM_DEV) |
V_FW_PARAMS_PARAM_X(FW_PARAMS_PARAM_DEV_DIAG) |
V_FW_PARAMS_PARAM_Y(FW_PARAM_DEV_DIAG_VDD);
rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 7, param, val);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
if (rc != 0) {
device_printf(sc->dev,
"failed to query parameters (post_init): %d.\n", rc);
return (rc);
}
sc->sge.iq_start = val[0];
sc->sge.eq_start = val[1];
cxgbe(4): Treat base/end of firmware parameters as signed integers when figuring out whether the range is valid or not. Approved by: re@ (rgrimes@) MFC after: 1 week Sponsored by: Chelsio Communications
2018-09-26 02:27:37 +00:00
if ((int)val[3] > (int)val[2]) {
cxgbe(4): Improvements in TID management. - Ignore any type of TID where the start/end values are not in the correct order. There are situations where the firmware isn't able to reserve room for the number requested in the config file but doesn't report a failure during configuration and instead sets end <= start. - Track start/end in tid_tab and remove some redundant copies from adapter->params. - Move all the start/end and other read-only parameters to a quiet part of tid_tab, away from the tid locks. MFC after: 1 week Sponsored by: Chelsio Communications
2018-08-02 22:52:05 +00:00
sc->tids.ftid_base = val[2];
sc->tids.ftid_end = val[3];
sc->tids.nftids = val[3] - val[2] + 1;
}
cxgbe(4): Updates to the hardware L2 table management code. - Add full support for IPv6 addresses. - Read the size of the L2 table during attach. Do not assume that PCIe physical function 4 of the card has all of the table to itself. - Use FNV instead of Jenkins to hash L3 addresses and drop the private copy of jhash.h from the driver. MFC after: 1 week
2013-01-14 20:36:22 +00:00
sc->vres.l2t.start = val[4];
sc->vres.l2t.size = val[5] - val[4] + 1;
KASSERT(sc->vres.l2t.size <= L2T_SIZE,
("%s: L2 table size (%u) larger than expected (%u)",
__func__, sc->vres.l2t.size, L2T_SIZE));
cxgbe(4): Add core Vdd to the sysctl MIB. Sponsored by: Chelsio Communications
2017-11-17 23:22:39 +00:00
sc->params.core_vdd = val[6];
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): Add support for high priority filters on T6+. They have their own region in the TCAM starting with T6, unlike previous chips where they were in the same region as normal filters. These filters "hit" before anything else in the LE's lookup. The exact order is: a) High priority filters b) TOE's active region (TCAM and/or hash) c) Servers (TOE hw listeners) d) Normal filters MFC after: 1 week Sponsored by: Chelsio Communications
2018-08-09 14:19:47 +00:00
if (chip_id(sc) >= CHELSIO_T6) {
sc->tids.tid_base = t4_read_reg(sc,
A_LE_DB_ACTIVE_TABLE_START_INDEX);
param[0] = FW_PARAM_PFVF(HPFILTER_START);
param[1] = FW_PARAM_PFVF(HPFILTER_END);
rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 2, param, val);
if (rc != 0) {
device_printf(sc->dev,
"failed to query hpfilter parameters: %d.\n", rc);
return (rc);
}
if ((int)val[1] > (int)val[0]) {
sc->tids.hpftid_base = val[0];
sc->tids.hpftid_end = val[1];
sc->tids.nhpftids = val[1] - val[0] + 1;
/*
* These should go off if the layout changes and the
* driver needs to catch up.
*/
MPASS(sc->tids.hpftid_base == 0);
MPASS(sc->tids.tid_base == sc->tids.nhpftids);
}
}
cxgbe(4): Read the MPS buffer group map from the firmware as it could be different from hardware defaults. The congestion channel map, which is still fixed, needs to be tracked separately now. Change the congestion setting for TOE rx queues to match the drivers on other OSes while here. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-10-24 05:41:48 +00:00
/*
* MPSBGMAP is queried separately because only recent firmwares support
* it as a parameter and we don't want the compound query above to fail
* on older firmwares.
*/
param[0] = FW_PARAM_DEV(MPSBGMAP);
val[0] = 0;
rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 1, param, val);
if (rc == 0)
sc->params.mps_bg_map = val[0];
else
sc->params.mps_bg_map = 0;
cxgbe(4): Determine whether the firmware supports the FILTER2 work request, which can be used to configure hardware NAT and swapmac. All firmwares released after Jan 2017 support this work request. Sponsored by: Chelsio Communications
2018-05-10 00:04:14 +00:00
/*
* Determine whether the firmware supports the filter2 work request.
* This is queried separately for the same reason as MPSBGMAP above.
*/
param[0] = FW_PARAM_DEV(FILTER2_WR);
val[0] = 0;
rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 1, param, val);
if (rc == 0)
sc->params.filter2_wr_support = val[0] != 0;
else
sc->params.filter2_wr_support = 0;
cxgbe/iw_cxgbe: Use DSGLs to write to card's memory when appropriate. Submitted by: Krishnamraju Eraparaju @ Chelsio Sponsored by: Chelsio Communications
2018-12-19 23:29:01 +00:00
/*
* Find out whether we're allowed to use the ULPTX MEMWRITE DSGL.
* This is queried separately for the same reason as other params above.
*/
param[0] = FW_PARAM_DEV(ULPTX_MEMWRITE_DSGL);
val[0] = 0;
rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 1, param, val);
if (rc == 0)
sc->params.ulptx_memwrite_dsgl = val[0] != 0;
else
sc->params.ulptx_memwrite_dsgl = false;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
/* get capabilites */
bzero(&caps, sizeof(caps));
caps.op_to_write = htobe32(V_FW_CMD_OP(FW_CAPS_CONFIG_CMD) |
F_FW_CMD_REQUEST | F_FW_CMD_READ);
caps.cfvalid_to_len16 = htobe32(FW_LEN16(caps));
rc = -t4_wr_mbox(sc, sc->mbox, &caps, sizeof(caps), &caps);
if (rc != 0) {
device_printf(sc->dev,
"failed to get card capabilities: %d.\n", rc);
return (rc);
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Read card capabilities after firmware initialization, instead of setting them up as part of firmware initialization (which the driver gets to do only if it's the master driver). Read the range of tids available for the ETHOFLD functionality if it's enabled. New is_ftid() and is_etid() functions to test whether a tid falls within the range of filter tids or ETHOFLD tids respectively. MFC after: 2 weeks
2013-12-14 03:08:03 +00:00
#define READ_CAPS(x) do { \
sc->x = htobe16(caps.x); \
} while (0)
cxgbe(4): Catch up with the latest list of card capabilities as reported by the firmware.
2016-03-12 02:54:55 +00:00
READ_CAPS(nbmcaps);
Read card capabilities after firmware initialization, instead of setting them up as part of firmware initialization (which the driver gets to do only if it's the master driver). Read the range of tids available for the ETHOFLD functionality if it's enabled. New is_ftid() and is_etid() functions to test whether a tid falls within the range of filter tids or ETHOFLD tids respectively. MFC after: 2 weeks
2013-12-14 03:08:03 +00:00
READ_CAPS(linkcaps);
cxgbe(4): Catch up with the latest list of card capabilities as reported by the firmware.
2016-03-12 02:54:55 +00:00
READ_CAPS(switchcaps);
Read card capabilities after firmware initialization, instead of setting them up as part of firmware initialization (which the driver gets to do only if it's the master driver). Read the range of tids available for the ETHOFLD functionality if it's enabled. New is_ftid() and is_etid() functions to test whether a tid falls within the range of filter tids or ETHOFLD tids respectively. MFC after: 2 weeks
2013-12-14 03:08:03 +00:00
READ_CAPS(niccaps);
READ_CAPS(toecaps);
READ_CAPS(rdmacaps);
cxgbe(4): Catch up with the rename of tlscaps -> cryptocaps. TLS is one of the capabilities of the crypto engine in T6. Sponsored by: Chelsio Communications
2016-09-12 00:15:40 +00:00
READ_CAPS(cryptocaps);
Read card capabilities after firmware initialization, instead of setting them up as part of firmware initialization (which the driver gets to do only if it's the master driver). Read the range of tids available for the ETHOFLD functionality if it's enabled. New is_ftid() and is_etid() functions to test whether a tid falls within the range of filter tids or ETHOFLD tids respectively. MFC after: 2 weeks
2013-12-14 03:08:03 +00:00
READ_CAPS(iscsicaps);
READ_CAPS(fcoecaps);
cxgbe(4): Add support for hash filters. These filters reside in the card's memory instead of its TCAM and can be configured via a new "hashfilter" subcommand in cxgbetool. Hash and normal TCAM filters can be used together. The hardware does an exact-match of packet fields for hash filters, unlike the masked match performed for TCAM filters. Any T5/T6 card with memory can support at least half a million hash filters. The sample config file with the driver configures 512K of these, it is possible to double this to 1 million+ in some cases. The chip does an exact-match of fields of incoming datagrams with hash filters and performs the action configured for the filter if it matches. The fields to match are specified in a "filter mask" in the firmware config file. The filter mask always includes the 5-tuple (sip, dip, sport, dport, ipproto). It can, optionally, also include any subset of the filter mode (see filterMode and filterMask in the firmware config file). For example: filterMode = fragmentation, mpshittype, protocol, vlan, port, fcoe filterMask = protocol, port, vlan Exact values of the 5-tuple, the physical port, and VLAN tag would have to be provided while setting up a hash filter with the chip configuration above. Hash filters support all actions supported by TCAM filters. A packet that hits a hash filter can be dropped, let through (with optional steering to a specific queue or RSS region), switched out of another port (with optional L2 rewrite of DMAC, SMAC, VLAN tag), or get NAT'ed. (Support for some of these will show up in the driver in a follow-up commit very shortly). Sponsored by: Chelsio Communications
2018-05-09 04:09:49 +00:00
if (sc->niccaps & FW_CAPS_CONFIG_NIC_HASHFILTER) {
MPASS(chip_id(sc) > CHELSIO_T4);
MPASS(sc->toecaps == 0);
sc->toecaps = 0;
cxgbe(4): Ignore capabilities that depend on TOE when the firmware reports TOE is not available. MFC after: 1 week Sponsored by: Chelsio Communications
2017-09-13 06:07:02 +00:00
cxgbe(4): Add support for hash filters. These filters reside in the card's memory instead of its TCAM and can be configured via a new "hashfilter" subcommand in cxgbetool. Hash and normal TCAM filters can be used together. The hardware does an exact-match of packet fields for hash filters, unlike the masked match performed for TCAM filters. Any T5/T6 card with memory can support at least half a million hash filters. The sample config file with the driver configures 512K of these, it is possible to double this to 1 million+ in some cases. The chip does an exact-match of fields of incoming datagrams with hash filters and performs the action configured for the filter if it matches. The fields to match are specified in a "filter mask" in the firmware config file. The filter mask always includes the 5-tuple (sip, dip, sport, dport, ipproto). It can, optionally, also include any subset of the filter mode (see filterMode and filterMask in the firmware config file). For example: filterMode = fragmentation, mpshittype, protocol, vlan, port, fcoe filterMask = protocol, port, vlan Exact values of the 5-tuple, the physical port, and VLAN tag would have to be provided while setting up a hash filter with the chip configuration above. Hash filters support all actions supported by TCAM filters. A packet that hits a hash filter can be dropped, let through (with optional steering to a specific queue or RSS region), switched out of another port (with optional L2 rewrite of DMAC, SMAC, VLAN tag), or get NAT'ed. (Support for some of these will show up in the driver in a follow-up commit very shortly). Sponsored by: Chelsio Communications
2018-05-09 04:09:49 +00:00
param[0] = FW_PARAM_DEV(NTID);
cxgbe(4): Use the correct number of parameters when querying the tid range for hashfilters. Approved by: re@ (gjb@)
2018-09-13 22:58:13 +00:00
rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 1, param, val);
cxgbe(4): Add support for hash filters. These filters reside in the card's memory instead of its TCAM and can be configured via a new "hashfilter" subcommand in cxgbetool. Hash and normal TCAM filters can be used together. The hardware does an exact-match of packet fields for hash filters, unlike the masked match performed for TCAM filters. Any T5/T6 card with memory can support at least half a million hash filters. The sample config file with the driver configures 512K of these, it is possible to double this to 1 million+ in some cases. The chip does an exact-match of fields of incoming datagrams with hash filters and performs the action configured for the filter if it matches. The fields to match are specified in a "filter mask" in the firmware config file. The filter mask always includes the 5-tuple (sip, dip, sport, dport, ipproto). It can, optionally, also include any subset of the filter mode (see filterMode and filterMask in the firmware config file). For example: filterMode = fragmentation, mpshittype, protocol, vlan, port, fcoe filterMask = protocol, port, vlan Exact values of the 5-tuple, the physical port, and VLAN tag would have to be provided while setting up a hash filter with the chip configuration above. Hash filters support all actions supported by TCAM filters. A packet that hits a hash filter can be dropped, let through (with optional steering to a specific queue or RSS region), switched out of another port (with optional L2 rewrite of DMAC, SMAC, VLAN tag), or get NAT'ed. (Support for some of these will show up in the driver in a follow-up commit very shortly). Sponsored by: Chelsio Communications
2018-05-09 04:09:49 +00:00
if (rc != 0) {
device_printf(sc->dev,
"failed to query HASHFILTER parameters: %d.\n", rc);
return (rc);
}
sc->tids.ntids = val[0];
cxgbe(4): Add support for high priority filters on T6+. They have their own region in the TCAM starting with T6, unlike previous chips where they were in the same region as normal filters. These filters "hit" before anything else in the LE's lookup. The exact order is: a) High priority filters b) TOE's active region (TCAM and/or hash) c) Servers (TOE hw listeners) d) Normal filters MFC after: 1 week Sponsored by: Chelsio Communications
2018-08-09 14:19:47 +00:00
if (sc->params.fw_vers <
(V_FW_HDR_FW_VER_MAJOR(1) | V_FW_HDR_FW_VER_MINOR(20) |
V_FW_HDR_FW_VER_MICRO(5) | V_FW_HDR_FW_VER_BUILD(0))) {
MPASS(sc->tids.ntids >= sc->tids.nhpftids);
sc->tids.ntids -= sc->tids.nhpftids;
}
cxgbe(4): Add support for hash filters. These filters reside in the card's memory instead of its TCAM and can be configured via a new "hashfilter" subcommand in cxgbetool. Hash and normal TCAM filters can be used together. The hardware does an exact-match of packet fields for hash filters, unlike the masked match performed for TCAM filters. Any T5/T6 card with memory can support at least half a million hash filters. The sample config file with the driver configures 512K of these, it is possible to double this to 1 million+ in some cases. The chip does an exact-match of fields of incoming datagrams with hash filters and performs the action configured for the filter if it matches. The fields to match are specified in a "filter mask" in the firmware config file. The filter mask always includes the 5-tuple (sip, dip, sport, dport, ipproto). It can, optionally, also include any subset of the filter mode (see filterMode and filterMask in the firmware config file). For example: filterMode = fragmentation, mpshittype, protocol, vlan, port, fcoe filterMask = protocol, port, vlan Exact values of the 5-tuple, the physical port, and VLAN tag would have to be provided while setting up a hash filter with the chip configuration above. Hash filters support all actions supported by TCAM filters. A packet that hits a hash filter can be dropped, let through (with optional steering to a specific queue or RSS region), switched out of another port (with optional L2 rewrite of DMAC, SMAC, VLAN tag), or get NAT'ed. (Support for some of these will show up in the driver in a follow-up commit very shortly). Sponsored by: Chelsio Communications
2018-05-09 04:09:49 +00:00
sc->tids.natids = min(sc->tids.ntids / 2, MAX_ATIDS);
sc->params.hash_filter = 1;
}
Read card capabilities after firmware initialization, instead of setting them up as part of firmware initialization (which the driver gets to do only if it's the master driver). Read the range of tids available for the ETHOFLD functionality if it's enabled. New is_ftid() and is_etid() functions to test whether a tid falls within the range of filter tids or ETHOFLD tids respectively. MFC after: 2 weeks
2013-12-14 03:08:03 +00:00
if (sc->niccaps & FW_CAPS_CONFIG_NIC_ETHOFLD) {
param[0] = FW_PARAM_PFVF(ETHOFLD_START);
param[1] = FW_PARAM_PFVF(ETHOFLD_END);
param[2] = FW_PARAM_DEV(FLOWC_BUFFIFO_SZ);
rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 3, param, val);
if (rc != 0) {
device_printf(sc->dev,
"failed to query NIC parameters: %d.\n", rc);
return (rc);
}
cxgbe(4): Treat base/end of firmware parameters as signed integers when figuring out whether the range is valid or not. Approved by: re@ (rgrimes@) MFC after: 1 week Sponsored by: Chelsio Communications
2018-09-26 02:27:37 +00:00
if ((int)val[1] > (int)val[0]) {
cxgbe(4): Improvements in TID management. - Ignore any type of TID where the start/end values are not in the correct order. There are situations where the firmware isn't able to reserve room for the number requested in the config file but doesn't report a failure during configuration and instead sets end <= start. - Track start/end in tid_tab and remove some redundant copies from adapter->params. - Move all the start/end and other read-only parameters to a quiet part of tid_tab, away from the tid locks. MFC after: 1 week Sponsored by: Chelsio Communications
2018-08-02 22:52:05 +00:00
sc->tids.etid_base = val[0];
sc->tids.etid_end = val[1];
sc->tids.netids = val[1] - val[0] + 1;
sc->params.eo_wr_cred = val[2];
sc->params.ethoffload = 1;
}
Read card capabilities after firmware initialization, instead of setting them up as part of firmware initialization (which the driver gets to do only if it's the master driver). Read the range of tids available for the ETHOFLD functionality if it's enabled. New is_ftid() and is_etid() functions to test whether a tid falls within the range of filter tids or ETHOFLD tids respectively. MFC after: 2 weeks
2013-12-14 03:08:03 +00:00
}
if (sc->toecaps) {
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
/* query offload-related parameters */
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
param[0] = FW_PARAM_DEV(NTID);
param[1] = FW_PARAM_PFVF(SERVER_START);
param[2] = FW_PARAM_PFVF(SERVER_END);
param[3] = FW_PARAM_PFVF(TDDP_START);
param[4] = FW_PARAM_PFVF(TDDP_END);
param[5] = FW_PARAM_DEV(FLOWC_BUFFIFO_SZ);
rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 6, param, val);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (rc != 0) {
device_printf(sc->dev,
"failed to query TOE parameters: %d.\n", rc);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
return (rc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
sc->tids.ntids = val[0];
cxgbe(4): Adjust ntids to account for nhptids in the TOE case too. This should have been part of r337538.
2018-08-17 20:28:31 +00:00
if (sc->params.fw_vers <
(V_FW_HDR_FW_VER_MAJOR(1) | V_FW_HDR_FW_VER_MINOR(20) |
V_FW_HDR_FW_VER_MICRO(5) | V_FW_HDR_FW_VER_BUILD(0))) {
MPASS(sc->tids.ntids >= sc->tids.nhpftids);
sc->tids.ntids -= sc->tids.nhpftids;
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
sc->tids.natids = min(sc->tids.ntids / 2, MAX_ATIDS);
cxgbe(4): Treat base/end of firmware parameters as signed integers when figuring out whether the range is valid or not. Approved by: re@ (rgrimes@) MFC after: 1 week Sponsored by: Chelsio Communications
2018-09-26 02:27:37 +00:00
if ((int)val[2] > (int)val[1]) {
cxgbe(4): Improvements in TID management. - Ignore any type of TID where the start/end values are not in the correct order. There are situations where the firmware isn't able to reserve room for the number requested in the config file but doesn't report a failure during configuration and instead sets end <= start. - Track start/end in tid_tab and remove some redundant copies from adapter->params. - Move all the start/end and other read-only parameters to a quiet part of tid_tab, away from the tid locks. MFC after: 1 week Sponsored by: Chelsio Communications
2018-08-02 22:52:05 +00:00
sc->tids.stid_base = val[1];
sc->tids.nstids = val[2] - val[1] + 1;
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
sc->vres.ddp.start = val[3];
sc->vres.ddp.size = val[4] - val[3] + 1;
sc->params.ofldq_wr_cred = val[5];
sc->params.offload = 1;
cxgbe(4): Add support for hash filters. These filters reside in the card's memory instead of its TCAM and can be configured via a new "hashfilter" subcommand in cxgbetool. Hash and normal TCAM filters can be used together. The hardware does an exact-match of packet fields for hash filters, unlike the masked match performed for TCAM filters. Any T5/T6 card with memory can support at least half a million hash filters. The sample config file with the driver configures 512K of these, it is possible to double this to 1 million+ in some cases. The chip does an exact-match of fields of incoming datagrams with hash filters and performs the action configured for the filter if it matches. The fields to match are specified in a "filter mask" in the firmware config file. The filter mask always includes the 5-tuple (sip, dip, sport, dport, ipproto). It can, optionally, also include any subset of the filter mode (see filterMode and filterMask in the firmware config file). For example: filterMode = fragmentation, mpshittype, protocol, vlan, port, fcoe filterMask = protocol, port, vlan Exact values of the 5-tuple, the physical port, and VLAN tag would have to be provided while setting up a hash filter with the chip configuration above. Hash filters support all actions supported by TCAM filters. A packet that hits a hash filter can be dropped, let through (with optional steering to a specific queue or RSS region), switched out of another port (with optional L2 rewrite of DMAC, SMAC, VLAN tag), or get NAT'ed. (Support for some of these will show up in the driver in a follow-up commit very shortly). Sponsored by: Chelsio Communications
2018-05-09 04:09:49 +00:00
} else {
/*
* The firmware attempts memfree TOE configuration for -SO cards
* and will report toecaps=0 if it runs out of resources (this
* depends on the config file). It may not report 0 for other
* capabilities dependent on the TOE in this case. Set them to
* 0 here so that the driver doesn't bother tracking resources
* that will never be used.
*/
sc->iscsicaps = 0;
sc->rdmacaps = 0;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
Read card capabilities after firmware initialization, instead of setting them up as part of firmware initialization (which the driver gets to do only if it's the master driver). Read the range of tids available for the ETHOFLD functionality if it's enabled. New is_ftid() and is_etid() functions to test whether a tid falls within the range of filter tids or ETHOFLD tids respectively. MFC after: 2 weeks
2013-12-14 03:08:03 +00:00
if (sc->rdmacaps) {
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
param[0] = FW_PARAM_PFVF(STAG_START);
param[1] = FW_PARAM_PFVF(STAG_END);
param[2] = FW_PARAM_PFVF(RQ_START);
param[3] = FW_PARAM_PFVF(RQ_END);
param[4] = FW_PARAM_PFVF(PBL_START);
param[5] = FW_PARAM_PFVF(PBL_END);
rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 6, param, val);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (rc != 0) {
device_printf(sc->dev,
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
"failed to query RDMA parameters(1): %d.\n", rc);
return (rc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
sc->vres.stag.start = val[0];
sc->vres.stag.size = val[1] - val[0] + 1;
sc->vres.rq.start = val[2];
sc->vres.rq.size = val[3] - val[2] + 1;
sc->vres.pbl.start = val[4];
sc->vres.pbl.size = val[5] - val[4] + 1;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
param[0] = FW_PARAM_PFVF(SQRQ_START);
param[1] = FW_PARAM_PFVF(SQRQ_END);
param[2] = FW_PARAM_PFVF(CQ_START);
param[3] = FW_PARAM_PFVF(CQ_END);
param[4] = FW_PARAM_PFVF(OCQ_START);
param[5] = FW_PARAM_PFVF(OCQ_END);
Use correct mailbox and PCIe PF number when querying RDMA parameters.
2013-08-26 19:02:52 +00:00
rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 6, param, val);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
if (rc != 0) {
device_printf(sc->dev,
"failed to query RDMA parameters(2): %d.\n", rc);
return (rc);
}
sc->vres.qp.start = val[0];
sc->vres.qp.size = val[1] - val[0] + 1;
sc->vres.cq.start = val[2];
sc->vres.cq.size = val[3] - val[2] + 1;
sc->vres.ocq.start = val[4];
sc->vres.ocq.size = val[5] - val[4] + 1;
cxgbe: Query some more RDMA related parameters from the firmware. MFC after: 3 days Sponsored by: Chelsio Communications
2017-04-13 17:16:36 +00:00
param[0] = FW_PARAM_PFVF(SRQ_START);
param[1] = FW_PARAM_PFVF(SRQ_END);
param[2] = FW_PARAM_DEV(MAXORDIRD_QP);
param[3] = FW_PARAM_DEV(MAXIRD_ADAPTER);
rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 4, param, val);
if (rc != 0) {
device_printf(sc->dev,
"failed to query RDMA parameters(3): %d.\n", rc);
return (rc);
}
sc->vres.srq.start = val[0];
sc->vres.srq.size = val[1] - val[0] + 1;
sc->params.max_ordird_qp = val[2];
sc->params.max_ird_adapter = val[3];
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
Read card capabilities after firmware initialization, instead of setting them up as part of firmware initialization (which the driver gets to do only if it's the master driver). Read the range of tids available for the ETHOFLD functionality if it's enabled. New is_ftid() and is_etid() functions to test whether a tid falls within the range of filter tids or ETHOFLD tids respectively. MFC after: 2 weeks
2013-12-14 03:08:03 +00:00
if (sc->iscsicaps) {
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
param[0] = FW_PARAM_PFVF(ISCSI_START);
param[1] = FW_PARAM_PFVF(ISCSI_END);
rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 2, param, val);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (rc != 0) {
device_printf(sc->dev,
"failed to query iSCSI parameters: %d.\n", rc);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
return (rc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
sc->vres.iscsi.start = val[0];
sc->vres.iscsi.size = val[1] - val[0] + 1;
}
Fetch TLS key parameters from the firmware. The parameters describe how much of the adapter's memory is reserved for storing TLS keys. The 'meminfo' sysctl now lists this region of adapter memory as 'TLS keys' if present. Sponsored by: Chelsio Communications
2018-02-26 21:56:06 +00:00
if (sc->cryptocaps & FW_CAPS_CONFIG_TLSKEYS) {
param[0] = FW_PARAM_PFVF(TLS_START);
param[1] = FW_PARAM_PFVF(TLS_END);
rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 2, param, val);
if (rc != 0) {
device_printf(sc->dev,
"failed to query TLS parameters: %d.\n", rc);
return (rc);
}
sc->vres.key.start = val[0];
sc->vres.key.size = val[1] - val[0] + 1;
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Make SGE parameter handling more VF-friendly. Add fields to hold the SGE control register and free list buffer sizes to the sge_params structure. Populate these new fields in t4_init_sge_params() for PF devices and change t4_read_chip_settings() to pull these values out of the params structure instead of reading registers directly. This will permit t4_read_chip_settings() to be reused for VF devices which cannot read SGE registers directly. While here, move the call to t4_init_sge_params() to get_params__post_init(). The VF driver will populate the SGE parameters structure via a different method before calling t4_read_chip_settings(). Reviewed by: np Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7476
2016-08-15 17:40:05 +00:00
t4_init_sge_params(sc);
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
/*
* We've got the params we wanted to query via the firmware. Now grab
* some others directly from the chip.
*/
rc = t4_read_chip_settings(sc);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
return (rc);
}
cxgbe(4): Ask the card's firmware to pad up tiny CPLs by encapsulating them in a firmware message if it is able to do so. This works out better for one of the FIFOs in the chip. MFC after: 5 days
2013-02-26 00:27:27 +00:00
static int
set_params__post_init(struct adapter *sc)
{
uint32_t param, val;
cxgbe(4): Allow the TOE timer tunables to be set with microsecond precision. These timers are already displayed in microseconds in the sysctl MIB. Add variables to track these tunables while here. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 15:57:10 +00:00
#ifdef TCP_OFFLOAD
cxgbe(4): Display some more TOE parameters related to retransmission and keepalive in the sysctl MIB. Provide tunables to change some of these parameters. These are supposed to be setup by the firmware so these tunables are for experimentation only. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-24 19:17:13 +00:00
int i, v, shift;
cxgbe(4): Allow the TOE timer tunables to be set with microsecond precision. These timers are already displayed in microseconds in the sysctl MIB. Add variables to track these tunables while here. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 15:57:10 +00:00
#endif
cxgbe(4): Ask the card's firmware to pad up tiny CPLs by encapsulating them in a firmware message if it is able to do so. This works out better for one of the FIFOs in the chip. MFC after: 5 days
2013-02-26 00:27:27 +00:00
There is no need for elaborate queries and error checking when trying to set FW4MSG_ENCAP. MFC after: 3 days
2013-04-11 21:15:35 +00:00
/* ask for encapsulated CPLs */
cxgbe(4): Ask the card's firmware to pad up tiny CPLs by encapsulating them in a firmware message if it is able to do so. This works out better for one of the FIFOs in the chip. MFC after: 5 days
2013-02-26 00:27:27 +00:00
param = FW_PARAM_PFVF(CPLFW4MSG_ENCAP);
There is no need for elaborate queries and error checking when trying to set FW4MSG_ENCAP. MFC after: 3 days
2013-04-11 21:15:35 +00:00
val = 1;
(void)t4_set_params(sc, sc->mbox, sc->pf, 0, 1, &param, &val);
cxgbe(4): Ask the card's firmware to pad up tiny CPLs by encapsulating them in a firmware message if it is able to do so. This works out better for one of the FIFOs in the chip. MFC after: 5 days
2013-02-26 00:27:27 +00:00
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
/* Enable 32b port caps if the firmware supports it. */
param = FW_PARAM_PFVF(PORT_CAPS32);
val = 1;
if (t4_set_params(sc, sc->mbox, sc->pf, 0, 1, &param, &val) == 0)
sc->params.port_caps32 = 1;
cxgbe(4): Allow "pass" filters to distribute matching traffic using a subset of a VI's RSS indirection table. This makes it possible to make groups out of rx queues and steer different kinds of traffic to different groups. For example, an interface with 8 rx queues could have all non-TCP traffic delivered to queues 0-3 and all TCP traffic to queues 4-7. Note that it is already possible for filters to steer traffic to a particular queue or to distribute it using the full indirection table (much like normal rx does). Sponsored by: Chelsio Communications
2018-10-25 14:37:26 +00:00
/* Let filter + maskhash steer to a part of the VI's RSS region. */
val = 1 << (G_MASKSIZE(t4_read_reg(sc, A_TP_RSS_CONFIG_TNL)) - 1);
t4_set_reg_field(sc, A_TP_RSS_CONFIG_TNL, V_MASKFILTER(M_MASKFILTER),
V_MASKFILTER(val - 1));
cxgbe(4): Allow the TOE timer tunables to be set with microsecond precision. These timers are already displayed in microseconds in the sysctl MIB. Add variables to track these tunables while here. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 15:57:10 +00:00
#ifdef TCP_OFFLOAD
cxgbe(4): Display some more TOE parameters related to retransmission and keepalive in the sysctl MIB. Provide tunables to change some of these parameters. These are supposed to be setup by the firmware so these tunables are for experimentation only. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-24 19:17:13 +00:00
/*
* Override the TOE timers with user provided tunables. This is not the
* recommended way to change the timers (the firmware config file is) so
* these tunables are not documented.
*
cxgbe(4): Allow the TOE timer tunables to be set with microsecond precision. These timers are already displayed in microseconds in the sysctl MIB. Add variables to track these tunables while here. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 15:57:10 +00:00
* All the timer tunables are in microseconds.
cxgbe(4): Display some more TOE parameters related to retransmission and keepalive in the sysctl MIB. Provide tunables to change some of these parameters. These are supposed to be setup by the firmware so these tunables are for experimentation only. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-24 19:17:13 +00:00
*/
cxgbe(4): Allow the TOE timer tunables to be set with microsecond precision. These timers are already displayed in microseconds in the sysctl MIB. Add variables to track these tunables while here. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 15:57:10 +00:00
if (t4_toe_keepalive_idle != 0) {
v = us_to_tcp_ticks(sc, t4_toe_keepalive_idle);
v &= M_KEEPALIVEIDLE;
cxgbe(4): Display some more TOE parameters related to retransmission and keepalive in the sysctl MIB. Provide tunables to change some of these parameters. These are supposed to be setup by the firmware so these tunables are for experimentation only. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-24 19:17:13 +00:00
t4_set_reg_field(sc, A_TP_KEEP_IDLE,
cxgbe(4): Allow the TOE timer tunables to be set with microsecond precision. These timers are already displayed in microseconds in the sysctl MIB. Add variables to track these tunables while here. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 15:57:10 +00:00
V_KEEPALIVEIDLE(M_KEEPALIVEIDLE), V_KEEPALIVEIDLE(v));
cxgbe(4): Display some more TOE parameters related to retransmission and keepalive in the sysctl MIB. Provide tunables to change some of these parameters. These are supposed to be setup by the firmware so these tunables are for experimentation only. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-24 19:17:13 +00:00
}
cxgbe(4): Allow the TOE timer tunables to be set with microsecond precision. These timers are already displayed in microseconds in the sysctl MIB. Add variables to track these tunables while here. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 15:57:10 +00:00
if (t4_toe_keepalive_interval != 0) {
v = us_to_tcp_ticks(sc, t4_toe_keepalive_interval);
v &= M_KEEPALIVEINTVL;
cxgbe(4): Display some more TOE parameters related to retransmission and keepalive in the sysctl MIB. Provide tunables to change some of these parameters. These are supposed to be setup by the firmware so these tunables are for experimentation only. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-24 19:17:13 +00:00
t4_set_reg_field(sc, A_TP_KEEP_INTVL,
cxgbe(4): Allow the TOE timer tunables to be set with microsecond precision. These timers are already displayed in microseconds in the sysctl MIB. Add variables to track these tunables while here. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 15:57:10 +00:00
V_KEEPALIVEINTVL(M_KEEPALIVEINTVL), V_KEEPALIVEINTVL(v));
cxgbe(4): Display some more TOE parameters related to retransmission and keepalive in the sysctl MIB. Provide tunables to change some of these parameters. These are supposed to be setup by the firmware so these tunables are for experimentation only. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-24 19:17:13 +00:00
}
cxgbe(4): Allow the TOE timer tunables to be set with microsecond precision. These timers are already displayed in microseconds in the sysctl MIB. Add variables to track these tunables while here. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 15:57:10 +00:00
if (t4_toe_keepalive_count != 0) {
v = t4_toe_keepalive_count & M_KEEPALIVEMAXR2;
cxgbe(4): Display some more TOE parameters related to retransmission and keepalive in the sysctl MIB. Provide tunables to change some of these parameters. These are supposed to be setup by the firmware so these tunables are for experimentation only. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-24 19:17:13 +00:00
t4_set_reg_field(sc, A_TP_SHIFT_CNT,
V_KEEPALIVEMAXR1(M_KEEPALIVEMAXR1) |
V_KEEPALIVEMAXR2(M_KEEPALIVEMAXR2),
V_KEEPALIVEMAXR1(1) | V_KEEPALIVEMAXR2(v));
}
cxgbe(4): Allow the TOE timer tunables to be set with microsecond precision. These timers are already displayed in microseconds in the sysctl MIB. Add variables to track these tunables while here. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 15:57:10 +00:00
if (t4_toe_rexmt_min != 0) {
v = us_to_tcp_ticks(sc, t4_toe_rexmt_min);
v &= M_RXTMIN;
cxgbe(4): Display some more TOE parameters related to retransmission and keepalive in the sysctl MIB. Provide tunables to change some of these parameters. These are supposed to be setup by the firmware so these tunables are for experimentation only. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-24 19:17:13 +00:00
t4_set_reg_field(sc, A_TP_RXT_MIN,
cxgbe(4): Allow the TOE timer tunables to be set with microsecond precision. These timers are already displayed in microseconds in the sysctl MIB. Add variables to track these tunables while here. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 15:57:10 +00:00
V_RXTMIN(M_RXTMIN), V_RXTMIN(v));
cxgbe(4): Display some more TOE parameters related to retransmission and keepalive in the sysctl MIB. Provide tunables to change some of these parameters. These are supposed to be setup by the firmware so these tunables are for experimentation only. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-24 19:17:13 +00:00
}
cxgbe(4): Allow the TOE timer tunables to be set with microsecond precision. These timers are already displayed in microseconds in the sysctl MIB. Add variables to track these tunables while here. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 15:57:10 +00:00
if (t4_toe_rexmt_max != 0) {
v = us_to_tcp_ticks(sc, t4_toe_rexmt_max);
v &= M_RXTMAX;
cxgbe(4): Display some more TOE parameters related to retransmission and keepalive in the sysctl MIB. Provide tunables to change some of these parameters. These are supposed to be setup by the firmware so these tunables are for experimentation only. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-24 19:17:13 +00:00
t4_set_reg_field(sc, A_TP_RXT_MAX,
cxgbe(4): Allow the TOE timer tunables to be set with microsecond precision. These timers are already displayed in microseconds in the sysctl MIB. Add variables to track these tunables while here. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 15:57:10 +00:00
V_RXTMAX(M_RXTMAX), V_RXTMAX(v));
cxgbe(4): Display some more TOE parameters related to retransmission and keepalive in the sysctl MIB. Provide tunables to change some of these parameters. These are supposed to be setup by the firmware so these tunables are for experimentation only. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-24 19:17:13 +00:00
}
cxgbe(4): Allow the TOE timer tunables to be set with microsecond precision. These timers are already displayed in microseconds in the sysctl MIB. Add variables to track these tunables while here. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 15:57:10 +00:00
if (t4_toe_rexmt_count != 0) {
v = t4_toe_rexmt_count & M_RXTSHIFTMAXR2;
cxgbe(4): Display some more TOE parameters related to retransmission and keepalive in the sysctl MIB. Provide tunables to change some of these parameters. These are supposed to be setup by the firmware so these tunables are for experimentation only. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-24 19:17:13 +00:00
t4_set_reg_field(sc, A_TP_SHIFT_CNT,
V_RXTSHIFTMAXR1(M_RXTSHIFTMAXR1) |
V_RXTSHIFTMAXR2(M_RXTSHIFTMAXR2),
V_RXTSHIFTMAXR1(1) | V_RXTSHIFTMAXR2(v));
}
cxgbe(4): Allow the TOE timer tunables to be set with microsecond precision. These timers are already displayed in microseconds in the sysctl MIB. Add variables to track these tunables while here. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 15:57:10 +00:00
for (i = 0; i < nitems(t4_toe_rexmt_backoff); i++) {
if (t4_toe_rexmt_backoff[i] != -1) {
v = t4_toe_rexmt_backoff[i] & M_TIMERBACKOFFINDEX0;
cxgbe(4): Display some more TOE parameters related to retransmission and keepalive in the sysctl MIB. Provide tunables to change some of these parameters. These are supposed to be setup by the firmware so these tunables are for experimentation only. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-24 19:17:13 +00:00
shift = (i & 3) << 3;
t4_set_reg_field(sc, A_TP_TCP_BACKOFF_REG0 + (i & ~3),
M_TIMERBACKOFFINDEX0 << shift, v << shift);
}
}
cxgbe(4): Allow the TOE timer tunables to be set with microsecond precision. These timers are already displayed in microseconds in the sysctl MIB. Add variables to track these tunables while here. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 15:57:10 +00:00
#endif
There is no need for elaborate queries and error checking when trying to set FW4MSG_ENCAP. MFC after: 3 days
2013-04-11 21:15:35 +00:00
return (0);
cxgbe(4): Ask the card's firmware to pad up tiny CPLs by encapsulating them in a firmware message if it is able to do so. This works out better for one of the FIFOs in the chip. MFC after: 5 days
2013-02-26 00:27:27 +00:00
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
#undef FW_PARAM_PFVF
#undef FW_PARAM_DEV
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
static void
t4_set_desc(struct adapter *sc)
{
char buf[128];
struct adapter_params *p = &sc->params;
cxgbe(4): Provide more details about the card in the sysctl MIB. dev.t5nex.0.%desc: Chelsio T580-CR dev.t5nex.0.hw_revision: 1 dev.t5nex.0.sn: PT13140042 dev.t5nex.0.pn: 110117150A0 dev.t5nex.0.ec: 0000000000000000 dev.t5nex.0.na: 0007432AF490 dev.t5nex.0.vpd_version: 3 dev.t5nex.0.scfg_version: 53255 dev.t5nex.0.bs_version: 1.1.0.0 dev.t5nex.0.er_version: 1.0.0.68 dev.t5nex.0.tp_version: 0.1.4.9 dev.t5nex.0.firmware_version: 1.16.2.0 Sponsored by: Chelsio Communications
2016-08-27 00:13:41 +00:00
snprintf(buf, sizeof(buf), "Chelsio %s", p->vpd.id);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
device_set_desc_copy(sc->dev, buf);
}
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
static inline void
cxgbe(4): Use ifm for ifmedia just like the rest of the kernel. No functional change.
2018-05-31 02:22:40 +00:00
ifmedia_add4(struct ifmedia *ifm, int m)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
{
cxgbe(4): Use ifm for ifmedia just like the rest of the kernel. No functional change.
2018-05-31 02:22:40 +00:00
ifmedia_add(ifm, m, 0, NULL);
ifmedia_add(ifm, m | IFM_ETH_TXPAUSE, 0, NULL);
ifmedia_add(ifm, m | IFM_ETH_RXPAUSE, 0, NULL);
ifmedia_add(ifm, m | IFM_ETH_TXPAUSE | IFM_ETH_RXPAUSE, 0, NULL);
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
}
- Provide accurate ifmedia information so that 40G ports/transceivers are displayed properly in ifconfig, etc. - Use the same number of tx and rx queues for a 40G port as for a 10G port. MFC after: 1 week
2013-04-30 05:51:52 +00:00
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
/*
* This is the selected media, which is not quite the same as the active media.
* The media line in ifconfig is "media: Ethernet selected (active)" if selected
* and active are not the same, and "media: Ethernet selected" otherwise.
*/
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
static void
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
set_current_media(struct port_info *pi)
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
{
struct link_config *lc;
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
struct ifmedia *ifm;
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
int mword;
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
u_int speed;
Deal correctly with 40G ports that don't have any transceiver plugged in. Do not claim that they have unknown tranceivers. MFC after: 3 days
2013-05-13 20:00:03 +00:00
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
PORT_LOCK_ASSERT_OWNED(pi);
- Provide accurate ifmedia information so that 40G ports/transceivers are displayed properly in ifconfig, etc. - Use the same number of tx and rx queues for a 40G port as for a 10G port. MFC after: 1 week
2013-04-30 05:51:52 +00:00
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
/* Leave current media alone if it's already set to IFM_NONE. */
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
ifm = &pi->media;
cxgbe(4): Use ifm for ifmedia just like the rest of the kernel. No functional change.
2018-05-31 02:22:40 +00:00
if (ifm->ifm_cur != NULL &&
IFM_SUBTYPE(ifm->ifm_cur->ifm_media) == IFM_NONE)
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return;
Deal correctly with 40G ports that don't have any transceiver plugged in. Do not claim that they have unknown tranceivers. MFC after: 3 days
2013-05-13 20:00:03 +00:00
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
lc = &pi->link_cfg;
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
if (lc->requested_aneg != AUTONEG_DISABLE &&
lc->supported & FW_PORT_CAP32_ANEG) {
cxgbe(4): Include full duplex mediaopt in media that can be reported as active. Always report full duplex in active media. Sponsored by: Chelsio Communications
2018-06-01 16:46:29 +00:00
ifmedia_set(ifm, IFM_ETHER | IFM_AUTO);
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return;
}
cxgbe(4): Include full duplex mediaopt in media that can be reported as active. Always report full duplex in active media. Sponsored by: Chelsio Communications
2018-06-01 16:46:29 +00:00
mword = IFM_ETHER | IFM_FDX;
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
if (lc->requested_fc & PAUSE_TX)
mword |= IFM_ETH_TXPAUSE;
if (lc->requested_fc & PAUSE_RX)
mword |= IFM_ETH_RXPAUSE;
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
if (lc->requested_speed == 0)
speed = port_top_speed(pi) * 1000; /* Gbps -> Mbps */
else
speed = lc->requested_speed;
mword |= port_mword(pi, speed_to_fwcap(speed));
cxgbe(4): Use ifm for ifmedia just like the rest of the kernel. No functional change.
2018-05-31 02:22:40 +00:00
ifmedia_set(ifm, mword);
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
/*
* Returns true if the ifmedia list for the port cannot change.
*/
static bool
fixed_ifmedia(struct port_info *pi)
{
return (pi->port_type == FW_PORT_TYPE_BT_SGMII ||
pi->port_type == FW_PORT_TYPE_BT_XFI ||
pi->port_type == FW_PORT_TYPE_BT_XAUI ||
pi->port_type == FW_PORT_TYPE_KX4 ||
pi->port_type == FW_PORT_TYPE_KX ||
pi->port_type == FW_PORT_TYPE_KR ||
pi->port_type == FW_PORT_TYPE_BP_AP ||
pi->port_type == FW_PORT_TYPE_BP4_AP ||
pi->port_type == FW_PORT_TYPE_BP40_BA ||
pi->port_type == FW_PORT_TYPE_KR4_100G ||
pi->port_type == FW_PORT_TYPE_KR_SFP28 ||
pi->port_type == FW_PORT_TYPE_KR_XLAUI);
}
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
static void
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
build_medialist(struct port_info *pi)
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
{
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
uint32_t ss, speed;
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
int unknown, mword, bit;
struct link_config *lc;
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
struct ifmedia *ifm;
cxgbe(4): Add support for additional port types and link speeds. Sponsored by: Chelsio Communications.
2016-09-11 23:08:57 +00:00
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
PORT_LOCK_ASSERT_OWNED(pi);
cxgbe(4): Add support for additional port types and link speeds. Sponsored by: Chelsio Communications.
2016-09-11 23:08:57 +00:00
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
if (pi->flags & FIXED_IFMEDIA)
return;
cxgbe(4): Add support for additional port types and link speeds. Sponsored by: Chelsio Communications.
2016-09-11 23:08:57 +00:00
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
/*
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
* Rebuild the ifmedia list.
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
*/
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
ifm = &pi->media;
cxgbe(4): Use ifm for ifmedia just like the rest of the kernel. No functional change.
2018-05-31 02:22:40 +00:00
ifmedia_removeall(ifm);
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
lc = &pi->link_cfg;
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
ss = G_FW_PORT_CAP32_SPEED(lc->supported); /* Supported Speeds */
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
if (__predict_false(ss == 0)) { /* not supposed to happen. */
MPASS(ss != 0);
no_media:
cxgbe(4): Use ifm for ifmedia just like the rest of the kernel. No functional change.
2018-05-31 02:22:40 +00:00
MPASS(LIST_EMPTY(&ifm->ifm_list));
ifmedia_add(ifm, IFM_ETHER | IFM_NONE, 0, NULL);
ifmedia_set(ifm, IFM_ETHER | IFM_NONE);
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return;
}
cxgbe(4): Add support for additional port types and link speeds. Sponsored by: Chelsio Communications.
2016-09-11 23:08:57 +00:00
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
unknown = 0;
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
for (bit = S_FW_PORT_CAP32_SPEED; bit < fls(ss); bit++) {
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
speed = 1 << bit;
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
MPASS(speed & M_FW_PORT_CAP32_SPEED);
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
if (ss & speed) {
mword = port_mword(pi, speed);
if (mword == IFM_NONE) {
goto no_media;
} else if (mword == IFM_UNKNOWN)
unknown++;
else
cxgbe(4): Include full duplex mediaopt in media that can be reported as active. Always report full duplex in active media. Sponsored by: Chelsio Communications
2018-06-01 16:46:29 +00:00
ifmedia_add4(ifm, IFM_ETHER | IFM_FDX | mword);
cxgbe(4): Add support for additional port types and link speeds. Sponsored by: Chelsio Communications.
2016-09-11 23:08:57 +00:00
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
if (unknown > 0) /* Add one unknown for all unknown media types. */
cxgbe(4): Include full duplex mediaopt in media that can be reported as active. Always report full duplex in active media. Sponsored by: Chelsio Communications
2018-06-01 16:46:29 +00:00
ifmedia_add4(ifm, IFM_ETHER | IFM_FDX | IFM_UNKNOWN);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
if (lc->supported & FW_PORT_CAP32_ANEG)
cxgbe(4): Use ifm for ifmedia just like the rest of the kernel. No functional change.
2018-05-31 02:22:40 +00:00
ifmedia_add(ifm, IFM_ETHER | IFM_AUTO, 0, NULL);
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
set_current_media(pi);
cxgbe(4): Various link/media related improvements. - Deal with changes to port_type, and not just port_mod when a transceiver is changed. This fixes hot swapping of transceivers of different types (QSFP+ or QSA or QSFP28 in a QSFP28 port, SFP+ or SFP28 in a SFP28 port, etc.). - Always refresh media information for ifconfig if the port is down. The firmware does not generate tranceiver-change interrupts unless at least one VI is enabled on the physical port. Before this change ifconfig diplayed potentially stale information for ports that were administratively down. - Always recalculate and reapply L1 config on a transceiver change. - Display PAUSE settings in ifconfig. The driver sysctls for this continue to work as well. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-17 00:42:13 +00:00
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Various link/media related improvements. - Deal with changes to port_type, and not just port_mod when a transceiver is changed. This fixes hot swapping of transceivers of different types (QSFP+ or QSA or QSFP28 in a QSFP28 port, SFP+ or SFP28 in a SFP28 port, etc.). - Always refresh media information for ifconfig if the port is down. The firmware does not generate tranceiver-change interrupts unless at least one VI is enabled on the physical port. Before this change ifconfig diplayed potentially stale information for ports that were administratively down. - Always recalculate and reapply L1 config on a transceiver change. - Display PAUSE settings in ifconfig. The driver sysctls for this continue to work as well. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-17 00:42:13 +00:00
/*
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
* Initialize the requested fields in the link config based on driver tunables.
cxgbe(4): Various link/media related improvements. - Deal with changes to port_type, and not just port_mod when a transceiver is changed. This fixes hot swapping of transceivers of different types (QSFP+ or QSA or QSFP28 in a QSFP28 port, SFP+ or SFP28 in a SFP28 port, etc.). - Always refresh media information for ifconfig if the port is down. The firmware does not generate tranceiver-change interrupts unless at least one VI is enabled on the physical port. Before this change ifconfig diplayed potentially stale information for ports that were administratively down. - Always recalculate and reapply L1 config on a transceiver change. - Display PAUSE settings in ifconfig. The driver sysctls for this continue to work as well. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-17 00:42:13 +00:00
*/
static void
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
init_link_config(struct port_info *pi)
cxgbe(4): Various link/media related improvements. - Deal with changes to port_type, and not just port_mod when a transceiver is changed. This fixes hot swapping of transceivers of different types (QSFP+ or QSA or QSFP28 in a QSFP28 port, SFP+ or SFP28 in a SFP28 port, etc.). - Always refresh media information for ifconfig if the port is down. The firmware does not generate tranceiver-change interrupts unless at least one VI is enabled on the physical port. Before this change ifconfig diplayed potentially stale information for ports that were administratively down. - Always recalculate and reapply L1 config on a transceiver change. - Display PAUSE settings in ifconfig. The driver sysctls for this continue to work as well. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-17 00:42:13 +00:00
{
struct link_config *lc = &pi->link_cfg;
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
PORT_LOCK_ASSERT_OWNED(pi);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
lc->requested_speed = 0;
cxgbe(4): Various link/media related improvements. - Deal with changes to port_type, and not just port_mod when a transceiver is changed. This fixes hot swapping of transceivers of different types (QSFP+ or QSA or QSFP28 in a QSFP28 port, SFP+ or SFP28 in a SFP28 port, etc.). - Always refresh media information for ifconfig if the port is down. The firmware does not generate tranceiver-change interrupts unless at least one VI is enabled on the physical port. Before this change ifconfig diplayed potentially stale information for ports that were administratively down. - Always recalculate and reapply L1 config on a transceiver change. - Display PAUSE settings in ifconfig. The driver sysctls for this continue to work as well. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-17 00:42:13 +00:00
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
if (t4_autoneg == 0)
cxgbe(4): Various link/media related improvements. - Deal with changes to port_type, and not just port_mod when a transceiver is changed. This fixes hot swapping of transceivers of different types (QSFP+ or QSA or QSFP28 in a QSFP28 port, SFP+ or SFP28 in a SFP28 port, etc.). - Always refresh media information for ifconfig if the port is down. The firmware does not generate tranceiver-change interrupts unless at least one VI is enabled on the physical port. Before this change ifconfig diplayed potentially stale information for ports that were administratively down. - Always recalculate and reapply L1 config on a transceiver change. - Display PAUSE settings in ifconfig. The driver sysctls for this continue to work as well. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-17 00:42:13 +00:00
lc->requested_aneg = AUTONEG_DISABLE;
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
else if (t4_autoneg == 1)
lc->requested_aneg = AUTONEG_ENABLE;
else
lc->requested_aneg = AUTONEG_AUTO;
lc->requested_fc = t4_pause_settings & (PAUSE_TX | PAUSE_RX |
PAUSE_AUTONEG);
if (t4_fec == -1 || t4_fec & FEC_AUTO)
lc->requested_fec = FEC_AUTO;
else {
lc->requested_fec = FEC_NONE;
if (t4_fec & FEC_RS)
lc->requested_fec |= FEC_RS;
if (t4_fec & FEC_BASER_RS)
lc->requested_fec |= FEC_BASER_RS;
cxgbe(4): Various link/media related improvements. - Deal with changes to port_type, and not just port_mod when a transceiver is changed. This fixes hot swapping of transceivers of different types (QSFP+ or QSA or QSFP28 in a QSFP28 port, SFP+ or SFP28 in a SFP28 port, etc.). - Always refresh media information for ifconfig if the port is down. The firmware does not generate tranceiver-change interrupts unless at least one VI is enabled on the physical port. Before this change ifconfig diplayed potentially stale information for ports that were administratively down. - Always recalculate and reapply L1 config on a transceiver change. - Display PAUSE settings in ifconfig. The driver sysctls for this continue to work as well. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-17 00:42:13 +00:00
}
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
}
cxgbe(4): Various link/media related improvements. - Deal with changes to port_type, and not just port_mod when a transceiver is changed. This fixes hot swapping of transceivers of different types (QSFP+ or QSA or QSFP28 in a QSFP28 port, SFP+ or SFP28 in a SFP28 port, etc.). - Always refresh media information for ifconfig if the port is down. The firmware does not generate tranceiver-change interrupts unless at least one VI is enabled on the physical port. Before this change ifconfig diplayed potentially stale information for ports that were administratively down. - Always recalculate and reapply L1 config on a transceiver change. - Display PAUSE settings in ifconfig. The driver sysctls for this continue to work as well. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-17 00:42:13 +00:00
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
/*
* Makes sure that all requested settings comply with what's supported by the
* port. Returns the number of settings that were invalid and had to be fixed.
*/
static int
fixup_link_config(struct port_info *pi)
{
int n = 0;
struct link_config *lc = &pi->link_cfg;
uint32_t fwspeed;
cxgbe(4): Various link/media related improvements. - Deal with changes to port_type, and not just port_mod when a transceiver is changed. This fixes hot swapping of transceivers of different types (QSFP+ or QSA or QSFP28 in a QSFP28 port, SFP+ or SFP28 in a SFP28 port, etc.). - Always refresh media information for ifconfig if the port is down. The firmware does not generate tranceiver-change interrupts unless at least one VI is enabled on the physical port. Before this change ifconfig diplayed potentially stale information for ports that were administratively down. - Always recalculate and reapply L1 config on a transceiver change. - Display PAUSE settings in ifconfig. The driver sysctls for this continue to work as well. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-17 00:42:13 +00:00
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
PORT_LOCK_ASSERT_OWNED(pi);
/* Speed (when not autonegotiating) */
if (lc->requested_speed != 0) {
fwspeed = speed_to_fwcap(lc->requested_speed);
if ((fwspeed & lc->supported) == 0) {
n++;
lc->requested_speed = 0;
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
}
cxgbe(4): Various link/media related improvements. - Deal with changes to port_type, and not just port_mod when a transceiver is changed. This fixes hot swapping of transceivers of different types (QSFP+ or QSA or QSFP28 in a QSFP28 port, SFP+ or SFP28 in a SFP28 port, etc.). - Always refresh media information for ifconfig if the port is down. The firmware does not generate tranceiver-change interrupts unless at least one VI is enabled on the physical port. Before this change ifconfig diplayed potentially stale information for ports that were administratively down. - Always recalculate and reapply L1 config on a transceiver change. - Display PAUSE settings in ifconfig. The driver sysctls for this continue to work as well. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-17 00:42:13 +00:00
}
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
/* Link autonegotiation */
MPASS(lc->requested_aneg == AUTONEG_ENABLE ||
lc->requested_aneg == AUTONEG_DISABLE ||
lc->requested_aneg == AUTONEG_AUTO);
if (lc->requested_aneg == AUTONEG_ENABLE &&
!(lc->supported & FW_PORT_CAP32_ANEG)) {
n++;
lc->requested_aneg = AUTONEG_AUTO;
}
/* Flow control */
MPASS((lc->requested_fc & ~(PAUSE_TX | PAUSE_RX | PAUSE_AUTONEG)) == 0);
if (lc->requested_fc & PAUSE_TX &&
!(lc->supported & FW_PORT_CAP32_FC_TX)) {
n++;
lc->requested_fc &= ~PAUSE_TX;
}
if (lc->requested_fc & PAUSE_RX &&
!(lc->supported & FW_PORT_CAP32_FC_RX)) {
n++;
lc->requested_fc &= ~PAUSE_RX;
}
if (!(lc->requested_fc & PAUSE_AUTONEG) &&
!(lc->supported & FW_PORT_CAP32_FORCE_PAUSE)) {
n++;
lc->requested_fc |= PAUSE_AUTONEG;
}
/* FEC */
if ((lc->requested_fec & FEC_RS &&
!(lc->supported & FW_PORT_CAP32_FEC_RS)) ||
(lc->requested_fec & FEC_BASER_RS &&
!(lc->supported & FW_PORT_CAP32_FEC_BASER_RS))) {
n++;
lc->requested_fec = FEC_AUTO;
}
return (n);
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
}
/*
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
* Apply the requested L1 settings, which are expected to be valid, to the
* hardware.
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
*/
static int
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
apply_link_config(struct port_info *pi)
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
{
struct adapter *sc = pi->adapter;
struct link_config *lc = &pi->link_cfg;
int rc;
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
#ifdef INVARIANTS
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
ASSERT_SYNCHRONIZED_OP(sc);
PORT_LOCK_ASSERT_OWNED(pi);
cxgbe(4): Various link/media related improvements. - Deal with changes to port_type, and not just port_mod when a transceiver is changed. This fixes hot swapping of transceivers of different types (QSFP+ or QSA or QSFP28 in a QSFP28 port, SFP+ or SFP28 in a SFP28 port, etc.). - Always refresh media information for ifconfig if the port is down. The firmware does not generate tranceiver-change interrupts unless at least one VI is enabled on the physical port. Before this change ifconfig diplayed potentially stale information for ports that were administratively down. - Always recalculate and reapply L1 config on a transceiver change. - Display PAUSE settings in ifconfig. The driver sysctls for this continue to work as well. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-17 00:42:13 +00:00
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
if (lc->requested_aneg == AUTONEG_ENABLE)
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
MPASS(lc->supported & FW_PORT_CAP32_ANEG);
if (!(lc->requested_fc & PAUSE_AUTONEG))
MPASS(lc->supported & FW_PORT_CAP32_FORCE_PAUSE);
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
if (lc->requested_fc & PAUSE_TX)
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
MPASS(lc->supported & FW_PORT_CAP32_FC_TX);
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
if (lc->requested_fc & PAUSE_RX)
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
MPASS(lc->supported & FW_PORT_CAP32_FC_RX);
if (lc->requested_fec & FEC_RS)
MPASS(lc->supported & FW_PORT_CAP32_FEC_RS);
if (lc->requested_fec & FEC_BASER_RS)
MPASS(lc->supported & FW_PORT_CAP32_FEC_BASER_RS);
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
#endif
cxgbe(4): Various link/media related improvements. - Deal with changes to port_type, and not just port_mod when a transceiver is changed. This fixes hot swapping of transceivers of different types (QSFP+ or QSA or QSFP28 in a QSFP28 port, SFP+ or SFP28 in a SFP28 port, etc.). - Always refresh media information for ifconfig if the port is down. The firmware does not generate tranceiver-change interrupts unless at least one VI is enabled on the physical port. Before this change ifconfig diplayed potentially stale information for ports that were administratively down. - Always recalculate and reapply L1 config on a transceiver change. - Display PAUSE settings in ifconfig. The driver sysctls for this continue to work as well. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-17 00:42:13 +00:00
rc = -t4_link_l1cfg(sc, sc->mbox, pi->tx_chan, lc);
if (rc != 0) {
cxgbev(4): Updates to the VF driver to cope with recent ifmedia and ctrlq changes in the base driver. MFC after: 1 week Sponsored by: Chelsio Communications
2018-08-23 00:58:10 +00:00
/* Don't complain if the VF driver gets back an EPERM. */
if (!(sc->flags & IS_VF) || rc != FW_EPERM)
device_printf(pi->dev, "l1cfg failed: %d\n", rc);
cxgbe(4): Various link/media related improvements. - Deal with changes to port_type, and not just port_mod when a transceiver is changed. This fixes hot swapping of transceivers of different types (QSFP+ or QSA or QSFP28 in a QSFP28 port, SFP+ or SFP28 in a SFP28 port, etc.). - Always refresh media information for ifconfig if the port is down. The firmware does not generate tranceiver-change interrupts unless at least one VI is enabled on the physical port. Before this change ifconfig diplayed potentially stale information for ports that were administratively down. - Always recalculate and reapply L1 config on a transceiver change. - Display PAUSE settings in ifconfig. The driver sysctls for this continue to work as well. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-17 00:42:13 +00:00
} else {
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
/*
* An L1_CFG will almost always result in a link-change event if
* the link is up, and the driver will refresh the actual
* fec/fc/etc. when the notification is processed. If the link
* is down then the actual settings are meaningless.
*
* This takes care of the case where a change in the L1 settings
* may not result in a notification.
*/
if (lc->link_ok && !(lc->requested_fc & PAUSE_AUTONEG))
lc->fc = lc->requested_fc & (PAUSE_TX | PAUSE_RX);
cxgbe(4): Various link/media related improvements. - Deal with changes to port_type, and not just port_mod when a transceiver is changed. This fixes hot swapping of transceivers of different types (QSFP+ or QSA or QSFP28 in a QSFP28 port, SFP+ or SFP28 in a SFP28 port, etc.). - Always refresh media information for ifconfig if the port is down. The firmware does not generate tranceiver-change interrupts unless at least one VI is enabled on the physical port. Before this change ifconfig diplayed potentially stale information for ports that were administratively down. - Always recalculate and reapply L1 config on a transceiver change. - Display PAUSE settings in ifconfig. The driver sysctls for this continue to work as well. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-17 00:42:13 +00:00
}
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
return (rc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
Program the MAC exact match table in batches of 7 addresses at a time when possible. This is more efficient than one at a time. Submitted by: gnn MFC after: 3 days
2012-02-08 00:36:36 +00:00
#define FW_MAC_EXACT_CHUNK 7
Convert to if_foreach_llmaddr() KPI.
2019-10-21 18:11:11 +00:00
struct mcaddr_ctx {
struct ifnet *ifp;
const uint8_t *mcaddr[FW_MAC_EXACT_CHUNK];
uint64_t hash;
int i;
int del;
int rc;
};
static u_int
add_maddr(void *arg, struct sockaddr_dl *sdl, u_int cnt)
{
struct mcaddr_ctx *ctx = arg;
struct vi_info *vi = ctx->ifp->if_softc;
struct port_info *pi = vi->pi;
struct adapter *sc = pi->adapter;
if (ctx->rc < 0)
return (0);
ctx->mcaddr[ctx->i] = LLADDR(sdl);
MPASS(ETHER_IS_MULTICAST(ctx->mcaddr[ctx->i]));
ctx->i++;
if (ctx->i == FW_MAC_EXACT_CHUNK) {
ctx->rc = t4_alloc_mac_filt(sc, sc->mbox, vi->viid, ctx->del,
ctx->i, ctx->mcaddr, NULL, &ctx->hash, 0);
if (ctx->rc < 0) {
int j;
for (j = 0; j < ctx->i; j++) {
if_printf(ctx->ifp,
"failed to add mc address"
" %02x:%02x:%02x:"
"%02x:%02x:%02x rc=%d\n",
ctx->mcaddr[j][0], ctx->mcaddr[j][1],
ctx->mcaddr[j][2], ctx->mcaddr[j][3],
ctx->mcaddr[j][4], ctx->mcaddr[j][5],
-ctx->rc);
}
return (0);
}
ctx->del = 0;
ctx->i = 0;
}
return (1);
}
Program the MAC exact match table in batches of 7 addresses at a time when possible. This is more efficient than one at a time. Submitted by: gnn MFC after: 3 days
2012-02-08 00:36:36 +00:00
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
/*
* Program the port's XGMAC based on parameters in ifnet. The caller also
* indicates which parameters should be programmed (the rest are left alone).
*/
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
int
update_mac_settings(struct ifnet *ifp, int flags)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
{
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
int rc = 0;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct vi_info *vi = ifp->if_softc;
struct port_info *pi = vi->pi;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
struct adapter *sc = pi->adapter;
int mtu = -1, promisc = -1, allmulti = -1, vlanex = -1;
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
ASSERT_SYNCHRONIZED_OP(sc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
KASSERT(flags, ("%s: not told what to update.", __func__));
if (flags & XGMAC_MTU)
mtu = ifp->if_mtu;
if (flags & XGMAC_PROMISC)
promisc = ifp->if_flags & IFF_PROMISC ? 1 : 0;
if (flags & XGMAC_ALLMULTI)
allmulti = ifp->if_flags & IFF_ALLMULTI ? 1 : 0;
if (flags & XGMAC_VLANEX)
vlanex = ifp->if_capenable & IFCAP_VLAN_HWTAGGING ? 1 : 0;
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
if (flags & (XGMAC_MTU|XGMAC_PROMISC|XGMAC_ALLMULTI|XGMAC_VLANEX)) {
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
rc = -t4_set_rxmode(sc, sc->mbox, vi->viid, mtu, promisc,
allmulti, 1, vlanex, false);
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
if (rc) {
if_printf(ifp, "set_rxmode (%x) failed: %d\n", flags,
rc);
return (rc);
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
if (flags & XGMAC_UCADDR) {
uint8_t ucaddr[ETHER_ADDR_LEN];
bcopy(IF_LLADDR(ifp), ucaddr, sizeof(ucaddr));
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
rc = t4_change_mac(sc, sc->mbox, vi->viid, vi->xact_addr_filt,
cxgbe(4): Treat the viid as an opaque identifier. Recent firmwares prefer to use a different format for viid internally and this change allows them to do so. MFC after: 1 week Sponsored by: Chelsio Communications
2019-03-20 17:27:11 +00:00
ucaddr, true, &vi->smt_idx);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (rc < 0) {
rc = -rc;
if_printf(ifp, "change_mac failed: %d\n", rc);
return (rc);
} else {
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
vi->xact_addr_filt = rc;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
rc = 0;
}
}
if (flags & XGMAC_MCADDRS) {
Convert to if_foreach_llmaddr() KPI.
2019-10-21 18:11:11 +00:00
struct epoch_tracker et;
struct mcaddr_ctx ctx;
int j;
ctx.ifp = ifp;
ctx.hash = 0;
ctx.i = 0;
ctx.del = 1;
/*
* Unlike other drivers, we accumulate list of pointers into
* interface address lists and we need to keep it safe even
* after if_foreach_llmaddr() returns, thus we must enter the
* network epoch.
*/
NET_EPOCH_ENTER(et);
if_foreach_llmaddr(ifp, add_maddr, &ctx);
if (ctx.rc < 0) {
NET_EPOCH_EXIT(et);
rc = -ctx.rc;
return (rc);
Program the MAC exact match table in batches of 7 addresses at a time when possible. This is more efficient than one at a time. Submitted by: gnn MFC after: 3 days
2012-02-08 00:36:36 +00:00
}
Convert to if_foreach_llmaddr() KPI.
2019-10-21 18:11:11 +00:00
if (ctx.i > 0) {
rc = t4_alloc_mac_filt(sc, sc->mbox, vi->viid,
ctx.del, ctx.i, ctx.mcaddr, NULL, &ctx.hash, 0);
NET_EPOCH_EXIT(et);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (rc < 0) {
rc = -rc;
Convert to if_foreach_llmaddr() KPI.
2019-10-21 18:11:11 +00:00
for (j = 0; j < ctx.i; j++) {
Program the MAC exact match table in batches of 7 addresses at a time when possible. This is more efficient than one at a time. Submitted by: gnn MFC after: 3 days
2012-02-08 00:36:36 +00:00
if_printf(ifp,
"failed to add mc address"
" %02x:%02x:%02x:"
"%02x:%02x:%02x rc=%d\n",
Convert to if_foreach_llmaddr() KPI.
2019-10-21 18:11:11 +00:00
ctx.mcaddr[j][0], ctx.mcaddr[j][1],
ctx.mcaddr[j][2], ctx.mcaddr[j][3],
ctx.mcaddr[j][4], ctx.mcaddr[j][5],
Program the MAC exact match table in batches of 7 addresses at a time when possible. This is more efficient than one at a time. Submitted by: gnn MFC after: 3 days
2012-02-08 00:36:36 +00:00
rc);
}
Convert to if_foreach_llmaddr() KPI.
2019-10-21 18:11:11 +00:00
return (rc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
Convert to if_foreach_llmaddr() KPI.
2019-10-21 18:11:11 +00:00
} else
NET_EPOCH_EXIT(et);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Convert to if_foreach_llmaddr() KPI.
2019-10-21 18:11:11 +00:00
rc = -t4_set_addr_hash(sc, sc->mbox, vi->viid, 0, ctx.hash, 0);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (rc != 0)
if_printf(ifp, "failed to set mc address hash: %d", rc);
}
return (rc);
}
cxgbe(4): a change to the synchronization rules within the the driver. This is purely cosmetic because the new rules are already followed. MFC after: 1 week
2015-02-08 08:42:45 +00:00
/*
* {begin|end}_synchronized_op must be called from the same thread.
*/
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
int
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
begin_synchronized_op(struct adapter *sc, struct vi_info *vi, int flags,
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
char *wmesg)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
{
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
int rc, pri;
#ifdef WITNESS
/* the caller thinks it's ok to sleep, but is it really? */
if (flags & SLEEP_OK)
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
"begin_synchronized_op");
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
#endif
if (INTR_OK)
pri = PCATCH;
else
pri = 0;
ADAPTER_LOCK(sc);
for (;;) {
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (vi && IS_DOOMED(vi)) {
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
rc = ENXIO;
goto done;
}
if (!IS_BUSY(sc)) {
rc = 0;
break;
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
if (!(flags & SLEEP_OK)) {
rc = EBUSY;
goto done;
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
if (mtx_sleep(&sc->flags, &sc->sc_lock, pri, wmesg, 0)) {
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
rc = EINTR;
goto done;
}
}
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
KASSERT(!IS_BUSY(sc), ("%s: controller busy.", __func__));
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
SET_BUSY(sc);
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
#ifdef INVARIANTS
sc->last_op = wmesg;
sc->last_op_thr = curthread;
cxgbe(4): Save the flags for the last adapter-wide synchronized operation that was initiated successfully. (The caller and thread are already recorded). MFC after: 1 week
2015-08-19 15:40:03 +00:00
sc->last_op_flags = flags;
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
#endif
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
done:
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
if (!(flags & HOLD_LOCK) || rc)
ADAPTER_UNLOCK(sc);
return (rc);
}
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
/*
* Tell if_ioctl and if_init that the VI is going away. This is
* special variant of begin_synchronized_op and must be paired with a
* call to end_synchronized_op.
*/
void
doom_vi(struct adapter *sc, struct vi_info *vi)
{
ADAPTER_LOCK(sc);
SET_DOOMED(vi);
wakeup(&sc->flags);
while (IS_BUSY(sc))
mtx_sleep(&sc->flags, &sc->sc_lock, 0, "t4detach", 0);
SET_BUSY(sc);
#ifdef INVARIANTS
sc->last_op = "t4detach";
sc->last_op_thr = curthread;
sc->last_op_flags = 0;
#endif
ADAPTER_UNLOCK(sc);
}
cxgbe(4): a change to the synchronization rules within the the driver. This is purely cosmetic because the new rules are already followed. MFC after: 1 week
2015-02-08 08:42:45 +00:00
/*
* {begin|end}_synchronized_op must be called from the same thread.
*/
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
void
end_synchronized_op(struct adapter *sc, int flags)
{
if (flags & LOCK_HELD)
ADAPTER_LOCK_ASSERT_OWNED(sc);
else
ADAPTER_LOCK(sc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
KASSERT(IS_BUSY(sc), ("%s: controller not busy.", __func__));
CLR_BUSY(sc);
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
wakeup(&sc->flags);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
ADAPTER_UNLOCK(sc);
}
static int
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
cxgbe_init_synchronized(struct vi_info *vi)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
{
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct port_info *pi = vi->pi;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
struct adapter *sc = pi->adapter;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct ifnet *ifp = vi->ifp;
cxgbe(4): major tx rework. a) Front load as much work as possible in if_transmit, before any driver lock or software queue has to get involved. b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This is specifically for the tx multiqueue model where one of the if_transmit producer threads becomes the consumer and other producers carry on as usual. mp_ring is implemented as standalone code and it should be possible to use it in any driver with tx multiqueue. It also has: - the ability to enqueue/dequeue multiple items. This might become significant if packet batching is ever implemented. - an abdication mechanism to allow a thread to give up writing tx descriptors and have another if_transmit thread take over. A thread that's writing tx descriptors can end up doing so for an unbounded time period if a) there are other if_transmit threads continuously feeding the sofware queue, and b) the chip keeps up with whatever the thread is throwing at it. - accurate statistics about interesting events even when the stats come at the expense of additional branches/conditional code. The NIC txq lock is uncontested on the fast path at this point. I've left it there for synchronization with the control events (interface up/down, modload/unload). c) Add support for "type 1" coalescing work request in the normal NIC tx path. This work request is optimized for frames with a single item in the DMA gather list. These are very common when forwarding packets. Note that netmap tx in cxgbe already uses these "type 1" work requests. d) Do not request automatic cidx updates every 32 descriptors. Instead, request updates via bits in individual work requests (still every 32 descriptors approximately). Also, request an automatic final update when the queue idles after activity. This means NIC tx reclaim is still performed lazily but it will catch up quickly as soon as the queue idles. This seems to be the best middle ground and I'll probably do something similar for netmap tx as well. e) Implement a faster tx path for WRQs (used by TOE tx and control queues, _not_ by the normal NIC tx). Allow work requests to be written directly to the hardware descriptor ring if room is available. I will convert t4_tom and iw_cxgbe modules to this faster style gradually. MFC after: 2 months
2014-12-31 23:19:16 +00:00
int rc = 0, i;
struct sge_txq *txq;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
ASSERT_SYNCHRONIZED_OP(sc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
return (0); /* already running */
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
if (!(sc->flags & FULL_INIT_DONE) &&
((rc = adapter_full_init(sc)) != 0))
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
return (rc); /* error message displayed already */
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (!(vi->flags & VI_INIT_DONE) &&
((rc = vi_full_init(vi)) != 0))
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
return (rc); /* error message displayed already */
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
rc = update_mac_settings(ifp, XGMAC_ALL);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (rc)
goto done; /* error message displayed already */
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
PORT_LOCK(pi);
if (pi->up_vis == 0) {
t4_update_port_info(pi);
fixup_link_config(pi);
build_medialist(pi);
apply_link_config(pi);
}
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
rc = -t4_enable_vi(sc, sc->mbox, vi->viid, true, true);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (rc != 0) {
if_printf(ifp, "enable_vi failed: %d\n", rc);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
PORT_UNLOCK(pi);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
goto done;
}
cxgbe(4): major tx rework. a) Front load as much work as possible in if_transmit, before any driver lock or software queue has to get involved. b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This is specifically for the tx multiqueue model where one of the if_transmit producer threads becomes the consumer and other producers carry on as usual. mp_ring is implemented as standalone code and it should be possible to use it in any driver with tx multiqueue. It also has: - the ability to enqueue/dequeue multiple items. This might become significant if packet batching is ever implemented. - an abdication mechanism to allow a thread to give up writing tx descriptors and have another if_transmit thread take over. A thread that's writing tx descriptors can end up doing so for an unbounded time period if a) there are other if_transmit threads continuously feeding the sofware queue, and b) the chip keeps up with whatever the thread is throwing at it. - accurate statistics about interesting events even when the stats come at the expense of additional branches/conditional code. The NIC txq lock is uncontested on the fast path at this point. I've left it there for synchronization with the control events (interface up/down, modload/unload). c) Add support for "type 1" coalescing work request in the normal NIC tx path. This work request is optimized for frames with a single item in the DMA gather list. These are very common when forwarding packets. Note that netmap tx in cxgbe already uses these "type 1" work requests. d) Do not request automatic cidx updates every 32 descriptors. Instead, request updates via bits in individual work requests (still every 32 descriptors approximately). Also, request an automatic final update when the queue idles after activity. This means NIC tx reclaim is still performed lazily but it will catch up quickly as soon as the queue idles. This seems to be the best middle ground and I'll probably do something similar for netmap tx as well. e) Implement a faster tx path for WRQs (used by TOE tx and control queues, _not_ by the normal NIC tx). Allow work requests to be written directly to the hardware descriptor ring if room is available. I will convert t4_tom and iw_cxgbe modules to this faster style gradually. MFC after: 2 months
2014-12-31 23:19:16 +00:00
/*
* Can't fail from this point onwards. Review cxgbe_uninit_synchronized
* if this changes.
*/
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
for_each_txq(vi, i, txq) {
cxgbe(4): major tx rework. a) Front load as much work as possible in if_transmit, before any driver lock or software queue has to get involved. b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This is specifically for the tx multiqueue model where one of the if_transmit producer threads becomes the consumer and other producers carry on as usual. mp_ring is implemented as standalone code and it should be possible to use it in any driver with tx multiqueue. It also has: - the ability to enqueue/dequeue multiple items. This might become significant if packet batching is ever implemented. - an abdication mechanism to allow a thread to give up writing tx descriptors and have another if_transmit thread take over. A thread that's writing tx descriptors can end up doing so for an unbounded time period if a) there are other if_transmit threads continuously feeding the sofware queue, and b) the chip keeps up with whatever the thread is throwing at it. - accurate statistics about interesting events even when the stats come at the expense of additional branches/conditional code. The NIC txq lock is uncontested on the fast path at this point. I've left it there for synchronization with the control events (interface up/down, modload/unload). c) Add support for "type 1" coalescing work request in the normal NIC tx path. This work request is optimized for frames with a single item in the DMA gather list. These are very common when forwarding packets. Note that netmap tx in cxgbe already uses these "type 1" work requests. d) Do not request automatic cidx updates every 32 descriptors. Instead, request updates via bits in individual work requests (still every 32 descriptors approximately). Also, request an automatic final update when the queue idles after activity. This means NIC tx reclaim is still performed lazily but it will catch up quickly as soon as the queue idles. This seems to be the best middle ground and I'll probably do something similar for netmap tx as well. e) Implement a faster tx path for WRQs (used by TOE tx and control queues, _not_ by the normal NIC tx). Allow work requests to be written directly to the hardware descriptor ring if room is available. I will convert t4_tom and iw_cxgbe modules to this faster style gradually. MFC after: 2 months
2014-12-31 23:19:16 +00:00
TXQ_LOCK(txq);
txq->eq.flags |= EQ_ENABLED;
TXQ_UNLOCK(txq);
}
Add support for packet-sniffing tracers to cxgbe(4). This works with all T4 and T5 based cards and is useful for analyzing TSO, LRO, TOE, and for general purpose monitoring without tapping any cxgbe or cxl ifnet directly. Tracers on the T4/T5 chips provide access to Ethernet frames exactly as they were received from or transmitted on the wire. On transmit, a tracer will capture a frame after TSO segmentation, hw VLAN tag insertion, hw L3 & L4 checksum insertion, etc. It will also capture frames generated by the TCP offload engine (TOE traffic is normally invisible to the kernel). On receive, a tracer will capture a frame before hw VLAN extraction, runt filtering, other badness filtering, before the steering/drop/L2-rewrite filters or the TOE have had a go at it, and of course before sw LRO in the driver. There are 4 tracers on a chip. A tracer can trace only in one direction (tx or rx). For now cxgbetool will set up tracers to capture the first 128B of every transmitted or received frame on a given port. This is a small subset of what the hardware can do. A pseudo ifnet with the same name as the nexus driver (t4nex0 or t5nex0) will be created for tracing. The data delivered to this ifnet is an additional copy made inside the chip. Normal delivery to cxgbe<n> or cxl<n> will be made as usual. /* watch cxl0, which is the first port hanging off t5nex0. */ # cxgbetool t5nex0 tracer 0 tx0 (watch what cxl0 is transmitting) # cxgbetool t5nex0 tracer 1 rx0 (watch what cxl0 is receiving) # cxgbetool t5nex0 tracer list # tcpdump -i t5nex0 <== all that cxl0 sees and puts on the wire If you were doing TSO, a tcpdump on cxl0 may have shown you ~64K "frames" with no L3/L4 checksum but this will show you the frames that were actually transmitted. /* all done */ # cxgbetool t5nex0 tracer 0 disable # cxgbetool t5nex0 tracer 1 disable # cxgbetool t5nex0 tracer list # ifconfig t5nex0 destroy
2013-07-26 22:04:11 +00:00
/*
* The first iq of the first port to come up is used for tracing.
*/
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (sc->traceq < 0 && IS_MAIN_VI(vi)) {
sc->traceq = sc->sge.rxq[vi->first_rxq].iq.abs_id;
Add support for packet-sniffing tracers to cxgbe(4). This works with all T4 and T5 based cards and is useful for analyzing TSO, LRO, TOE, and for general purpose monitoring without tapping any cxgbe or cxl ifnet directly. Tracers on the T4/T5 chips provide access to Ethernet frames exactly as they were received from or transmitted on the wire. On transmit, a tracer will capture a frame after TSO segmentation, hw VLAN tag insertion, hw L3 & L4 checksum insertion, etc. It will also capture frames generated by the TCP offload engine (TOE traffic is normally invisible to the kernel). On receive, a tracer will capture a frame before hw VLAN extraction, runt filtering, other badness filtering, before the steering/drop/L2-rewrite filters or the TOE have had a go at it, and of course before sw LRO in the driver. There are 4 tracers on a chip. A tracer can trace only in one direction (tx or rx). For now cxgbetool will set up tracers to capture the first 128B of every transmitted or received frame on a given port. This is a small subset of what the hardware can do. A pseudo ifnet with the same name as the nexus driver (t4nex0 or t5nex0) will be created for tracing. The data delivered to this ifnet is an additional copy made inside the chip. Normal delivery to cxgbe<n> or cxl<n> will be made as usual. /* watch cxl0, which is the first port hanging off t5nex0. */ # cxgbetool t5nex0 tracer 0 tx0 (watch what cxl0 is transmitting) # cxgbetool t5nex0 tracer 1 rx0 (watch what cxl0 is receiving) # cxgbetool t5nex0 tracer list # tcpdump -i t5nex0 <== all that cxl0 sees and puts on the wire If you were doing TSO, a tcpdump on cxl0 may have shown you ~64K "frames" with no L3/L4 checksum but this will show you the frames that were actually transmitted. /* all done */ # cxgbetool t5nex0 tracer 0 disable # cxgbetool t5nex0 tracer 1 disable # cxgbetool t5nex0 tracer list # ifconfig t5nex0 destroy
2013-07-26 22:04:11 +00:00
t4_write_reg(sc, is_t4(sc) ? A_MPS_TRC_RSS_CONTROL :
A_MPS_T5_TRC_RSS_CONTROL, V_RSSCONTROL(pi->tx_chan) |
V_QUEUENUMBER(sc->traceq));
pi->flags |= HAS_TRACEQ;
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
/* all ok */
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
pi->up_vis++;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
ifp->if_drv_flags |= IFF_DRV_RUNNING;
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
if (pi->nvi > 1 || sc->flags & IS_VF)
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
callout_reset(&vi->tick, hz, vi_tick, vi);
else
callout_reset(&pi->tick, hz, cxgbe_tick, pi);
cxgbe(4): Don't forget to report link state to the kernel if the link is already up at attach. Reported by: Fabrice Bruel @ Orange Business Service MFC after: 1 week Sponsored by: Chelsio Communications
2019-03-01 02:43:30 +00:00
if (pi->link_cfg.link_ok)
t4_os_link_changed(pi);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
PORT_UNLOCK(pi);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
done:
if (rc != 0)
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
cxgbe_uninit_synchronized(vi);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
return (rc);
}
/*
* Idempotent.
*/
static int
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
cxgbe_uninit_synchronized(struct vi_info *vi)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
{
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct port_info *pi = vi->pi;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
struct adapter *sc = pi->adapter;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct ifnet *ifp = vi->ifp;
cxgbe(4): major tx rework. a) Front load as much work as possible in if_transmit, before any driver lock or software queue has to get involved. b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This is specifically for the tx multiqueue model where one of the if_transmit producer threads becomes the consumer and other producers carry on as usual. mp_ring is implemented as standalone code and it should be possible to use it in any driver with tx multiqueue. It also has: - the ability to enqueue/dequeue multiple items. This might become significant if packet batching is ever implemented. - an abdication mechanism to allow a thread to give up writing tx descriptors and have another if_transmit thread take over. A thread that's writing tx descriptors can end up doing so for an unbounded time period if a) there are other if_transmit threads continuously feeding the sofware queue, and b) the chip keeps up with whatever the thread is throwing at it. - accurate statistics about interesting events even when the stats come at the expense of additional branches/conditional code. The NIC txq lock is uncontested on the fast path at this point. I've left it there for synchronization with the control events (interface up/down, modload/unload). c) Add support for "type 1" coalescing work request in the normal NIC tx path. This work request is optimized for frames with a single item in the DMA gather list. These are very common when forwarding packets. Note that netmap tx in cxgbe already uses these "type 1" work requests. d) Do not request automatic cidx updates every 32 descriptors. Instead, request updates via bits in individual work requests (still every 32 descriptors approximately). Also, request an automatic final update when the queue idles after activity. This means NIC tx reclaim is still performed lazily but it will catch up quickly as soon as the queue idles. This seems to be the best middle ground and I'll probably do something similar for netmap tx as well. e) Implement a faster tx path for WRQs (used by TOE tx and control queues, _not_ by the normal NIC tx). Allow work requests to be written directly to the hardware descriptor ring if room is available. I will convert t4_tom and iw_cxgbe modules to this faster style gradually. MFC after: 2 months
2014-12-31 23:19:16 +00:00
int rc, i;
struct sge_txq *txq;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
ASSERT_SYNCHRONIZED_OP(sc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (!(vi->flags & VI_INIT_DONE)) {
cxgbe(4): Always display an error message if SIOCSIFFLAGS will leave IFF_UP and IFF_DRV_RUNNING out of sync. ifhwioctl in the kernel pays no attention to the return code from the driver ioctl during SIOCSIFFLAGS so these messages are the only indication that the ioctl was called but failed. MFC after: 1 week Sponsored by: Chelsio Communications
2018-04-04 22:52:24 +00:00
if (__predict_false(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
KASSERT(0, ("uninited VI is running"));
if_printf(ifp, "uninited VI with running ifnet. "
"vi->flags 0x%016lx, if_flags 0x%08x, "
"if_drv_flags 0x%08x\n", vi->flags, ifp->if_flags,
ifp->if_drv_flags);
}
Fix a couple of panics when detaching from a cxgbe/cxl interface that was never brought up: - Allow NULL to be passed to sglist_free(). - Don't try to stop an interface that was never fully initialized. Reviewed by: np
2015-01-26 16:26:28 +00:00
return (0);
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
/*
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
* Disable the VI so that all its data in either direction is discarded
* by the MPS. Leave everything else (the queues, interrupts, and 1Hz
* tick) intact as the TP can deliver negative advice or data that it's
* holding in its RAM (for an offloaded connection) even after the VI is
* disabled.
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
*/
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
rc = -t4_enable_vi(sc, sc->mbox, vi->viid, false, false);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
if (rc) {
if_printf(ifp, "disable_vi failed: %d\n", rc);
return (rc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
for_each_txq(vi, i, txq) {
cxgbe(4): major tx rework. a) Front load as much work as possible in if_transmit, before any driver lock or software queue has to get involved. b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This is specifically for the tx multiqueue model where one of the if_transmit producer threads becomes the consumer and other producers carry on as usual. mp_ring is implemented as standalone code and it should be possible to use it in any driver with tx multiqueue. It also has: - the ability to enqueue/dequeue multiple items. This might become significant if packet batching is ever implemented. - an abdication mechanism to allow a thread to give up writing tx descriptors and have another if_transmit thread take over. A thread that's writing tx descriptors can end up doing so for an unbounded time period if a) there are other if_transmit threads continuously feeding the sofware queue, and b) the chip keeps up with whatever the thread is throwing at it. - accurate statistics about interesting events even when the stats come at the expense of additional branches/conditional code. The NIC txq lock is uncontested on the fast path at this point. I've left it there for synchronization with the control events (interface up/down, modload/unload). c) Add support for "type 1" coalescing work request in the normal NIC tx path. This work request is optimized for frames with a single item in the DMA gather list. These are very common when forwarding packets. Note that netmap tx in cxgbe already uses these "type 1" work requests. d) Do not request automatic cidx updates every 32 descriptors. Instead, request updates via bits in individual work requests (still every 32 descriptors approximately). Also, request an automatic final update when the queue idles after activity. This means NIC tx reclaim is still performed lazily but it will catch up quickly as soon as the queue idles. This seems to be the best middle ground and I'll probably do something similar for netmap tx as well. e) Implement a faster tx path for WRQs (used by TOE tx and control queues, _not_ by the normal NIC tx). Allow work requests to be written directly to the hardware descriptor ring if room is available. I will convert t4_tom and iw_cxgbe modules to this faster style gradually. MFC after: 2 months
2014-12-31 23:19:16 +00:00
TXQ_LOCK(txq);
txq->eq.flags &= ~EQ_ENABLED;
TXQ_UNLOCK(txq);
}
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
PORT_LOCK(pi);
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
if (pi->nvi > 1 || sc->flags & IS_VF)
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
callout_stop(&vi->tick);
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
else
callout_stop(&pi->tick);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
PORT_UNLOCK(pi);
return (0);
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
pi->up_vis--;
if (pi->up_vis > 0) {
PORT_UNLOCK(pi);
return (0);
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
pi->link_cfg.link_ok = false;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
pi->link_cfg.speed = 0;
cxgbe(4): Updates to link configuration. - Update struct link_settings and associated shared code. - Add tunables to control FEC and autonegotiation. All ports inherit these values as their initial settings. hw.cxgbe.fec hw.cxgbe.autoneg - Add per-port sysctls to control FEC and autonegotiation. These can be modified at any time. dev.<port>.<n>.fec dev.<port>.<n>.autoneg MFC after: 3 days Sponsored by: Chelsio Communications
2016-12-30 08:59:49 +00:00
pi->link_cfg.link_down_rc = 255;
cxgbe(4): Save the last reported link parameters and compare them with the current state to determine whether to generate a link-state change notification. This fixes a bug introduced in r321063 that caused the driver to sometimes skip these notifications. Reported by: Jason Eggleston @ LLNW MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-12 14:02:19 +00:00
t4_os_link_changed(pi);
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
PORT_UNLOCK(pi);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
return (0);
}
Install interrupt handlers early, during attach, for the reason explained in r239913 by jhb. MFC after: 1 week
2012-09-13 09:18:13 +00:00
/*
* It is ok for this function to fail midway and return right away. t4_detach
* will walk the entire sc->irq list and clean up whatever is valid.
*/
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
int
t4_setup_intr_handlers(struct adapter *sc)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
{
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
int rc, rid, p, q, v;
- Specialized ingress queues that take interrupts for other ingress queues. Try to have a set of these per port when possible, fall back to sharing a common pool between all ports otherwise. - One control queue per port (used to be one per hardware channel). - t4_eth_rx now handles Ethernet rx only. - sysctls to display pidx/cidx for some queues. MFC after: 1 week
2011-05-30 21:34:44 +00:00
char s[8];
struct irq *irq;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
struct port_info *pi;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct vi_info *vi;
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
struct sge *sge = &sc->sge;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
struct sge_rxq *rxq;
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
#ifdef TCP_OFFLOAD
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
struct sge_ofld_rxq *ofld_rxq;
#endif
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
#ifdef DEV_NETMAP
struct sge_nm_rxq *nm_rxq;
#endif
cxgbe: bind the ithreads that handle NIC rx to the correct CPU if the kernel is built with option RSS.
2016-01-11 17:52:42 +00:00
#ifdef RSS
int nbuckets = rss_getnumbuckets();
#endif
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
/*
* Setup interrupts.
*/
- Specialized ingress queues that take interrupts for other ingress queues. Try to have a set of these per port when possible, fall back to sharing a common pool between all ports otherwise. - One control queue per port (used to be one per hardware channel). - t4_eth_rx now handles Ethernet rx only. - sysctls to display pidx/cidx for some queues. MFC after: 1 week
2011-05-30 21:34:44 +00:00
irq = &sc->irq[0];
rid = sc->intr_type == INTR_INTX ? 0 : 1;
cxgbe(4): Do not forward interrupts to queues with freelists. This leaves the firmware event queue (fwq) as the only queue that can take interrupts for others. This simplifies cfg_itype_and_nqueues and queue allocation in the driver at the cost of a little (never?) used configuration. It also allows service_iq to be split into two specialized variants in the future. MFC after: 2 months Sponsored by: Chelsio Communications
2017-12-22 19:10:19 +00:00
if (forwarding_intr_to_fwq(sc))
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
return (t4_alloc_irq(sc, irq, rid, t4_intr_all, sc, "all"));
- Specialized ingress queues that take interrupts for other ingress queues. Try to have a set of these per port when possible, fall back to sharing a common pool between all ports otherwise. - One control queue per port (used to be one per hardware channel). - t4_eth_rx now handles Ethernet rx only. - sysctls to display pidx/cidx for some queues. MFC after: 1 week
2011-05-30 21:34:44 +00:00
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
/* Multiple interrupts. */
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
if (sc->flags & IS_VF)
KASSERT(sc->intr_count >= T4VF_EXTRA_INTR + sc->params.nports,
("%s: too few intr.", __func__));
else
KASSERT(sc->intr_count >= T4_EXTRA_INTR + sc->params.nports,
("%s: too few intr.", __func__));
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
/* The first one is always error intr on PFs */
if (!(sc->flags & IS_VF)) {
rc = t4_alloc_irq(sc, irq, rid, t4_intr_err, sc, "err");
if (rc != 0)
return (rc);
irq++;
rid++;
}
- Specialized ingress queues that take interrupts for other ingress queues. Try to have a set of these per port when possible, fall back to sharing a common pool between all ports otherwise. - One control queue per port (used to be one per hardware channel). - t4_eth_rx now handles Ethernet rx only. - sysctls to display pidx/cidx for some queues. MFC after: 1 week
2011-05-30 21:34:44 +00:00
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
/* The second one is always the firmware event queue (first on VFs) */
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
rc = t4_alloc_irq(sc, irq, rid, t4_intr_evt, &sge->fwq, "evt");
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
if (rc != 0)
return (rc);
irq++;
rid++;
- Specialized ingress queues that take interrupts for other ingress queues. Try to have a set of these per port when possible, fall back to sharing a common pool between all ports otherwise. - One control queue per port (used to be one per hardware channel). - t4_eth_rx now handles Ethernet rx only. - sysctls to display pidx/cidx for some queues. MFC after: 1 week
2011-05-30 21:34:44 +00:00
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
for_each_port(sc, p) {
pi = sc->port[p];
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
for_each_vi(pi, v, vi) {
vi->first_intr = rid - 1;
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
if (vi->nnmrxq > 0) {
int n = max(vi->nrxq, vi->nnmrxq);
rxq = &sge->rxq[vi->first_rxq];
#ifdef DEV_NETMAP
nm_rxq = &sge->nm_rxq[vi->first_nm_rxq];
#endif
for (q = 0; q < n; q++) {
snprintf(s, sizeof(s), "%x%c%x", p,
'a' + v, q);
if (q < vi->nrxq)
irq->rxq = rxq++;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
#ifdef DEV_NETMAP
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
if (q < vi->nnmrxq)
irq->nm_rxq = nm_rxq++;
cxgbe(4): Create two variants of service_iq, one for queues with freelists and one for those without. MFH: 3 weeks Sponsored by: Chelsio Communications
2018-08-11 04:55:47 +00:00
if (irq->nm_rxq != NULL &&
irq->rxq == NULL) {
/* Netmap rx only */
rc = t4_alloc_irq(sc, irq, rid,
t4_nm_intr, irq->nm_rxq, s);
}
if (irq->nm_rxq != NULL &&
irq->rxq != NULL) {
/* NIC and Netmap rx */
rc = t4_alloc_irq(sc, irq, rid,
t4_vi_intr, irq, s);
}
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
#endif
cxgbe(4): Create two variants of service_iq, one for queues with freelists and one for those without. MFH: 3 weeks Sponsored by: Chelsio Communications
2018-08-11 04:55:47 +00:00
if (irq->rxq != NULL &&
irq->nm_rxq == NULL) {
/* NIC rx only */
rc = t4_alloc_irq(sc, irq, rid,
t4_intr, irq->rxq, s);
}
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (rc != 0)
return (rc);
cxgbe(4): Do not forward interrupts to queues with freelists. This leaves the firmware event queue (fwq) as the only queue that can take interrupts for others. This simplifies cfg_itype_and_nqueues and queue allocation in the driver at the cost of a little (never?) used configuration. It also allows service_iq to be split into two specialized variants in the future. MFC after: 2 months Sponsored by: Chelsio Communications
2017-12-22 19:10:19 +00:00
#ifdef RSS
if (q < vi->nrxq) {
bus_bind_intr(sc->dev, irq->res,
rss_getcpu(q % nbuckets));
}
#endif
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
irq++;
rid++;
vi->nintr++;
}
cxgbe(4): Do not forward interrupts to queues with freelists. This leaves the firmware event queue (fwq) as the only queue that can take interrupts for others. This simplifies cfg_itype_and_nqueues and queue allocation in the driver at the cost of a little (never?) used configuration. It also allows service_iq to be split into two specialized variants in the future. MFC after: 2 months Sponsored by: Chelsio Communications
2017-12-22 19:10:19 +00:00
} else {
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
for_each_rxq(vi, q, rxq) {
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
snprintf(s, sizeof(s), "%x%c%x", p,
'a' + v, q);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
rc = t4_alloc_irq(sc, irq, rid,
t4_intr, rxq, s);
if (rc != 0)
return (rc);
cxgbe: bind the ithreads that handle NIC rx to the correct CPU if the kernel is built with option RSS.
2016-01-11 17:52:42 +00:00
#ifdef RSS
bus_bind_intr(sc->dev, irq->res,
rss_getcpu(q % nbuckets));
#endif
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
irq++;
rid++;
vi->nintr++;
}
}
#ifdef TCP_OFFLOAD
cxgbe(4): Do not forward interrupts to queues with freelists. This leaves the firmware event queue (fwq) as the only queue that can take interrupts for others. This simplifies cfg_itype_and_nqueues and queue allocation in the driver at the cost of a little (never?) used configuration. It also allows service_iq to be split into two specialized variants in the future. MFC after: 2 months Sponsored by: Chelsio Communications
2017-12-22 19:10:19 +00:00
for_each_ofld_rxq(vi, q, ofld_rxq) {
snprintf(s, sizeof(s), "%x%c%x", p, 'A' + v, q);
rc = t4_alloc_irq(sc, irq, rid, t4_intr,
ofld_rxq, s);
if (rc != 0)
return (rc);
irq++;
rid++;
vi->nintr++;
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
}
#endif
Fix build for !TCP_OFFLOAD case.
2015-12-03 10:33:57 +00:00
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
MPASS(irq == &sc->irq[sc->intr_count]);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Install interrupt handlers early, during attach, for the reason explained in r239913 by jhb. MFC after: 1 week
2012-09-13 09:18:13 +00:00
return (0);
}
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
int
Install interrupt handlers early, during attach, for the reason explained in r239913 by jhb. MFC after: 1 week
2012-09-13 09:18:13 +00:00
adapter_full_init(struct adapter *sc)
{
int rc, i;
cxgbe(4): Program the global RSS key once instead of once per ifnet. MFC after: 3 days Sponsored by: Chelsio Communications
2017-04-04 23:14:03 +00:00
#ifdef RSS
uint32_t raw_rss_key[RSS_KEYSIZE / sizeof(uint32_t)];
uint32_t rss_key[RSS_KEYSIZE / sizeof(uint32_t)];
#endif
Install interrupt handlers early, during attach, for the reason explained in r239913 by jhb. MFC after: 1 week
2012-09-13 09:18:13 +00:00
cxgbe(4): adapter_full_init is always a synchronized operation. MFC after: 1 week
2015-02-08 08:52:18 +00:00
ASSERT_SYNCHRONIZED_OP(sc);
Install interrupt handlers early, during attach, for the reason explained in r239913 by jhb. MFC after: 1 week
2012-09-13 09:18:13 +00:00
ADAPTER_LOCK_ASSERT_NOTOWNED(sc);
KASSERT((sc->flags & FULL_INIT_DONE) == 0,
("%s: FULL_INIT_DONE already", __func__));
/*
* queues that belong to the adapter (not any particular port).
*/
rc = t4_setup_adapter_queues(sc);
if (rc != 0)
goto done;
for (i = 0; i < nitems(sc->tq); i++) {
sc->tq[i] = taskqueue_create("t4 taskq", M_NOWAIT,
taskqueue_thread_enqueue, &sc->tq[i]);
if (sc->tq[i] == NULL) {
device_printf(sc->dev,
"failed to allocate task queue %d\n", i);
rc = ENOMEM;
goto done;
}
taskqueue_start_threads(&sc->tq[i], 1, PI_NET, "%s tq%d",
device_get_nameunit(sc->dev), i);
}
cxgbe(4): Program the global RSS key once instead of once per ifnet. MFC after: 3 days Sponsored by: Chelsio Communications
2017-04-04 23:14:03 +00:00
#ifdef RSS
MPASS(RSS_KEYSIZE == 40);
rss_getkey((void *)&raw_rss_key[0]);
for (i = 0; i < nitems(rss_key); i++) {
rss_key[i] = htobe32(raw_rss_key[nitems(rss_key) - 1 - i]);
}
cxgbe(4): Some updates to the common code. - Updated register ranges. - Helper routines for access to TP registers. - Updated routine to read flash parameters. Obtained from: Chelsio Communications MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-26 20:20:58 +00:00
t4_write_rss_key(sc, &rss_key[0], -1, 1);
cxgbe(4): Program the global RSS key once instead of once per ifnet. MFC after: 3 days Sponsored by: Chelsio Communications
2017-04-04 23:14:03 +00:00
#endif
Install interrupt handlers early, during attach, for the reason explained in r239913 by jhb. MFC after: 1 week
2012-09-13 09:18:13 +00:00
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
if (!(sc->flags & IS_VF))
t4_intr_enable(sc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
sc->flags |= FULL_INIT_DONE;
done:
if (rc != 0)
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
adapter_full_uninit(sc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
return (rc);
}
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
int
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
adapter_full_uninit(struct adapter *sc)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
{
int i;
ADAPTER_LOCK_ASSERT_NOTOWNED(sc);
- Move all Ethernet specific items from sge_eq to sge_txq. sge_eq is now a suitable base for all kinds of egress queues. - Add control queues (sge_ctrlq) and allocate one of these per hardware channel. They can be used to program filters and steer traffic (and more). MFC after: 1 week
2011-04-19 22:08:28 +00:00
t4_teardown_adapter_queues(sc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Use native FreeBSD facilities everywhere except the shared code in common/ MFC after: 1 week
2012-09-13 09:10:10 +00:00
for (i = 0; i < nitems(sc->tq) && sc->tq[i]; i++) {
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
taskqueue_free(sc->tq[i]);
sc->tq[i] = NULL;
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
sc->flags &= ~FULL_INIT_DONE;
return (0);
}
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
#ifdef RSS
#define SUPPORTED_RSS_HASHTYPES (RSS_HASHTYPE_RSS_IPV4 | \
RSS_HASHTYPE_RSS_TCP_IPV4 | RSS_HASHTYPE_RSS_IPV6 | \
RSS_HASHTYPE_RSS_TCP_IPV6 | RSS_HASHTYPE_RSS_UDP_IPV4 | \
RSS_HASHTYPE_RSS_UDP_IPV6)
/* Translates kernel hash types to hardware. */
static int
hashconfig_to_hashen(int hashconfig)
{
int hashen = 0;
if (hashconfig & RSS_HASHTYPE_RSS_IPV4)
hashen |= F_FW_RSS_VI_CONFIG_CMD_IP4TWOTUPEN;
if (hashconfig & RSS_HASHTYPE_RSS_IPV6)
hashen |= F_FW_RSS_VI_CONFIG_CMD_IP6TWOTUPEN;
if (hashconfig & RSS_HASHTYPE_RSS_UDP_IPV4) {
hashen |= F_FW_RSS_VI_CONFIG_CMD_UDPEN |
F_FW_RSS_VI_CONFIG_CMD_IP4FOURTUPEN;
}
if (hashconfig & RSS_HASHTYPE_RSS_UDP_IPV6) {
hashen |= F_FW_RSS_VI_CONFIG_CMD_UDPEN |
F_FW_RSS_VI_CONFIG_CMD_IP6FOURTUPEN;
}
if (hashconfig & RSS_HASHTYPE_RSS_TCP_IPV4)
hashen |= F_FW_RSS_VI_CONFIG_CMD_IP4FOURTUPEN;
if (hashconfig & RSS_HASHTYPE_RSS_TCP_IPV6)
hashen |= F_FW_RSS_VI_CONFIG_CMD_IP6FOURTUPEN;
return (hashen);
}
/* Translates hardware hash types to kernel. */
static int
hashen_to_hashconfig(int hashen)
{
int hashconfig = 0;
if (hashen & F_FW_RSS_VI_CONFIG_CMD_UDPEN) {
/*
* If UDP hashing was enabled it must have been enabled for
* either IPv4 or IPv6 (inclusive or). Enabling UDP without
* enabling any 4-tuple hash is nonsense configuration.
*/
MPASS(hashen & (F_FW_RSS_VI_CONFIG_CMD_IP4FOURTUPEN |
F_FW_RSS_VI_CONFIG_CMD_IP6FOURTUPEN));
if (hashen & F_FW_RSS_VI_CONFIG_CMD_IP4FOURTUPEN)
hashconfig |= RSS_HASHTYPE_RSS_UDP_IPV4;
if (hashen & F_FW_RSS_VI_CONFIG_CMD_IP6FOURTUPEN)
hashconfig |= RSS_HASHTYPE_RSS_UDP_IPV6;
}
if (hashen & F_FW_RSS_VI_CONFIG_CMD_IP4FOURTUPEN)
hashconfig |= RSS_HASHTYPE_RSS_TCP_IPV4;
if (hashen & F_FW_RSS_VI_CONFIG_CMD_IP6FOURTUPEN)
hashconfig |= RSS_HASHTYPE_RSS_TCP_IPV6;
if (hashen & F_FW_RSS_VI_CONFIG_CMD_IP4TWOTUPEN)
hashconfig |= RSS_HASHTYPE_RSS_IPV4;
if (hashen & F_FW_RSS_VI_CONFIG_CMD_IP6TWOTUPEN)
hashconfig |= RSS_HASHTYPE_RSS_IPV6;
return (hashconfig);
}
#endif
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
int
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
vi_full_init(struct vi_info *vi)
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
{
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct adapter *sc = vi->pi->adapter;
struct ifnet *ifp = vi->ifp;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
uint16_t *rss;
struct sge_rxq *rxq;
cxgbe(4): Update the VI's default queue when netmap is enabled/disabled. Sponsored by: Chelsio Communications
2018-10-25 06:24:42 +00:00
int rc, i, j;
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
#ifdef RSS
int nbuckets = rss_getnumbuckets();
int hashconfig = rss_gethashconfig();
int extra;
#endif
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
ASSERT_SYNCHRONIZED_OP(sc);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
KASSERT((vi->flags & VI_INIT_DONE) == 0,
("%s: VI_INIT_DONE already", __func__));
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
sysctl_ctx_init(&vi->ctx);
vi->flags |= VI_SYSCTL_CTX;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
/*
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
* Allocate tx/rx/fl queues for this VI.
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
*/
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
rc = t4_setup_vi_queues(vi);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
if (rc != 0)
goto done; /* error message displayed already */
/*
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
* Setup RSS for this VI. Save a copy of the RSS table for later use.
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
*/
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (vi->nrxq > vi->rss_size) {
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
if_printf(ifp, "nrxq (%d) > hw RSS table size (%d); "
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
"some queues will never receive traffic.\n", vi->nrxq,
vi->rss_size);
} else if (vi->rss_size % vi->nrxq) {
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
if_printf(ifp, "nrxq (%d), hw RSS table size (%d); "
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
"expect uneven traffic distribution.\n", vi->nrxq,
vi->rss_size);
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
}
#ifdef RSS
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (vi->nrxq != nbuckets) {
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
if_printf(ifp, "nrxq (%d) != kernel RSS buckets (%d);"
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
"performance will be impacted.\n", vi->nrxq, nbuckets);
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
}
#endif
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
rss = malloc(vi->rss_size * sizeof (*rss), M_CXGBE, M_ZERO | M_WAITOK);
for (i = 0; i < vi->rss_size;) {
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
#ifdef RSS
j = rss_get_indirection_to_bucket(i);
Fix RSS build. Reported by: arybchik@
2015-12-05 10:10:18 +00:00
j %= vi->nrxq;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
rxq = &sc->sge.rxq[vi->first_rxq + j];
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
rss[i++] = rxq->iq.abs_id;
#else
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
for_each_rxq(vi, j, rxq) {
cxgbe(4): save a copy of the RSS map for each port for the driver's use.
2013-12-08 17:47:37 +00:00
rss[i++] = rxq->iq.abs_id;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (i == vi->rss_size)
cxgbe(4): save a copy of the RSS map for each port for the driver's use.
2013-12-08 17:47:37 +00:00
break;
}
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
#endif
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
cxgbe(4): save a copy of the RSS map for each port for the driver's use.
2013-12-08 17:47:37 +00:00
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
rc = -t4_config_rss_range(sc, sc->mbox, vi->viid, 0, vi->rss_size, rss,
vi->rss_size);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
if (rc != 0) {
cxgbe(4): Do not leak memory in case of errors during VI initialization. Reported by: Coverity (CID 1392026) MFC after: 1 week Sponsored by: Chelsio Communications
2018-08-21 22:15:57 +00:00
free(rss, M_CXGBE);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
if_printf(ifp, "rss_config failed: %d\n", rc);
goto done;
}
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
#ifdef RSS
cxgbe(4): Update the VI's default queue when netmap is enabled/disabled. Sponsored by: Chelsio Communications
2018-10-25 06:24:42 +00:00
vi->hashen = hashconfig_to_hashen(hashconfig);
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
/*
* We may have had to enable some hashes even though the global config
* wants them disabled. This is a potential problem that must be
* reported to the user.
*/
cxgbe(4): Update the VI's default queue when netmap is enabled/disabled. Sponsored by: Chelsio Communications
2018-10-25 06:24:42 +00:00
extra = hashen_to_hashconfig(vi->hashen) ^ hashconfig;
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
/*
* If we consider only the supported hash types, then the enabled hashes
* are a superset of the requested hashes. In other words, there cannot
* be any supported hash that was requested but not enabled, but there
* can be hashes that were not requested but had to be enabled.
*/
extra &= SUPPORTED_RSS_HASHTYPES;
MPASS((extra & hashconfig) == 0);
if (extra) {
if_printf(ifp,
sys/dev: minor spelling fixes. Most affect comments, very few have user-visible effects.
2016-05-03 03:41:25 +00:00
"global RSS config (0x%x) cannot be accommodated.\n",
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
hashconfig);
}
if (extra & RSS_HASHTYPE_RSS_IPV4)
if_printf(ifp, "IPv4 2-tuple hashing forced on.\n");
if (extra & RSS_HASHTYPE_RSS_TCP_IPV4)
if_printf(ifp, "TCP/IPv4 4-tuple hashing forced on.\n");
if (extra & RSS_HASHTYPE_RSS_IPV6)
if_printf(ifp, "IPv6 2-tuple hashing forced on.\n");
if (extra & RSS_HASHTYPE_RSS_TCP_IPV6)
if_printf(ifp, "TCP/IPv6 4-tuple hashing forced on.\n");
if (extra & RSS_HASHTYPE_RSS_UDP_IPV4)
if_printf(ifp, "UDP/IPv4 4-tuple hashing forced on.\n");
if (extra & RSS_HASHTYPE_RSS_UDP_IPV6)
if_printf(ifp, "UDP/IPv6 4-tuple hashing forced on.\n");
#else
cxgbe(4): Update the VI's default queue when netmap is enabled/disabled. Sponsored by: Chelsio Communications
2018-10-25 06:24:42 +00:00
vi->hashen = F_FW_RSS_VI_CONFIG_CMD_IP6FOURTUPEN |
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
F_FW_RSS_VI_CONFIG_CMD_IP6TWOTUPEN |
F_FW_RSS_VI_CONFIG_CMD_IP4FOURTUPEN |
F_FW_RSS_VI_CONFIG_CMD_IP4TWOTUPEN | F_FW_RSS_VI_CONFIG_CMD_UDPEN;
#endif
cxgbe(4): Update the VI's default queue when netmap is enabled/disabled. Sponsored by: Chelsio Communications
2018-10-25 06:24:42 +00:00
rc = -t4_config_vi_rss(sc, sc->mbox, vi->viid, vi->hashen, rss[0], 0, 0);
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
if (rc != 0) {
cxgbe(4): Do not leak memory in case of errors during VI initialization. Reported by: Coverity (CID 1392026) MFC after: 1 week Sponsored by: Chelsio Communications
2018-08-21 22:15:57 +00:00
free(rss, M_CXGBE);
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
if_printf(ifp, "rss hash/defaultq config failed: %d\n", rc);
goto done;
}
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
vi->rss = rss;
vi->flags |= VI_INIT_DONE;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
done:
if (rc != 0)
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
vi_full_uninit(vi);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
return (rc);
}
/*
* Idempotent.
*/
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
int
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
vi_full_uninit(struct vi_info *vi)
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
{
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct port_info *pi = vi->pi;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
struct adapter *sc = pi->adapter;
int i;
struct sge_rxq *rxq;
struct sge_txq *txq;
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
#ifdef TCP_OFFLOAD
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
struct sge_ofld_rxq *ofld_rxq;
Fix a few more places to handle ofld tx queues for RATELIMIT. - Drain offload transmit queues when RATELIMIT is enabled but TCP_OFFLOAD is not. - Expose the per-VI nofldtxq and first_ofld_txq sysctls when RATELIMIT is enabled but TCP_OFFLOAD is not. - Clear offload transmit queue stats as part of a 'cxgbetool clearstats' request when RATELIMIT is enabled but TCP_OFFLOAD is not. Reviewed by: np MFC after: 2 weeks Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18966
2019-01-25 20:54:18 +00:00
#endif
#if defined(TCP_OFFLOAD) || defined(RATELIMIT)
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
struct sge_wrq *ofld_txq;
#endif
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (vi->flags & VI_INIT_DONE) {
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): major tx rework. a) Front load as much work as possible in if_transmit, before any driver lock or software queue has to get involved. b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This is specifically for the tx multiqueue model where one of the if_transmit producer threads becomes the consumer and other producers carry on as usual. mp_ring is implemented as standalone code and it should be possible to use it in any driver with tx multiqueue. It also has: - the ability to enqueue/dequeue multiple items. This might become significant if packet batching is ever implemented. - an abdication mechanism to allow a thread to give up writing tx descriptors and have another if_transmit thread take over. A thread that's writing tx descriptors can end up doing so for an unbounded time period if a) there are other if_transmit threads continuously feeding the sofware queue, and b) the chip keeps up with whatever the thread is throwing at it. - accurate statistics about interesting events even when the stats come at the expense of additional branches/conditional code. The NIC txq lock is uncontested on the fast path at this point. I've left it there for synchronization with the control events (interface up/down, modload/unload). c) Add support for "type 1" coalescing work request in the normal NIC tx path. This work request is optimized for frames with a single item in the DMA gather list. These are very common when forwarding packets. Note that netmap tx in cxgbe already uses these "type 1" work requests. d) Do not request automatic cidx updates every 32 descriptors. Instead, request updates via bits in individual work requests (still every 32 descriptors approximately). Also, request an automatic final update when the queue idles after activity. This means NIC tx reclaim is still performed lazily but it will catch up quickly as soon as the queue idles. This seems to be the best middle ground and I'll probably do something similar for netmap tx as well. e) Implement a faster tx path for WRQs (used by TOE tx and control queues, _not_ by the normal NIC tx). Allow work requests to be written directly to the hardware descriptor ring if room is available. I will convert t4_tom and iw_cxgbe modules to this faster style gradually. MFC after: 2 months
2014-12-31 23:19:16 +00:00
/* Need to quiesce queues. */
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
/* XXX: Only for the first VI? */
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
if (IS_MAIN_VI(vi) && !(sc->flags & IS_VF))
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
quiesce_wrq(sc, &sc->sge.ctrlq[pi->port_id]);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
for_each_txq(vi, i, txq) {
cxgbe(4): major tx rework. a) Front load as much work as possible in if_transmit, before any driver lock or software queue has to get involved. b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This is specifically for the tx multiqueue model where one of the if_transmit producer threads becomes the consumer and other producers carry on as usual. mp_ring is implemented as standalone code and it should be possible to use it in any driver with tx multiqueue. It also has: - the ability to enqueue/dequeue multiple items. This might become significant if packet batching is ever implemented. - an abdication mechanism to allow a thread to give up writing tx descriptors and have another if_transmit thread take over. A thread that's writing tx descriptors can end up doing so for an unbounded time period if a) there are other if_transmit threads continuously feeding the sofware queue, and b) the chip keeps up with whatever the thread is throwing at it. - accurate statistics about interesting events even when the stats come at the expense of additional branches/conditional code. The NIC txq lock is uncontested on the fast path at this point. I've left it there for synchronization with the control events (interface up/down, modload/unload). c) Add support for "type 1" coalescing work request in the normal NIC tx path. This work request is optimized for frames with a single item in the DMA gather list. These are very common when forwarding packets. Note that netmap tx in cxgbe already uses these "type 1" work requests. d) Do not request automatic cidx updates every 32 descriptors. Instead, request updates via bits in individual work requests (still every 32 descriptors approximately). Also, request an automatic final update when the queue idles after activity. This means NIC tx reclaim is still performed lazily but it will catch up quickly as soon as the queue idles. This seems to be the best middle ground and I'll probably do something similar for netmap tx as well. e) Implement a faster tx path for WRQs (used by TOE tx and control queues, _not_ by the normal NIC tx). Allow work requests to be written directly to the hardware descriptor ring if room is available. I will convert t4_tom and iw_cxgbe modules to this faster style gradually. MFC after: 2 months
2014-12-31 23:19:16 +00:00
quiesce_txq(sc, txq);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
Fix a few more places to handle ofld tx queues for RATELIMIT. - Drain offload transmit queues when RATELIMIT is enabled but TCP_OFFLOAD is not. - Expose the per-VI nofldtxq and first_ofld_txq sysctls when RATELIMIT is enabled but TCP_OFFLOAD is not. - Clear offload transmit queue stats as part of a 'cxgbetool clearstats' request when RATELIMIT is enabled but TCP_OFFLOAD is not. Reviewed by: np MFC after: 2 weeks Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18966
2019-01-25 20:54:18 +00:00
#if defined(TCP_OFFLOAD) || defined(RATELIMIT)
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
for_each_ofld_txq(vi, i, ofld_txq) {
cxgbe(4): major tx rework. a) Front load as much work as possible in if_transmit, before any driver lock or software queue has to get involved. b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This is specifically for the tx multiqueue model where one of the if_transmit producer threads becomes the consumer and other producers carry on as usual. mp_ring is implemented as standalone code and it should be possible to use it in any driver with tx multiqueue. It also has: - the ability to enqueue/dequeue multiple items. This might become significant if packet batching is ever implemented. - an abdication mechanism to allow a thread to give up writing tx descriptors and have another if_transmit thread take over. A thread that's writing tx descriptors can end up doing so for an unbounded time period if a) there are other if_transmit threads continuously feeding the sofware queue, and b) the chip keeps up with whatever the thread is throwing at it. - accurate statistics about interesting events even when the stats come at the expense of additional branches/conditional code. The NIC txq lock is uncontested on the fast path at this point. I've left it there for synchronization with the control events (interface up/down, modload/unload). c) Add support for "type 1" coalescing work request in the normal NIC tx path. This work request is optimized for frames with a single item in the DMA gather list. These are very common when forwarding packets. Note that netmap tx in cxgbe already uses these "type 1" work requests. d) Do not request automatic cidx updates every 32 descriptors. Instead, request updates via bits in individual work requests (still every 32 descriptors approximately). Also, request an automatic final update when the queue idles after activity. This means NIC tx reclaim is still performed lazily but it will catch up quickly as soon as the queue idles. This seems to be the best middle ground and I'll probably do something similar for netmap tx as well. e) Implement a faster tx path for WRQs (used by TOE tx and control queues, _not_ by the normal NIC tx). Allow work requests to be written directly to the hardware descriptor ring if room is available. I will convert t4_tom and iw_cxgbe modules to this faster style gradually. MFC after: 2 months
2014-12-31 23:19:16 +00:00
quiesce_wrq(sc, ofld_txq);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
#endif
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
for_each_rxq(vi, i, rxq) {
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
quiesce_iq(sc, &rxq->iq);
quiesce_fl(sc, &rxq->fl);
}
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
#ifdef TCP_OFFLOAD
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
for_each_ofld_rxq(vi, i, ofld_rxq) {
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
quiesce_iq(sc, &ofld_rxq->iq);
quiesce_fl(sc, &ofld_rxq->fl);
}
#endif
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
free(vi->rss, M_CXGBE);
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
free(vi->nm_rss, M_CXGBE);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
t4_teardown_vi_queues(vi);
vi->flags &= ~VI_INIT_DONE;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
return (0);
}
static void
cxgbe(4): major tx rework. a) Front load as much work as possible in if_transmit, before any driver lock or software queue has to get involved. b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This is specifically for the tx multiqueue model where one of the if_transmit producer threads becomes the consumer and other producers carry on as usual. mp_ring is implemented as standalone code and it should be possible to use it in any driver with tx multiqueue. It also has: - the ability to enqueue/dequeue multiple items. This might become significant if packet batching is ever implemented. - an abdication mechanism to allow a thread to give up writing tx descriptors and have another if_transmit thread take over. A thread that's writing tx descriptors can end up doing so for an unbounded time period if a) there are other if_transmit threads continuously feeding the sofware queue, and b) the chip keeps up with whatever the thread is throwing at it. - accurate statistics about interesting events even when the stats come at the expense of additional branches/conditional code. The NIC txq lock is uncontested on the fast path at this point. I've left it there for synchronization with the control events (interface up/down, modload/unload). c) Add support for "type 1" coalescing work request in the normal NIC tx path. This work request is optimized for frames with a single item in the DMA gather list. These are very common when forwarding packets. Note that netmap tx in cxgbe already uses these "type 1" work requests. d) Do not request automatic cidx updates every 32 descriptors. Instead, request updates via bits in individual work requests (still every 32 descriptors approximately). Also, request an automatic final update when the queue idles after activity. This means NIC tx reclaim is still performed lazily but it will catch up quickly as soon as the queue idles. This seems to be the best middle ground and I'll probably do something similar for netmap tx as well. e) Implement a faster tx path for WRQs (used by TOE tx and control queues, _not_ by the normal NIC tx). Allow work requests to be written directly to the hardware descriptor ring if room is available. I will convert t4_tom and iw_cxgbe modules to this faster style gradually. MFC after: 2 months
2014-12-31 23:19:16 +00:00
quiesce_txq(struct adapter *sc, struct sge_txq *txq)
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
{
cxgbe(4): major tx rework. a) Front load as much work as possible in if_transmit, before any driver lock or software queue has to get involved. b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This is specifically for the tx multiqueue model where one of the if_transmit producer threads becomes the consumer and other producers carry on as usual. mp_ring is implemented as standalone code and it should be possible to use it in any driver with tx multiqueue. It also has: - the ability to enqueue/dequeue multiple items. This might become significant if packet batching is ever implemented. - an abdication mechanism to allow a thread to give up writing tx descriptors and have another if_transmit thread take over. A thread that's writing tx descriptors can end up doing so for an unbounded time period if a) there are other if_transmit threads continuously feeding the sofware queue, and b) the chip keeps up with whatever the thread is throwing at it. - accurate statistics about interesting events even when the stats come at the expense of additional branches/conditional code. The NIC txq lock is uncontested on the fast path at this point. I've left it there for synchronization with the control events (interface up/down, modload/unload). c) Add support for "type 1" coalescing work request in the normal NIC tx path. This work request is optimized for frames with a single item in the DMA gather list. These are very common when forwarding packets. Note that netmap tx in cxgbe already uses these "type 1" work requests. d) Do not request automatic cidx updates every 32 descriptors. Instead, request updates via bits in individual work requests (still every 32 descriptors approximately). Also, request an automatic final update when the queue idles after activity. This means NIC tx reclaim is still performed lazily but it will catch up quickly as soon as the queue idles. This seems to be the best middle ground and I'll probably do something similar for netmap tx as well. e) Implement a faster tx path for WRQs (used by TOE tx and control queues, _not_ by the normal NIC tx). Allow work requests to be written directly to the hardware descriptor ring if room is available. I will convert t4_tom and iw_cxgbe modules to this faster style gradually. MFC after: 2 months
2014-12-31 23:19:16 +00:00
struct sge_eq *eq = &txq->eq;
struct sge_qstat *spg = (void *)&eq->desc[eq->sidx];
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): major tx rework. a) Front load as much work as possible in if_transmit, before any driver lock or software queue has to get involved. b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This is specifically for the tx multiqueue model where one of the if_transmit producer threads becomes the consumer and other producers carry on as usual. mp_ring is implemented as standalone code and it should be possible to use it in any driver with tx multiqueue. It also has: - the ability to enqueue/dequeue multiple items. This might become significant if packet batching is ever implemented. - an abdication mechanism to allow a thread to give up writing tx descriptors and have another if_transmit thread take over. A thread that's writing tx descriptors can end up doing so for an unbounded time period if a) there are other if_transmit threads continuously feeding the sofware queue, and b) the chip keeps up with whatever the thread is throwing at it. - accurate statistics about interesting events even when the stats come at the expense of additional branches/conditional code. The NIC txq lock is uncontested on the fast path at this point. I've left it there for synchronization with the control events (interface up/down, modload/unload). c) Add support for "type 1" coalescing work request in the normal NIC tx path. This work request is optimized for frames with a single item in the DMA gather list. These are very common when forwarding packets. Note that netmap tx in cxgbe already uses these "type 1" work requests. d) Do not request automatic cidx updates every 32 descriptors. Instead, request updates via bits in individual work requests (still every 32 descriptors approximately). Also, request an automatic final update when the queue idles after activity. This means NIC tx reclaim is still performed lazily but it will catch up quickly as soon as the queue idles. This seems to be the best middle ground and I'll probably do something similar for netmap tx as well. e) Implement a faster tx path for WRQs (used by TOE tx and control queues, _not_ by the normal NIC tx). Allow work requests to be written directly to the hardware descriptor ring if room is available. I will convert t4_tom and iw_cxgbe modules to this faster style gradually. MFC after: 2 months
2014-12-31 23:19:16 +00:00
(void) sc; /* unused */
#ifdef INVARIANTS
TXQ_LOCK(txq);
MPASS((eq->flags & EQ_ENABLED) == 0);
TXQ_UNLOCK(txq);
#endif
/* Wait for the mp_ring to empty. */
while (!mp_ring_is_idle(txq->r)) {
mp_ring_check_drainage(txq->r, 0);
pause("rquiesce", 1);
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): major tx rework. a) Front load as much work as possible in if_transmit, before any driver lock or software queue has to get involved. b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This is specifically for the tx multiqueue model where one of the if_transmit producer threads becomes the consumer and other producers carry on as usual. mp_ring is implemented as standalone code and it should be possible to use it in any driver with tx multiqueue. It also has: - the ability to enqueue/dequeue multiple items. This might become significant if packet batching is ever implemented. - an abdication mechanism to allow a thread to give up writing tx descriptors and have another if_transmit thread take over. A thread that's writing tx descriptors can end up doing so for an unbounded time period if a) there are other if_transmit threads continuously feeding the sofware queue, and b) the chip keeps up with whatever the thread is throwing at it. - accurate statistics about interesting events even when the stats come at the expense of additional branches/conditional code. The NIC txq lock is uncontested on the fast path at this point. I've left it there for synchronization with the control events (interface up/down, modload/unload). c) Add support for "type 1" coalescing work request in the normal NIC tx path. This work request is optimized for frames with a single item in the DMA gather list. These are very common when forwarding packets. Note that netmap tx in cxgbe already uses these "type 1" work requests. d) Do not request automatic cidx updates every 32 descriptors. Instead, request updates via bits in individual work requests (still every 32 descriptors approximately). Also, request an automatic final update when the queue idles after activity. This means NIC tx reclaim is still performed lazily but it will catch up quickly as soon as the queue idles. This seems to be the best middle ground and I'll probably do something similar for netmap tx as well. e) Implement a faster tx path for WRQs (used by TOE tx and control queues, _not_ by the normal NIC tx). Allow work requests to be written directly to the hardware descriptor ring if room is available. I will convert t4_tom and iw_cxgbe modules to this faster style gradually. MFC after: 2 months
2014-12-31 23:19:16 +00:00
/* Then wait for the hardware to finish. */
while (spg->cidx != htobe16(eq->pidx))
pause("equiesce", 1);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): major tx rework. a) Front load as much work as possible in if_transmit, before any driver lock or software queue has to get involved. b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This is specifically for the tx multiqueue model where one of the if_transmit producer threads becomes the consumer and other producers carry on as usual. mp_ring is implemented as standalone code and it should be possible to use it in any driver with tx multiqueue. It also has: - the ability to enqueue/dequeue multiple items. This might become significant if packet batching is ever implemented. - an abdication mechanism to allow a thread to give up writing tx descriptors and have another if_transmit thread take over. A thread that's writing tx descriptors can end up doing so for an unbounded time period if a) there are other if_transmit threads continuously feeding the sofware queue, and b) the chip keeps up with whatever the thread is throwing at it. - accurate statistics about interesting events even when the stats come at the expense of additional branches/conditional code. The NIC txq lock is uncontested on the fast path at this point. I've left it there for synchronization with the control events (interface up/down, modload/unload). c) Add support for "type 1" coalescing work request in the normal NIC tx path. This work request is optimized for frames with a single item in the DMA gather list. These are very common when forwarding packets. Note that netmap tx in cxgbe already uses these "type 1" work requests. d) Do not request automatic cidx updates every 32 descriptors. Instead, request updates via bits in individual work requests (still every 32 descriptors approximately). Also, request an automatic final update when the queue idles after activity. This means NIC tx reclaim is still performed lazily but it will catch up quickly as soon as the queue idles. This seems to be the best middle ground and I'll probably do something similar for netmap tx as well. e) Implement a faster tx path for WRQs (used by TOE tx and control queues, _not_ by the normal NIC tx). Allow work requests to be written directly to the hardware descriptor ring if room is available. I will convert t4_tom and iw_cxgbe modules to this faster style gradually. MFC after: 2 months
2014-12-31 23:19:16 +00:00
/* Finally, wait for the driver to reclaim all descriptors. */
while (eq->cidx != eq->pidx)
pause("dquiesce", 1);
}
static void
quiesce_wrq(struct adapter *sc, struct sge_wrq *wrq)
{
/* XXXTX */
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
static void
quiesce_iq(struct adapter *sc, struct sge_iq *iq)
{
(void) sc; /* unused */
/* Synchronize with the interrupt handler */
while (!atomic_cmpset_int(&iq->state, IQS_IDLE, IQS_DISABLED))
pause("iqfree", 1);
}
static void
quiesce_fl(struct adapter *sc, struct sge_fl *fl)
{
mtx_lock(&sc->sfl_lock);
FL_LOCK(fl);
fl->flags |= FL_DOOMED;
FL_UNLOCK(fl);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
callout_stop(&sc->sfl_callout);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
mtx_unlock(&sc->sfl_lock);
KASSERT((fl->flags & FL_STARVING) == 0,
("%s: still starving", __func__));
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
static int
t4_alloc_irq(struct adapter *sc, struct irq *irq, int rid,
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
driver_intr_t *handler, void *arg, char *name)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
{
int rc;
irq->rid = rid;
irq->res = bus_alloc_resource_any(sc->dev, SYS_RES_IRQ, &irq->rid,
RF_SHAREABLE | RF_ACTIVE);
if (irq->res == NULL) {
device_printf(sc->dev,
"failed to allocate IRQ for rid %d, name %s.\n", rid, name);
return (ENOMEM);
}
rc = bus_setup_intr(sc->dev, irq->res, INTR_MPSAFE | INTR_TYPE_NET,
NULL, handler, arg, &irq->tag);
if (rc != 0) {
device_printf(sc->dev,
"failed to setup interrupt for rid %d, name %s: %d\n",
rid, name, rc);
} else if (name)
Add __printflike() to bus_describe_intr() to enable -Wformat checks. Fix a few places that were passing a raw string as the format to use a "%s" format string instead. MFC after: 2 months
2016-08-04 18:29:16 +00:00
bus_describe_intr(sc->dev, irq->res, irq->tag, "%s", name);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
return (rc);
}
static int
t4_free_irq(struct adapter *sc, struct irq *irq)
{
if (irq->tag)
bus_teardown_intr(sc->dev, irq->res, irq->tag);
if (irq->res)
bus_release_resource(sc->dev, SYS_RES_IRQ, irq->rid, irq->res);
bzero(irq, sizeof(*irq));
return (0);
}
static void
cxgbe(4): Updated register dumps. - Get the list of registers to read during a regdump from the shared code instead of the OS specific code. This follows a similar move internally. The shared code includes the list for T6. - Update cxgbetool to be able to decode T5 VF, T6, and T6 VF register dumps (and catch up with some updates to T4 and T5 register decode). Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-07 21:11:35 +00:00
get_regs(struct adapter *sc, struct t4_regdump *regs, uint8_t *buf)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
{
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
regs->version = chip_id(sc) | chip_rev(sc) << 10;
cxgbe(4): Updated register dumps. - Get the list of registers to read during a regdump from the shared code instead of the OS specific code. This follows a similar move internally. The shared code includes the list for T6. - Update cxgbetool to be able to decode T5 VF, T6, and T6 VF register dumps (and catch up with some updates to T4 and T5 register decode). Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-07 21:11:35 +00:00
t4_get_regs(sc, buf, regs->len);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
#define A_PL_INDIR_CMD 0x1f8
#define S_PL_AUTOINC 31
#define M_PL_AUTOINC 0x1U
#define V_PL_AUTOINC(x) ((x) << S_PL_AUTOINC)
#define G_PL_AUTOINC(x) (((x) >> S_PL_AUTOINC) & M_PL_AUTOINC)
#define S_PL_VFID 20
#define M_PL_VFID 0xffU
#define V_PL_VFID(x) ((x) << S_PL_VFID)
#define G_PL_VFID(x) (((x) >> S_PL_VFID) & M_PL_VFID)
#define S_PL_ADDR 0
#define M_PL_ADDR 0xfffffU
#define V_PL_ADDR(x) ((x) << S_PL_ADDR)
#define G_PL_ADDR(x) (((x) >> S_PL_ADDR) & M_PL_ADDR)
#define A_PL_INDIR_DATA 0x1fc
static uint64_t
cxgbe(4): Treat the viid as an opaque identifier. Recent firmwares prefer to use a different format for viid internally and this change allows them to do so. MFC after: 1 week Sponsored by: Chelsio Communications
2019-03-20 17:27:11 +00:00
read_vf_stat(struct adapter *sc, u_int vin, int reg)
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
{
u32 stats[2];
cxgbe(4): Rename regwin_lock to reg_lock. It is used to protect access to indirect registers only.
2016-03-08 22:23:30 +00:00
mtx_assert(&sc->reg_lock, MA_OWNED);
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
if (sc->flags & IS_VF) {
stats[0] = t4_read_reg(sc, VF_MPS_REG(reg));
stats[1] = t4_read_reg(sc, VF_MPS_REG(reg + 4));
} else {
t4_write_reg(sc, A_PL_INDIR_CMD, V_PL_AUTOINC(1) |
cxgbe(4): Treat the viid as an opaque identifier. Recent firmwares prefer to use a different format for viid internally and this change allows them to do so. MFC after: 1 week Sponsored by: Chelsio Communications
2019-03-20 17:27:11 +00:00
V_PL_VFID(vin) | V_PL_ADDR(VF_MPS_REG(reg)));
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
stats[0] = t4_read_reg(sc, A_PL_INDIR_DATA);
stats[1] = t4_read_reg(sc, A_PL_INDIR_DATA);
}
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
return (((uint64_t)stats[1]) << 32 | stats[0]);
}
static void
cxgbe(4): Treat the viid as an opaque identifier. Recent firmwares prefer to use a different format for viid internally and this change allows them to do so. MFC after: 1 week Sponsored by: Chelsio Communications
2019-03-20 17:27:11 +00:00
t4_get_vi_stats(struct adapter *sc, u_int vin, struct fw_vi_stats_vf *stats)
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
{
#define GET_STAT(name) \
cxgbe(4): Treat the viid as an opaque identifier. Recent firmwares prefer to use a different format for viid internally and this change allows them to do so. MFC after: 1 week Sponsored by: Chelsio Communications
2019-03-20 17:27:11 +00:00
read_vf_stat(sc, vin, A_MPS_VF_STAT_##name##_L)
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
stats->tx_bcast_bytes = GET_STAT(TX_VF_BCAST_BYTES);
stats->tx_bcast_frames = GET_STAT(TX_VF_BCAST_FRAMES);
stats->tx_mcast_bytes = GET_STAT(TX_VF_MCAST_BYTES);
stats->tx_mcast_frames = GET_STAT(TX_VF_MCAST_FRAMES);
stats->tx_ucast_bytes = GET_STAT(TX_VF_UCAST_BYTES);
stats->tx_ucast_frames = GET_STAT(TX_VF_UCAST_FRAMES);
stats->tx_drop_frames = GET_STAT(TX_VF_DROP_FRAMES);
stats->tx_offload_bytes = GET_STAT(TX_VF_OFFLOAD_BYTES);
stats->tx_offload_frames = GET_STAT(TX_VF_OFFLOAD_FRAMES);
stats->rx_bcast_bytes = GET_STAT(RX_VF_BCAST_BYTES);
stats->rx_bcast_frames = GET_STAT(RX_VF_BCAST_FRAMES);
stats->rx_mcast_bytes = GET_STAT(RX_VF_MCAST_BYTES);
stats->rx_mcast_frames = GET_STAT(RX_VF_MCAST_FRAMES);
stats->rx_ucast_bytes = GET_STAT(RX_VF_UCAST_BYTES);
stats->rx_ucast_frames = GET_STAT(RX_VF_UCAST_FRAMES);
stats->rx_err_frames = GET_STAT(RX_VF_ERR_FRAMES);
#undef GET_STAT
}
static void
cxgbe(4): Treat the viid as an opaque identifier. Recent firmwares prefer to use a different format for viid internally and this change allows them to do so. MFC after: 1 week Sponsored by: Chelsio Communications
2019-03-20 17:27:11 +00:00
t4_clr_vi_stats(struct adapter *sc, u_int vin)
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
{
int reg;
cxgbe(4): Treat the viid as an opaque identifier. Recent firmwares prefer to use a different format for viid internally and this change allows them to do so. MFC after: 1 week Sponsored by: Chelsio Communications
2019-03-20 17:27:11 +00:00
t4_write_reg(sc, A_PL_INDIR_CMD, V_PL_AUTOINC(1) | V_PL_VFID(vin) |
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
V_PL_ADDR(VF_MPS_REG(A_MPS_VF_STAT_TX_VF_BCAST_BYTES_L)));
for (reg = A_MPS_VF_STAT_TX_VF_BCAST_BYTES_L;
reg <= A_MPS_VF_STAT_RX_VF_ERR_FRAMES_H; reg += 4)
t4_write_reg(sc, A_PL_INDIR_DATA, 0);
}
static void
vi_refresh_stats(struct adapter *sc, struct vi_info *vi)
{
struct timeval tv;
const struct timeval interval = {0, 250000}; /* 250ms */
if (!(vi->flags & VI_INIT_DONE))
return;
getmicrotime(&tv);
timevalsub(&tv, &interval);
if (timevalcmp(&tv, &vi->last_refreshed, <))
return;
cxgbe(4): Rename regwin_lock to reg_lock. It is used to protect access to indirect registers only.
2016-03-08 22:23:30 +00:00
mtx_lock(&sc->reg_lock);
cxgbe(4): Treat the viid as an opaque identifier. Recent firmwares prefer to use a different format for viid internally and this change allows them to do so. MFC after: 1 week Sponsored by: Chelsio Communications
2019-03-20 17:27:11 +00:00
t4_get_vi_stats(sc, vi->vin, &vi->stats);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
getmicrotime(&vi->last_refreshed);
cxgbe(4): Rename regwin_lock to reg_lock. It is used to protect access to indirect registers only.
2016-03-08 22:23:30 +00:00
mtx_unlock(&sc->reg_lock);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
}
cxgbe(4): implement if_get_counter.
2014-09-27 05:50:31 +00:00
static void
cxgbe_refresh_stats(struct adapter *sc, struct port_info *pi)
{
cxgbe(4): Read the MPS buffer group map from the firmware as it could be different from hardware defaults. The congestion channel map, which is still fixed, needs to be tracked separately now. Change the congestion setting for TOE rx queues to match the drivers on other OSes while here. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-10-24 05:41:48 +00:00
u_int i, v, tnl_cong_drops, bg_map;
cxgbe(4): implement if_get_counter.
2014-09-27 05:50:31 +00:00
struct timeval tv;
const struct timeval interval = {0, 250000}; /* 250ms */
getmicrotime(&tv);
timevalsub(&tv, &interval);
if (timevalcmp(&tv, &pi->last_refreshed, <))
return;
tnl_cong_drops = 0;
t4_get_port_stats(sc, pi->tx_chan, &pi->stats);
cxgbe(4): Read the MPS buffer group map from the firmware as it could be different from hardware defaults. The congestion channel map, which is still fixed, needs to be tracked separately now. Change the congestion setting for TOE rx queues to match the drivers on other OSes while here. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-10-24 05:41:48 +00:00
bg_map = pi->mps_bg_map;
while (bg_map) {
i = ffs(bg_map) - 1;
mtx_lock(&sc->reg_lock);
t4_read_indirect(sc, A_TP_MIB_INDEX, A_TP_MIB_DATA, &v, 1,
A_TP_MIB_TNL_CNG_DROP_0 + i);
mtx_unlock(&sc->reg_lock);
tnl_cong_drops += v;
bg_map &= ~(1 << i);
cxgbe(4): implement if_get_counter.
2014-09-27 05:50:31 +00:00
}
pi->tnl_cong_drops = tnl_cong_drops;
getmicrotime(&pi->last_refreshed);
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
static void
cxgbe_tick(void *arg)
{
struct port_info *pi = arg;
cxgbe(4): if_iqdrops statistic should include tunnel congestion drops. MFC after: 1 week
2014-03-20 01:58:04 +00:00
struct adapter *sc = pi->adapter;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
PORT_LOCK_ASSERT_OWNED(pi);
cxgbe(4): implement if_get_counter.
2014-09-27 05:50:31 +00:00
cxgbe_refresh_stats(sc, pi);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
callout_schedule(&pi->tick, hz);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
}
void
vi_tick(void *arg)
{
struct vi_info *vi = arg;
struct adapter *sc = vi->pi->adapter;
vi_refresh_stats(sc, vi);
callout_schedule(&vi->tick, hz);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
cxgbe(4): Catch up with the latest list of card capabilities as reported by the firmware.
2016-03-12 02:54:55 +00:00
/*
* Should match fw_caps_config_<foo> enums in t4fw_interface.h
*/
static char *caps_decoder[] = {
cxgbe(4): Fix typo in previous commit.
2016-03-12 03:02:33 +00:00
"\20\001IPMI\002NCSI", /* 0: NBM */
cxgbe(4): Catch up with the latest list of card capabilities as reported by the firmware.
2016-03-12 02:54:55 +00:00
"\20\001PPP\002QFC\003DCBX", /* 1: link */
"\20\001INGRESS\002EGRESS", /* 2: switch */
"\20\001NIC\002VM\003IDS\004UM\005UM_ISGL" /* 3: NIC */
"\006HASHFILTER\007ETHOFLD",
"\20\001TOE", /* 4: TOE */
"\20\001RDDP\002RDMAC", /* 5: RDMA */
"\20\001INITIATOR_PDU\002TARGET_PDU" /* 6: iSCSI */
"\003INITIATOR_CNXOFLD\004TARGET_CNXOFLD"
"\005INITIATOR_SSNOFLD\006TARGET_SSNOFLD"
"\007T10DIF"
"\010INITIATOR_CMDOFLD\011TARGET_CMDOFLD",
cxgbe(4): Catch up with the rename of tlscaps -> cryptocaps. TLS is one of the capabilities of the crypto engine in T6. Sponsored by: Chelsio Communications
2016-09-12 00:15:40 +00:00
"\20\001LOOKASIDE\002TLSKEYS", /* 7: Crypto */
cxgbe(4): Catch up with the latest list of card capabilities as reported by the firmware.
2016-03-12 02:54:55 +00:00
"\20\001INITIATOR\002TARGET\003CTRL_OFLD" /* 8: FCoE */
"\004PO_INITIATOR\005PO_TARGET",
};
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
void
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
t4_sysctls(struct adapter *sc)
{
struct sysctl_ctx_list *ctx;
struct sysctl_oid *oid;
struct sysctl_oid_list *children, *c0;
Cosmetic change (s/wrwc/wcwr/;s/WRWC/WCWR/). MFC after: 3 days.
2013-04-11 22:49:29 +00:00
static char *doorbells = {"\20\1UDB\2WCWR\3UDBWC\4KDB"};
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
ctx = device_get_sysctl_ctx(sc->dev);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
/*
* dev.t4nex.X.
*/
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
oid = device_get_sysctl_tree(sc->dev);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
c0 = children = SYSCTL_CHILDREN(oid);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Add an option to enable or disable the small RX packet copying that is done to improve performance of small frames. When doing RX packing, the RX copying isn't necessarily required. Reviewed by: np
2014-01-02 23:23:33 +00:00
sc->sc_do_rxcopy = 1;
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "do_rx_copy", CTLFLAG_RW,
&sc->sc_do_rxcopy, 1, "Do RX copy of small frames");
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "nports", CTLFLAG_RD, NULL,
sc->params.nports, "# of ports");
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Reorder sysctls so that nodes shared with the VF driver are added first. This permits a single early return for VF devices in the routines that add sysctl nodes. Reviewed by: np Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7512
2016-08-19 17:54:51 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "doorbells",
cxgbe(4): Break up sysctl_bitfield into 8 bit and 16 bit variants. Have them display the current value of the bitfield rather than the fixed value that was provided when the sysctl node was created. MFC after: 1 week Sponsored by: Chelsio Communications
2018-08-06 21:54:51 +00:00
CTLTYPE_STRING | CTLFLAG_RD, doorbells, (uintptr_t)&sc->doorbells,
sysctl_bitfield_8b, "A", "available doorbells");
Reorder sysctls so that nodes shared with the VF driver are added first. This permits a single early return for VF devices in the routines that add sysctl nodes. Reviewed by: np Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7512
2016-08-19 17:54:51 +00:00
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "core_clock", CTLFLAG_RD, NULL,
sc->params.vpd.cclk, "core clock frequency (in KHz)");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "holdoff_timers",
CTLTYPE_STRING | CTLFLAG_RD, sc->params.sge.timer_val,
sizeof(sc->params.sge.timer_val), sysctl_int_array, "A",
"interrupt holdoff timer values (us)");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "holdoff_pkt_counts",
CTLTYPE_STRING | CTLFLAG_RD, sc->params.sge.counter_val,
sizeof(sc->params.sge.counter_val), sysctl_int_array, "A",
"interrupt holdoff packet counter values");
t4_sge_sysctls(sc, ctx, children);
sc->lro_timeout = 100;
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "lro_timeout", CTLFLAG_RW,
&sc->lro_timeout, 0, "lro inactive-flush timeout (in us)");
cxgbe(4): Rename the debug_flags driver tunable/sysctl to dflags. Tunables that end with _flags are special. Sponsored by: Chelsio Communications
2016-09-11 18:05:37 +00:00
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "dflags", CTLFLAG_RW,
Reorder sysctls so that nodes shared with the VF driver are added first. This permits a single early return for VF devices in the routines that add sysctl nodes. Reviewed by: np Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7512
2016-08-19 17:54:51 +00:00
&sc->debug_flags, 0, "flags to enable runtime debugging");
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
SYSCTL_ADD_STRING(ctx, children, OID_AUTO, "tp_version",
CTLFLAG_RD, sc->tp_version, 0, "TP microcode version");
SYSCTL_ADD_STRING(ctx, children, OID_AUTO, "firmware_version",
CTLFLAG_RD, sc->fw_version, 0, "firmware version");
if (sc->flags & IS_VF)
return;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "hw_revision", CTLFLAG_RD,
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
NULL, chip_rev(sc), "chip hardware revision");
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Provide more details about the card in the sysctl MIB. dev.t5nex.0.%desc: Chelsio T580-CR dev.t5nex.0.hw_revision: 1 dev.t5nex.0.sn: PT13140042 dev.t5nex.0.pn: 110117150A0 dev.t5nex.0.ec: 0000000000000000 dev.t5nex.0.na: 0007432AF490 dev.t5nex.0.vpd_version: 3 dev.t5nex.0.scfg_version: 53255 dev.t5nex.0.bs_version: 1.1.0.0 dev.t5nex.0.er_version: 1.0.0.68 dev.t5nex.0.tp_version: 0.1.4.9 dev.t5nex.0.firmware_version: 1.16.2.0 Sponsored by: Chelsio Communications
2016-08-27 00:13:41 +00:00
SYSCTL_ADD_STRING(ctx, children, OID_AUTO, "sn",
CTLFLAG_RD, sc->params.vpd.sn, 0, "serial number");
SYSCTL_ADD_STRING(ctx, children, OID_AUTO, "pn",
CTLFLAG_RD, sc->params.vpd.pn, 0, "part number");
SYSCTL_ADD_STRING(ctx, children, OID_AUTO, "ec",
CTLFLAG_RD, sc->params.vpd.ec, 0, "engineering change");
cxgbe(4): Read the MFG diags version from the VPD and make it available in the sysctl MIB. MFC after: 1 week Sponsored by: Chelsio Communications
2017-12-21 15:19:43 +00:00
SYSCTL_ADD_STRING(ctx, children, OID_AUTO, "md_version",
CTLFLAG_RD, sc->params.vpd.md, 0, "manufacturing diags version");
cxgbe(4): Provide more details about the card in the sysctl MIB. dev.t5nex.0.%desc: Chelsio T580-CR dev.t5nex.0.hw_revision: 1 dev.t5nex.0.sn: PT13140042 dev.t5nex.0.pn: 110117150A0 dev.t5nex.0.ec: 0000000000000000 dev.t5nex.0.na: 0007432AF490 dev.t5nex.0.vpd_version: 3 dev.t5nex.0.scfg_version: 53255 dev.t5nex.0.bs_version: 1.1.0.0 dev.t5nex.0.er_version: 1.0.0.68 dev.t5nex.0.tp_version: 0.1.4.9 dev.t5nex.0.firmware_version: 1.16.2.0 Sponsored by: Chelsio Communications
2016-08-27 00:13:41 +00:00
SYSCTL_ADD_STRING(ctx, children, OID_AUTO, "na",
CTLFLAG_RD, sc->params.vpd.na, 0, "network address");
SYSCTL_ADD_STRING(ctx, children, OID_AUTO, "er_version", CTLFLAG_RD,
sc->er_version, 0, "expansion ROM version");
cxgbe(4): Add sysctls to display the TP microcode version and the expansion rom version (if there's one). trantor:~# sysctl dev.t4nex dev.t5nex | grep _version dev.t4nex.0.firmware_version: 1.15.28.0 dev.t4nex.0.tp_version: 0.1.9.4 dev.t5nex.0.firmware_version: 1.15.28.0 dev.t5nex.0.exprom_version: 1.0.0.68 dev.t5nex.0.tp_version: 0.1.4.9
2016-03-11 03:15:17 +00:00
cxgbe(4): Provide more details about the card in the sysctl MIB. dev.t5nex.0.%desc: Chelsio T580-CR dev.t5nex.0.hw_revision: 1 dev.t5nex.0.sn: PT13140042 dev.t5nex.0.pn: 110117150A0 dev.t5nex.0.ec: 0000000000000000 dev.t5nex.0.na: 0007432AF490 dev.t5nex.0.vpd_version: 3 dev.t5nex.0.scfg_version: 53255 dev.t5nex.0.bs_version: 1.1.0.0 dev.t5nex.0.er_version: 1.0.0.68 dev.t5nex.0.tp_version: 0.1.4.9 dev.t5nex.0.firmware_version: 1.16.2.0 Sponsored by: Chelsio Communications
2016-08-27 00:13:41 +00:00
SYSCTL_ADD_STRING(ctx, children, OID_AUTO, "bs_version", CTLFLAG_RD,
sc->bs_version, 0, "bootstrap firmware version");
SYSCTL_ADD_UINT(ctx, children, OID_AUTO, "scfg_version", CTLFLAG_RD,
NULL, sc->params.scfg_vers, "serial config version");
SYSCTL_ADD_UINT(ctx, children, OID_AUTO, "vpd_version", CTLFLAG_RD,
NULL, sc->params.vpd_vers, "VPD version");
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
SYSCTL_ADD_STRING(ctx, children, OID_AUTO, "cf",
Fix multiple incorrect SYSCTL arguments in the kernel: - Wrong integer type was specified. - Wrong or missing "access" specifier. The "access" specifier sometimes included the SYSCTL type, which it should not, except for procedural SYSCTL nodes. - Logical OR where binary OR was expected. - Properly assert the "access" argument passed to all SYSCTL macros, using the CTASSERT macro. This applies to both static- and dynamically created SYSCTLs. - Properly assert the the data type for both static and dynamic SYSCTLs. In the case of static SYSCTLs we only assert that the data pointed to by the SYSCTL data pointer has the correct size, hence there is no easy way to assert types in the C language outside a C-function. - Rewrote some code which doesn't pass a constant "access" specifier when creating dynamic SYSCTL nodes, which is now a requirement. - Updated "EXAMPLES" section in SYSCTL manual page. MFC after: 3 days Sponsored by: Mellanox Technologies
2014-10-21 07:31:21 +00:00
CTLFLAG_RD, sc->cfg_file, 0, "configuration file");
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
SYSCTL_ADD_UINT(ctx, children, OID_AUTO, "cfcsum", CTLFLAG_RD, NULL,
sc->cfcsum, "config file checksum");
cxgbe(4): Catch up with the latest list of card capabilities as reported by the firmware.
2016-03-12 02:54:55 +00:00
#define SYSCTL_CAP(name, n, text) \
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, #name, \
cxgbe(4): Break up sysctl_bitfield into 8 bit and 16 bit variants. Have them display the current value of the bitfield rather than the fixed value that was provided when the sysctl node was created. MFC after: 1 week Sponsored by: Chelsio Communications
2018-08-06 21:54:51 +00:00
CTLTYPE_STRING | CTLFLAG_RD, caps_decoder[n], (uintptr_t)&sc->name, \
sysctl_bitfield_16b, "A", "available " text " capabilities")
cxgbe(4): Catch up with the latest list of card capabilities as reported by the firmware.
2016-03-12 02:54:55 +00:00
SYSCTL_CAP(nbmcaps, 0, "NBM");
SYSCTL_CAP(linkcaps, 1, "link");
SYSCTL_CAP(switchcaps, 2, "switch");
SYSCTL_CAP(niccaps, 3, "NIC");
SYSCTL_CAP(toecaps, 4, "TCP offload");
SYSCTL_CAP(rdmacaps, 5, "RDMA");
SYSCTL_CAP(iscsicaps, 6, "iSCSI");
cxgbe(4): Catch up with the rename of tlscaps -> cryptocaps. TLS is one of the capabilities of the crypto engine in T6. Sponsored by: Chelsio Communications
2016-09-12 00:15:40 +00:00
SYSCTL_CAP(cryptocaps, 7, "crypto");
cxgbe(4): Catch up with the latest list of card capabilities as reported by the firmware.
2016-03-12 02:54:55 +00:00
SYSCTL_CAP(fcoecaps, 8, "FCoE");
#undef SYSCTL_CAP
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Add a sysctl to get the number of filters available. sysctl dev.t4nex.<N>.nfilters sysctl dev.t5nex.<N>.nfilters MFC after: 3 days
2013-07-01 17:31:04 +00:00
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "nfilters", CTLFLAG_RD,
NULL, sc->tids.nftids, "number of filters");
Display temperature sensor data. Shows -1 if sensor not available on the card. # sysctl dev.t4nex.0.temperature # sysctl dev.t5nex.0.temperature
2013-08-02 18:05:42 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "temperature", CTLTYPE_INT |
use correct (integer) type for the temperature sysctl Reviewed by: np, scottl Obtained from: Netflix MFC after: 3 days
2014-04-17 19:29:15 +00:00
CTLFLAG_RD, sc, 0, sysctl_temperature, "I",
Display temperature sensor data. Shows -1 if sensor not available on the card. # sysctl dev.t4nex.0.temperature # sysctl dev.t5nex.0.temperature
2013-08-02 18:05:42 +00:00
"chip temperature (in Celsius)");
cxgbe(4): Add a sysctl to report the chip's microprocessor's load averages. This works with debug or custom firmwares only. sysctl dev.<nexus>.<instance>.loadavg sysctl dev.t6nex.0.loadavg MFC after: 1 month Sponsored by: Chelsio Communications
2018-07-10 03:03:10 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "loadavg", CTLTYPE_STRING |
CTLFLAG_RD, sc, 0, sysctl_loadavg, "A",
"microprocessor load averages (debug firmwares only)");
cxgbe(4): Add core Vdd to the sysctl MIB. Sponsored by: Chelsio Communications
2017-11-17 23:22:39 +00:00
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "core_vdd", CTLFLAG_RD,
&sc->params.core_vdd, 0, "core Vdd (in mV)");
cxgbe(4): sysctls to display the local and intr CPUs for the adapter. The driver assumes the list can change (even though it does't right now) and queries it every time the sysctl runs. sysctl dev.<nexus>.<inst>.local_cpus sysctl dev.<nexus>.<inst>.intr_cpus sysctl dev.t6nex.0.local_cpus sysctl dev.t6nex.0.intr_cpus Sponsored by: Chelsio Communications
2018-06-15 18:04:44 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "local_cpus",
CTLTYPE_STRING | CTLFLAG_RD, sc, LOCAL_CPUS,
sysctl_cpus, "A", "local CPUs");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "intr_cpus",
CTLTYPE_STRING | CTLFLAG_RD, sc, INTR_CPUS,
sysctl_cpus, "A", "preferred CPUs for interrupts");
cxgbe(4): Count and clear interrupts generated at the software's request. An interrupt can be requested by setting the F_SWINT bit in PL_PF_CTL. MFC after: 1 week Sponsored by: Chelsio Communications
2019-03-28 21:22:28 +00:00
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "swintr", CTLFLAG_RW,
&sc->swintr, 0, "software triggered interrupts");
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
/*
* dev.t4nex.X.misc. Marked CTLFLAG_SKIP to avoid information overload.
*/
oid = SYSCTL_ADD_NODE(ctx, c0, OID_AUTO, "misc",
CTLFLAG_RD | CTLFLAG_SKIP, NULL,
"logs and miscellaneous information");
children = SYSCTL_CHILDREN(oid);
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cctrl",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
sysctl_cctrl, "A", "congestion control");
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_ibq_tp0",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
sysctl_cim_ibq_obq, "A", "CIM IBQ 0 (TP0)");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_ibq_tp1",
CTLTYPE_STRING | CTLFLAG_RD, sc, 1,
sysctl_cim_ibq_obq, "A", "CIM IBQ 1 (TP1)");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_ibq_ulp",
CTLTYPE_STRING | CTLFLAG_RD, sc, 2,
sysctl_cim_ibq_obq, "A", "CIM IBQ 2 (ULP)");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_ibq_sge0",
CTLTYPE_STRING | CTLFLAG_RD, sc, 3,
sysctl_cim_ibq_obq, "A", "CIM IBQ 3 (SGE0)");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_ibq_sge1",
CTLTYPE_STRING | CTLFLAG_RD, sc, 4,
sysctl_cim_ibq_obq, "A", "CIM IBQ 4 (SGE1)");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_ibq_ncsi",
CTLTYPE_STRING | CTLFLAG_RD, sc, 5,
sysctl_cim_ibq_obq, "A", "CIM IBQ 5 (NCSI)");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_la",
cxgbe(4): Auto-dump the CIM block's logic analyzer on a TIMER0 interrupt. Sponsored by: Chelsio Communications
2019-02-07 05:40:51 +00:00
CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_cim_la,
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
"A", "CIM logic analyzer");
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_ma_la",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
sysctl_cim_ma_la, "A", "CIM MA logic analyzer");
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_obq_ulp0",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0 + CIM_NUM_IBQ,
sysctl_cim_ibq_obq, "A", "CIM OBQ 0 (ULP0)");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_obq_ulp1",
CTLTYPE_STRING | CTLFLAG_RD, sc, 1 + CIM_NUM_IBQ,
sysctl_cim_ibq_obq, "A", "CIM OBQ 1 (ULP1)");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_obq_ulp2",
CTLTYPE_STRING | CTLFLAG_RD, sc, 2 + CIM_NUM_IBQ,
sysctl_cim_ibq_obq, "A", "CIM OBQ 2 (ULP2)");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_obq_ulp3",
CTLTYPE_STRING | CTLFLAG_RD, sc, 3 + CIM_NUM_IBQ,
sysctl_cim_ibq_obq, "A", "CIM OBQ 3 (ULP3)");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_obq_sge",
CTLTYPE_STRING | CTLFLAG_RD, sc, 4 + CIM_NUM_IBQ,
sysctl_cim_ibq_obq, "A", "CIM OBQ 4 (SGE)");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_obq_ncsi",
CTLTYPE_STRING | CTLFLAG_RD, sc, 5 + CIM_NUM_IBQ,
sysctl_cim_ibq_obq, "A", "CIM OBQ 5 (NCSI)");
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
if (chip_id(sc) > CHELSIO_T4) {
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_obq_sge0_rx",
CTLTYPE_STRING | CTLFLAG_RD, sc, 6 + CIM_NUM_IBQ,
sysctl_cim_ibq_obq, "A", "CIM OBQ 6 (SGE0-RX)");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_obq_sge1_rx",
CTLTYPE_STRING | CTLFLAG_RD, sc, 7 + CIM_NUM_IBQ,
sysctl_cim_ibq_obq, "A", "CIM OBQ 7 (SGE1-RX)");
}
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_pif_la",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
sysctl_cim_pif_la, "A", "CIM PIF logic analyzer");
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_qcfg",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
sysctl_cim_qcfg, "A", "CIM queue configuration");
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cpl_stats",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
sysctl_cpl_stats, "A", "CPL statistics");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "ddp_stats",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
cxgbe(4): fix the description of a strange bunch of counters. MFC after: 1 week
2015-01-05 23:43:24 +00:00
sysctl_ddp_stats, "A", "non-TCP DDP statistics");
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Firmware device log. # sysctl dev.t4nex.0.devlog MFC after: mdf's sysctl+sbuf changes are MFC'd
2011-05-31 23:49:13 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "devlog",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
sysctl_devlog, "A", "firmware's device log");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "fcoe_stats",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
sysctl_fcoe_stats, "A", "FCoE statistics");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "hw_sched",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
sysctl_hw_sched, "A", "hardware scheduler ");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "l2t",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
sysctl_l2t, "A", "hardware L2 table");
cxgbe(4): Add code to deal with the chip's source MAC table (aka SMT). Submitted by: Krishnamraju Eraparaju @ Chelsio Sponsored by: Chelsio Communications
2018-05-31 21:31:08 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "smt",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
sysctl_smt, "A", "hardware source MAC table");
Make most of the CLIP code conditional on #ifdef INET6. This fixes builds of kernels without INET6 such as LINT-NOINET6. Reported by: arybchik Reviewed by: np Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18384
2018-11-29 23:14:54 +00:00
#ifdef INET6
Move CLIP table handling out of TOM and into the base driver. - Store the clip table in 'struct adapter' instead of in the TOM softc. - Init the clip table during attach and teardown during detach. - While here, add a dev.<nexus>.<unit>.misc.clip sysctl to dump the CLIP table. This does mean that we update the clip table even if TOE is not enabled, but non-TOE things need the CLIP table anyway. Reviewed by: np, Krishnamraju Eraparaju @ Chelsio Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18010
2018-11-29 01:15:53 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "clip",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
sysctl_clip, "A", "active CLIP table entries");
Make most of the CLIP code conditional on #ifdef INET6. This fixes builds of kernels without INET6 such as LINT-NOINET6. Reported by: arybchik Reviewed by: np Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18384
2018-11-29 23:14:54 +00:00
#endif
Move CLIP table handling out of TOM and into the base driver. - Store the clip table in 'struct adapter' instead of in the TOM softc. - Init the clip table during attach and teardown during detach. - While here, add a dev.<nexus>.<unit>.misc.clip sysctl to dump the CLIP table. This does mean that we update the clip table even if TOE is not enabled, but non-TOE things need the CLIP table anyway. Reviewed by: np, Krishnamraju Eraparaju @ Chelsio Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18010
2018-11-29 01:15:53 +00:00
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "lb_stats",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
sysctl_lb_stats, "A", "loopback statistics");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "meminfo",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
sysctl_meminfo, "A", "memory regions");
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "mps_tcam",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
chip_id(sc) <= CHELSIO_T5 ? sysctl_mps_tcam : sysctl_mps_tcam_t6,
"A", "MPS TCAM entries");
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "path_mtus",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
sysctl_path_mtus, "A", "path MTUs");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "pm_stats",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
sysctl_pm_stats, "A", "PM statistics");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "rdma_stats",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
sysctl_rdma_stats, "A", "RDMA statistics");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "tcp_stats",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
sysctl_tcp_stats, "A", "TCP statistics");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "tids",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
sysctl_tids, "A", "TID information");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "tp_err_stats",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
sysctl_tp_err_stats, "A", "TP error statistics");
cxgbe(4): Add a sysctl for the event capture mask of the TP block's logic analyzer. dev.t5nex.<n>.misc.tp_la_mask dev.t4nex.<n>.misc.tp_la_mask
2016-03-11 01:54:43 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "tp_la_mask",
CTLTYPE_INT | CTLFLAG_RW, sc, 0, sysctl_tp_la_mask, "I",
"TP logic analyzer event capture mask");
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "tp_la",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
sysctl_tp_la, "A", "TP logic analyzer");
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "tx_rate",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
sysctl_tx_rate, "A", "Tx rate");
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "ulprx_la",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
sysctl_ulprx_la, "A", "ULPRX logic analyzer");
cxgbe(4): Show wcwr_stats for T6 cards.
2016-09-21 00:46:08 +00:00
if (chip_id(sc) >= CHELSIO_T5) {
Cosmetic change (s/wrwc/wcwr/;s/WRWC/WCWR/). MFC after: 3 days.
2013-04-11 22:49:29 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "wcwr_stats",
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
Cosmetic change (s/wrwc/wcwr/;s/WRWC/WCWR/). MFC after: 3 days.
2013-04-11 22:49:29 +00:00
sysctl_wcwr_stats, "A", "write combined work requests");
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
#ifdef TCP_OFFLOAD
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
if (is_offload(sc)) {
cxgbe(4): Display some more TOE parameters related to retransmission and keepalive in the sysctl MIB. Provide tunables to change some of these parameters. These are supposed to be setup by the firmware so these tunables are for experimentation only. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-24 19:17:13 +00:00
int i;
char s[4];
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
/*
* dev.t4nex.X.toe.
*/
oid = SYSCTL_ADD_NODE(ctx, c0, OID_AUTO, "toe", CTLFLAG_RD,
NULL, "TOE parameters");
children = SYSCTL_CHILDREN(oid);
cxgbe/t4_tom: Add a knob to select the congestion control algorigthm used by the TOE hardware for fully offloaded connections. The knob affects new connections only. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-08-31 20:33:22 +00:00
sc->tt.cong_algorithm = -1;
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "cong_algorithm",
CTLFLAG_RW, &sc->tt.cong_algorithm, 0, "congestion control "
"(-1 = default, 0 = reno, 1 = tahoe, 2 = newreno, "
"3 = highspeed)");
cxgbe/t4_tom: Initialize all TOE connection parameters in one place. Remove now-redundant items from toepcb and synq_entry and the code to support them. Let the driver calculate tx_align, rx_coalesce, and sndbuf by default. Reviewed by: jhb@ MFC after: 1 week Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D21387
2019-08-27 04:19:40 +00:00
sc->tt.sndbuf = -1;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "sndbuf", CTLFLAG_RW,
cxgbe/t4_tom: Initialize all TOE connection parameters in one place. Remove now-redundant items from toepcb and synq_entry and the code to support them. Let the driver calculate tx_align, rx_coalesce, and sndbuf by default. Reviewed by: jhb@ MFC after: 1 week Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D21387
2019-08-27 04:19:40 +00:00
&sc->tt.sndbuf, 0, "hardware send buffer");
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
sc->tt.ddp = 0;
cxgbe(4): Rename the DDP sysctl to rx_zcopy to match the tx_zcopy sysctl and update its description. The old name continues to work for now. Sponsored by: Chelsio Communications
2019-06-07 05:03:03 +00:00
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "ddp",
CTLFLAG_RW | CTLFLAG_SKIP, &sc->tt.ddp, 0, "");
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "rx_zcopy", CTLFLAG_RW,
&sc->tt.ddp, 0, "Enable zero-copy aio_read(2)");
Initialize various DDP parameters in the main cxgbe(4) driver: - Setup multiple DDP page sizes. When the driver attempts DDP it will try to combine physically contiguous pages into regions of these sizes. - Set the indicate size such that the payload carried in the indicate can be copied in the header mbuf (and the 16K rx buffer can be recycled). - Set DDP threshold to the max payload that the chip will coalesce and deliver to the driver (this is ~16K by default, which is also why the offload rx queue is backed by 16K buffers). If the chip is able to coalesce up to the max it's allowed to, it's a good sign that the peer is transmitting in bulk without any TCP PSH. MFC after: 2 weeks
2012-08-16 22:33:56 +00:00
cxgbe/t4_tom: Initialize all TOE connection parameters in one place. Remove now-redundant items from toepcb and synq_entry and the code to support them. Let the driver calculate tx_align, rx_coalesce, and sndbuf by default. Reviewed by: jhb@ MFC after: 1 week Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D21387
2019-08-27 04:19:40 +00:00
sc->tt.rx_coalesce = -1;
- Make note of interface MTU change if the rx queues exist, and not just when the interface is up. - Add a tunable to control the TOE's rx coalesce feature (enabled by default as it always has been). Consider the interface MTU or the coalesce size when deciding which cluster zone to use to fill the offload rx queue's free list. The tunable is: dev.{t4nex,t5nex}.<N>.toe.rx_coalesce MFC after: 1 day
2013-07-04 21:19:01 +00:00
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "rx_coalesce",
CTLFLAG_RW, &sc->tt.rx_coalesce, 0, "receive coalescing");
cxgbe/tom: do not engage the TOE's payload chopper for payload < 2 MSS or for 10Gbps ports. MFC after: 2 weeks
2015-01-03 00:09:21 +00:00
Support for TLS offload of TOE connections on T6 adapters. The TOE engine in Chelsio T6 adapters supports offloading of TLS encryption and TCP segmentation for offloaded connections. Sockets using TLS are required to use a set of custom socket options to upload RX and TX keys to the NIC and to enable RX processing. Currently these socket options are implemented as TCP options in the vendor specific range. A patched OpenSSL library will be made available in a port / package for use with the TLS TOE support. TOE sockets can either offload both transmit and reception of TLS records or just transmit. TLS offload (both RX and TX) is enabled by setting the dev.t6nex.<x>.tls sysctl to 1 and requires TOE to be enabled on the relevant interface. Transmit offload can be used on any "normal" or TLS TOE socket by using the custom socket option to program a transmit key. This permits most TOE sockets to transparently offload TLS when applications use a patched SSL library (e.g. using LD_LIBRARY_PATH to request use of a patched OpenSSL library). Receive offload can only be used with TOE sockets using the TLS mode. The dev.t6nex.0.toe.tls_rx_ports sysctl can be set to a list of TCP port numbers. Any connection with either a local or remote port number in that list will be created as a TLS socket rather than a plain TOE socket. Note that although this sysctl accepts an arbitrary list of port numbers, the sysctl(8) tool is only able to set sysctl nodes to a single value. A TLS socket will hang without receiving data if used by an application that is not using a patched SSL library. Thus, the tls_rx_ports node should be used with care. For a server mostly concerned with offloading TLS transmit, this node is not needed as plain TOE sockets will fall back to software crypto when using an unpatched SSL library. New per-interface statistics nodes are added giving counts of TLS packets and payload bytes (payload bytes do not include TLS headers or authentication tags/MACs) offloaded via the TOE engine, e.g.: dev.cc.0.stats.rx_tls_octets: 149 dev.cc.0.stats.rx_tls_records: 13 dev.cc.0.stats.tx_tls_octets: 26501823 dev.cc.0.stats.tx_tls_records: 1620 TLS transmit work requests are constructed by a new variant of t4_push_frames() called t4_push_tls_records() in tom/t4_tls.c. TLS transmit work requests require a buffer containing IVs. If the IVs are too large to fit into the work request, a separate buffer is allocated when constructing a work request. This buffer is associated with the transmit descriptor and freed when the descriptor is ACKed by the adapter. Received TLS frames use two new CPL messages. The first message is a CPL_TLS_DATA containing the decryped payload of a single TLS record. The handler places the mbuf containing the received payload on an mbufq in the TOE pcb. The second message is a CPL_RX_TLS_CMP message which includes a copy of the TLS header and indicates if there were any errors. The handler for this message places the TLS header into the socket buffer followed by the saved mbuf with the payload data. Both of these handlers are contained in tom/t4_tls.c. A few routines were exposed from t4_cpl_io.c for use by t4_tls.c including send_rx_credits(), a new send_rx_modulate(), and t4_close_conn(). TLS keys for both transmit and receive are stored in onboard memory in the NIC in the "TLS keys" memory region. In some cases a TLS socket can hang with pending data available in the NIC that is not delivered to the host. As a workaround, TLS sockets are more aggressive about sending CPL_RX_DATA_ACK messages anytime that any data is read from a TLS socket. In addition, a fallback timer will periodically send CPL_RX_DATA_ACK messages to the NIC for connections that are still in the handshake phase. Once the connection has finished the handshake and programmed RX keys via the socket option, the timer is stopped. A new function select_ulp_mode() is used to determine what sub-mode a given TOE socket should use (plain TOE, DDP, or TLS). The existing set_tcpddp_ulp_mode() function has been renamed to set_ulp_mode() and handles initialization of TLS-specific state when necessary in addition to DDP-specific state. Since TLS sockets do not receive individual TCP segments but always receive full TLS records, they can receive more data than is available in the current window (e.g. if a 16k TLS record is received but the socket buffer is itself 16k). To cope with this, just drop the window to 0 when this happens, but track the overage and "eat" the overage as it is read from the socket buffer not opening the window (or adding rx_credits) for the overage bytes. Reviewed by: np (earlier version) Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D14529
2018-03-13 23:05:51 +00:00
sc->tt.tls = 0;
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "tls", CTLFLAG_RW,
&sc->tt.tls, 0, "Inline TLS allowed");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "tls_rx_ports",
CTLTYPE_INT | CTLFLAG_RW, sc, 0, sysctl_tls_rx_ports,
"I", "TCP ports that use inline TLS+TOE RX");
cxgbe/t4_tom: Initialize all TOE connection parameters in one place. Remove now-redundant items from toepcb and synq_entry and the code to support them. Let the driver calculate tx_align, rx_coalesce, and sndbuf by default. Reviewed by: jhb@ MFC after: 1 week Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D21387
2019-08-27 04:19:40 +00:00
sc->tt.tx_align = -1;
cxgbe/tom: do not engage the TOE's payload chopper for payload < 2 MSS or for 10Gbps ports. MFC after: 2 weeks
2015-01-03 00:09:21 +00:00
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "tx_align",
CTLFLAG_RW, &sc->tt.tx_align, 0, "chop and align payload");
cxgbe(4): sysctls to display the TOE's TCP timers. cask:~# sysctl -d dev.t5nex.0.toe dev.t5nex.0.toe.finwait2_timer: FINWAIT2 timer (us) dev.t5nex.0.toe.initial_srtt: Initial SRTT (us) dev.t5nex.0.toe.keepalive_intvl: Keepidle interval (us) dev.t5nex.0.toe.keepalive_idle: Keepidle idle timer (us) dev.t5nex.0.toe.persist_max: Persist timer max (us) dev.t5nex.0.toe.persist_min: Persist timer min (us) dev.t5nex.0.toe.rexmt_max: Retransmit max (us) dev.t5nex.0.toe.rexmt_min: Retransmit min (us) dev.t5nex.0.toe.dack_timer: DACK timer (us) dev.t5nex.0.toe.dack_tick: DACK tick (us) dev.t5nex.0.toe.timestamp_tick: TCP timestamp tick (us) dev.t5nex.0.toe.timer_tick: TP timer tick (us) ... cask:~# sysctl dev.t5nex.0.toe dev.t5nex.0.toe.finwait2_timer: 9765440 dev.t5nex.0.toe.initial_srtt: 244128 dev.t5nex.0.toe.keepalive_intvl: 73240800 dev.t5nex.0.toe.keepalive_idle: 7031116800 dev.t5nex.0.toe.persist_max: 9765440 dev.t5nex.0.toe.persist_min: 976544 dev.t5nex.0.toe.rexmt_max: 9765440 dev.t5nex.0.toe.rexmt_min: 244128 dev.t5nex.0.toe.dack_timer: 19520 dev.t5nex.0.toe.dack_tick: 32.768 dev.t5nex.0.toe.timestamp_tick: 1048.576 dev.t5nex.0.toe.timer_tick: 32.768 ...
2016-03-11 23:24:04 +00:00
Add support for zero-copy aio_write() on TOE sockets. AIO write requests for a TOE socket on a Chelsio T4+ adapter can now DMA directly from the user-supplied buffer. This is implemented by wiring the pages backing the user-supplied buffer and queueing special mbufs backed by raw VM pages to the socket buffer. The TOE code recognizes these special mbufs and builds a sglist from the VM page array associated with the mbuf when queueing a work request to the TOE. Because these mbufs do not have an associated virtual address, m_data is not valid. Thus, the AIO handler does not invoke sosend() directly for these mbufs but instead inlines portions of sosend_generic() and tcp_usr_send(). An aiotx_buffer structure is used to describe the user buffer (e.g. it holds the array of VM pages and a reference to the AIO job). The special mbufs reference this structure via m_ext. Note that a single job might be split across multiple mbufs (e.g. if it is larger than the socket buffer size). The 'ext_arg2' member of each mbuf gives an offset relative to the backing aiotx_buffer. The AIO job associated with an aiotx_buffer structure is completed when the last reference to the structure is released. Zero-copy aio_write()'s for connections associated with a given adapter can be enabled/disabled at runtime via the 'dev.t[45]nex.N.toe.tx_zcopy' sysctl. MFC after: 1 month Relnotes: yes Sponsored by: Chelsio Communications
2016-07-27 18:29:35 +00:00
sc->tt.tx_zcopy = 0;
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "tx_zcopy",
CTLFLAG_RW, &sc->tt.tx_zcopy, 0,
"Enable zero-copy aio_write(2)");
cxgbe(4): Add support for Connection Offload Policy (aka COP). COP allows fine-grained control on whether to offload a TCP connection using t4_tom, and what settings to apply to a connection selected for offload. t4_tom must still be loaded and IFCAP_TOE must still be enabled for full TCP offload to take place on an interface. The difference is that IFCAP_TOE used to be the only knob and would enable TOE for all new connections on the inteface, but now the driver will also consult the COP, if any, before offloading to the hardware TOE. A policy is a plain text file with any number of rules, one per line. Each rule has a "match" part consisting of a socket-type (L = listen, A = active open, P = passive open, D = don't care) and a pcap-filter(7) expression, and a "settings" part that specifies whether to offload the connection or not and the parameters to use if so. The general format of a rule is: [socket-type] expr => settings Example. See cxgbetool(8) for more information. [L] ip && port http => offload [L] port 443 => !offload [L] port ssh => offload [P] src net 192.168/16 && dst port ssh => offload !nagle !timestamp cong newreno [P] dst port ssh => offload !nagle ecn cong tahoe [P] dst port http => offload [A] dst port 443 => offload tls [A] dst net 192.168/16 => offload !timestamp cong highspeed The driver processes the rules for each new listen, active open, or passive open and stops at the first match. There is an implicit rule at the end of every policy that prohibits offload when no rule in the policy matches: [D] all => !offload This is a reworked and expanded version of a patch submitted by Krishnamraju Eraparaju @ Chelsio. Sponsored by: Chelsio Communications
2018-04-14 19:07:56 +00:00
sc->tt.cop_managed_offloading = !!t4_cop_managed_offloading;
SYSCTL_ADD_INT(ctx, children, OID_AUTO,
"cop_managed_offloading", CTLFLAG_RW,
&sc->tt.cop_managed_offloading, 0,
"COP (Connection Offload Policy) controls all TOE offload");
cxgbe/t4_tom: Catch up with r344433, which removed tcb_autorcvbuf_inc. The declaration in tcp_var.h is still around so t4_tom continued to compile but wouldn't load. A separate commit will fix tcp_var.h Reported By: Dustin Marquess (dmarquess at gmail) Sponsored by: Chelsio Communications
2019-03-29 16:43:24 +00:00
sc->tt.autorcvbuf_inc = 16 * 1024;
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "autorcvbuf_inc",
CTLFLAG_RW, &sc->tt.autorcvbuf_inc, 0,
"autorcvbuf increment");
cxgbe(4): sysctls to display the TOE's TCP timers. cask:~# sysctl -d dev.t5nex.0.toe dev.t5nex.0.toe.finwait2_timer: FINWAIT2 timer (us) dev.t5nex.0.toe.initial_srtt: Initial SRTT (us) dev.t5nex.0.toe.keepalive_intvl: Keepidle interval (us) dev.t5nex.0.toe.keepalive_idle: Keepidle idle timer (us) dev.t5nex.0.toe.persist_max: Persist timer max (us) dev.t5nex.0.toe.persist_min: Persist timer min (us) dev.t5nex.0.toe.rexmt_max: Retransmit max (us) dev.t5nex.0.toe.rexmt_min: Retransmit min (us) dev.t5nex.0.toe.dack_timer: DACK timer (us) dev.t5nex.0.toe.dack_tick: DACK tick (us) dev.t5nex.0.toe.timestamp_tick: TCP timestamp tick (us) dev.t5nex.0.toe.timer_tick: TP timer tick (us) ... cask:~# sysctl dev.t5nex.0.toe dev.t5nex.0.toe.finwait2_timer: 9765440 dev.t5nex.0.toe.initial_srtt: 244128 dev.t5nex.0.toe.keepalive_intvl: 73240800 dev.t5nex.0.toe.keepalive_idle: 7031116800 dev.t5nex.0.toe.persist_max: 9765440 dev.t5nex.0.toe.persist_min: 976544 dev.t5nex.0.toe.rexmt_max: 9765440 dev.t5nex.0.toe.rexmt_min: 244128 dev.t5nex.0.toe.dack_timer: 19520 dev.t5nex.0.toe.dack_tick: 32.768 dev.t5nex.0.toe.timestamp_tick: 1048.576 dev.t5nex.0.toe.timer_tick: 32.768 ...
2016-03-11 23:24:04 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "timer_tick",
CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_tp_tick, "A",
"TP timer tick (us)");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "timestamp_tick",
CTLTYPE_STRING | CTLFLAG_RD, sc, 1, sysctl_tp_tick, "A",
"TCP timestamp tick (us)");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "dack_tick",
CTLTYPE_STRING | CTLFLAG_RD, sc, 2, sysctl_tp_tick, "A",
"DACK tick (us)");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "dack_timer",
CTLTYPE_UINT | CTLFLAG_RD, sc, 0, sysctl_tp_dack_timer,
"IU", "DACK timer (us)");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "rexmt_min",
CTLTYPE_ULONG | CTLFLAG_RD, sc, A_TP_RXT_MIN,
cxgbe(4): Display some more TOE parameters related to retransmission and keepalive in the sysctl MIB. Provide tunables to change some of these parameters. These are supposed to be setup by the firmware so these tunables are for experimentation only. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-24 19:17:13 +00:00
sysctl_tp_timer, "LU", "Minimum retransmit interval (us)");
cxgbe(4): sysctls to display the TOE's TCP timers. cask:~# sysctl -d dev.t5nex.0.toe dev.t5nex.0.toe.finwait2_timer: FINWAIT2 timer (us) dev.t5nex.0.toe.initial_srtt: Initial SRTT (us) dev.t5nex.0.toe.keepalive_intvl: Keepidle interval (us) dev.t5nex.0.toe.keepalive_idle: Keepidle idle timer (us) dev.t5nex.0.toe.persist_max: Persist timer max (us) dev.t5nex.0.toe.persist_min: Persist timer min (us) dev.t5nex.0.toe.rexmt_max: Retransmit max (us) dev.t5nex.0.toe.rexmt_min: Retransmit min (us) dev.t5nex.0.toe.dack_timer: DACK timer (us) dev.t5nex.0.toe.dack_tick: DACK tick (us) dev.t5nex.0.toe.timestamp_tick: TCP timestamp tick (us) dev.t5nex.0.toe.timer_tick: TP timer tick (us) ... cask:~# sysctl dev.t5nex.0.toe dev.t5nex.0.toe.finwait2_timer: 9765440 dev.t5nex.0.toe.initial_srtt: 244128 dev.t5nex.0.toe.keepalive_intvl: 73240800 dev.t5nex.0.toe.keepalive_idle: 7031116800 dev.t5nex.0.toe.persist_max: 9765440 dev.t5nex.0.toe.persist_min: 976544 dev.t5nex.0.toe.rexmt_max: 9765440 dev.t5nex.0.toe.rexmt_min: 244128 dev.t5nex.0.toe.dack_timer: 19520 dev.t5nex.0.toe.dack_tick: 32.768 dev.t5nex.0.toe.timestamp_tick: 1048.576 dev.t5nex.0.toe.timer_tick: 32.768 ...
2016-03-11 23:24:04 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "rexmt_max",
CTLTYPE_ULONG | CTLFLAG_RD, sc, A_TP_RXT_MAX,
cxgbe(4): Display some more TOE parameters related to retransmission and keepalive in the sysctl MIB. Provide tunables to change some of these parameters. These are supposed to be setup by the firmware so these tunables are for experimentation only. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-24 19:17:13 +00:00
sysctl_tp_timer, "LU", "Maximum retransmit interval (us)");
cxgbe(4): sysctls to display the TOE's TCP timers. cask:~# sysctl -d dev.t5nex.0.toe dev.t5nex.0.toe.finwait2_timer: FINWAIT2 timer (us) dev.t5nex.0.toe.initial_srtt: Initial SRTT (us) dev.t5nex.0.toe.keepalive_intvl: Keepidle interval (us) dev.t5nex.0.toe.keepalive_idle: Keepidle idle timer (us) dev.t5nex.0.toe.persist_max: Persist timer max (us) dev.t5nex.0.toe.persist_min: Persist timer min (us) dev.t5nex.0.toe.rexmt_max: Retransmit max (us) dev.t5nex.0.toe.rexmt_min: Retransmit min (us) dev.t5nex.0.toe.dack_timer: DACK timer (us) dev.t5nex.0.toe.dack_tick: DACK tick (us) dev.t5nex.0.toe.timestamp_tick: TCP timestamp tick (us) dev.t5nex.0.toe.timer_tick: TP timer tick (us) ... cask:~# sysctl dev.t5nex.0.toe dev.t5nex.0.toe.finwait2_timer: 9765440 dev.t5nex.0.toe.initial_srtt: 244128 dev.t5nex.0.toe.keepalive_intvl: 73240800 dev.t5nex.0.toe.keepalive_idle: 7031116800 dev.t5nex.0.toe.persist_max: 9765440 dev.t5nex.0.toe.persist_min: 976544 dev.t5nex.0.toe.rexmt_max: 9765440 dev.t5nex.0.toe.rexmt_min: 244128 dev.t5nex.0.toe.dack_timer: 19520 dev.t5nex.0.toe.dack_tick: 32.768 dev.t5nex.0.toe.timestamp_tick: 1048.576 dev.t5nex.0.toe.timer_tick: 32.768 ...
2016-03-11 23:24:04 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "persist_min",
CTLTYPE_ULONG | CTLFLAG_RD, sc, A_TP_PERS_MIN,
sysctl_tp_timer, "LU", "Persist timer min (us)");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "persist_max",
CTLTYPE_ULONG | CTLFLAG_RD, sc, A_TP_PERS_MAX,
sysctl_tp_timer, "LU", "Persist timer max (us)");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "keepalive_idle",
CTLTYPE_ULONG | CTLFLAG_RD, sc, A_TP_KEEP_IDLE,
cxgbe(4): Display some more TOE parameters related to retransmission and keepalive in the sysctl MIB. Provide tunables to change some of these parameters. These are supposed to be setup by the firmware so these tunables are for experimentation only. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-24 19:17:13 +00:00
sysctl_tp_timer, "LU", "Keepalive idle timer (us)");
cxgbe(4): sysctls to display the TOE's TCP timers. cask:~# sysctl -d dev.t5nex.0.toe dev.t5nex.0.toe.finwait2_timer: FINWAIT2 timer (us) dev.t5nex.0.toe.initial_srtt: Initial SRTT (us) dev.t5nex.0.toe.keepalive_intvl: Keepidle interval (us) dev.t5nex.0.toe.keepalive_idle: Keepidle idle timer (us) dev.t5nex.0.toe.persist_max: Persist timer max (us) dev.t5nex.0.toe.persist_min: Persist timer min (us) dev.t5nex.0.toe.rexmt_max: Retransmit max (us) dev.t5nex.0.toe.rexmt_min: Retransmit min (us) dev.t5nex.0.toe.dack_timer: DACK timer (us) dev.t5nex.0.toe.dack_tick: DACK tick (us) dev.t5nex.0.toe.timestamp_tick: TCP timestamp tick (us) dev.t5nex.0.toe.timer_tick: TP timer tick (us) ... cask:~# sysctl dev.t5nex.0.toe dev.t5nex.0.toe.finwait2_timer: 9765440 dev.t5nex.0.toe.initial_srtt: 244128 dev.t5nex.0.toe.keepalive_intvl: 73240800 dev.t5nex.0.toe.keepalive_idle: 7031116800 dev.t5nex.0.toe.persist_max: 9765440 dev.t5nex.0.toe.persist_min: 976544 dev.t5nex.0.toe.rexmt_max: 9765440 dev.t5nex.0.toe.rexmt_min: 244128 dev.t5nex.0.toe.dack_timer: 19520 dev.t5nex.0.toe.dack_tick: 32.768 dev.t5nex.0.toe.timestamp_tick: 1048.576 dev.t5nex.0.toe.timer_tick: 32.768 ...
2016-03-11 23:24:04 +00:00
cxgbe(4): Display some more TOE parameters related to retransmission and keepalive in the sysctl MIB. Provide tunables to change some of these parameters. These are supposed to be setup by the firmware so these tunables are for experimentation only. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-24 19:17:13 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "keepalive_interval",
cxgbe(4): sysctls to display the TOE's TCP timers. cask:~# sysctl -d dev.t5nex.0.toe dev.t5nex.0.toe.finwait2_timer: FINWAIT2 timer (us) dev.t5nex.0.toe.initial_srtt: Initial SRTT (us) dev.t5nex.0.toe.keepalive_intvl: Keepidle interval (us) dev.t5nex.0.toe.keepalive_idle: Keepidle idle timer (us) dev.t5nex.0.toe.persist_max: Persist timer max (us) dev.t5nex.0.toe.persist_min: Persist timer min (us) dev.t5nex.0.toe.rexmt_max: Retransmit max (us) dev.t5nex.0.toe.rexmt_min: Retransmit min (us) dev.t5nex.0.toe.dack_timer: DACK timer (us) dev.t5nex.0.toe.dack_tick: DACK tick (us) dev.t5nex.0.toe.timestamp_tick: TCP timestamp tick (us) dev.t5nex.0.toe.timer_tick: TP timer tick (us) ... cask:~# sysctl dev.t5nex.0.toe dev.t5nex.0.toe.finwait2_timer: 9765440 dev.t5nex.0.toe.initial_srtt: 244128 dev.t5nex.0.toe.keepalive_intvl: 73240800 dev.t5nex.0.toe.keepalive_idle: 7031116800 dev.t5nex.0.toe.persist_max: 9765440 dev.t5nex.0.toe.persist_min: 976544 dev.t5nex.0.toe.rexmt_max: 9765440 dev.t5nex.0.toe.rexmt_min: 244128 dev.t5nex.0.toe.dack_timer: 19520 dev.t5nex.0.toe.dack_tick: 32.768 dev.t5nex.0.toe.timestamp_tick: 1048.576 dev.t5nex.0.toe.timer_tick: 32.768 ...
2016-03-11 23:24:04 +00:00
CTLTYPE_ULONG | CTLFLAG_RD, sc, A_TP_KEEP_INTVL,
cxgbe(4): Display some more TOE parameters related to retransmission and keepalive in the sysctl MIB. Provide tunables to change some of these parameters. These are supposed to be setup by the firmware so these tunables are for experimentation only. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-24 19:17:13 +00:00
sysctl_tp_timer, "LU", "Keepalive interval timer (us)");
cxgbe(4): sysctls to display the TOE's TCP timers. cask:~# sysctl -d dev.t5nex.0.toe dev.t5nex.0.toe.finwait2_timer: FINWAIT2 timer (us) dev.t5nex.0.toe.initial_srtt: Initial SRTT (us) dev.t5nex.0.toe.keepalive_intvl: Keepidle interval (us) dev.t5nex.0.toe.keepalive_idle: Keepidle idle timer (us) dev.t5nex.0.toe.persist_max: Persist timer max (us) dev.t5nex.0.toe.persist_min: Persist timer min (us) dev.t5nex.0.toe.rexmt_max: Retransmit max (us) dev.t5nex.0.toe.rexmt_min: Retransmit min (us) dev.t5nex.0.toe.dack_timer: DACK timer (us) dev.t5nex.0.toe.dack_tick: DACK tick (us) dev.t5nex.0.toe.timestamp_tick: TCP timestamp tick (us) dev.t5nex.0.toe.timer_tick: TP timer tick (us) ... cask:~# sysctl dev.t5nex.0.toe dev.t5nex.0.toe.finwait2_timer: 9765440 dev.t5nex.0.toe.initial_srtt: 244128 dev.t5nex.0.toe.keepalive_intvl: 73240800 dev.t5nex.0.toe.keepalive_idle: 7031116800 dev.t5nex.0.toe.persist_max: 9765440 dev.t5nex.0.toe.persist_min: 976544 dev.t5nex.0.toe.rexmt_max: 9765440 dev.t5nex.0.toe.rexmt_min: 244128 dev.t5nex.0.toe.dack_timer: 19520 dev.t5nex.0.toe.dack_tick: 32.768 dev.t5nex.0.toe.timestamp_tick: 1048.576 dev.t5nex.0.toe.timer_tick: 32.768 ...
2016-03-11 23:24:04 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "initial_srtt",
CTLTYPE_ULONG | CTLFLAG_RD, sc, A_TP_INIT_SRTT,
sysctl_tp_timer, "LU", "Initial SRTT (us)");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "finwait2_timer",
CTLTYPE_ULONG | CTLFLAG_RD, sc, A_TP_FINWAIT2_TIMER,
sysctl_tp_timer, "LU", "FINWAIT2 timer (us)");
cxgbe(4): Display some more TOE parameters related to retransmission and keepalive in the sysctl MIB. Provide tunables to change some of these parameters. These are supposed to be setup by the firmware so these tunables are for experimentation only. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-24 19:17:13 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "syn_rexmt_count",
CTLTYPE_UINT | CTLFLAG_RD, sc, S_SYNSHIFTMAX,
sysctl_tp_shift_cnt, "IU",
"Number of SYN retransmissions before abort");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "rexmt_count",
CTLTYPE_UINT | CTLFLAG_RD, sc, S_RXTSHIFTMAXR2,
sysctl_tp_shift_cnt, "IU",
"Number of retransmissions before abort");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "keepalive_count",
CTLTYPE_UINT | CTLFLAG_RD, sc, S_KEEPALIVEMAXR2,
sysctl_tp_shift_cnt, "IU",
"Number of keepalive probes before abort");
oid = SYSCTL_ADD_NODE(ctx, children, OID_AUTO, "rexmt_backoff",
CTLFLAG_RD, NULL, "TOE retransmit backoffs");
children = SYSCTL_CHILDREN(oid);
for (i = 0; i < 16; i++) {
snprintf(s, sizeof(s), "%u", i);
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, s,
CTLTYPE_UINT | CTLFLAG_RD, sc, i, sysctl_tp_backoff,
"IU", "TOE retransmit backoff");
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
#endif
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
void
vi_sysctls(struct vi_info *vi)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
{
struct sysctl_ctx_list *ctx;
struct sysctl_oid *oid;
struct sysctl_oid_list *children;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
ctx = device_get_sysctl_ctx(vi->dev);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
/*
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
* dev.v?(cxgbe|cxl).X.
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
*/
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
oid = device_get_sysctl_tree(vi->dev);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
children = SYSCTL_CHILDREN(oid);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
SYSCTL_ADD_UINT(ctx, children, OID_AUTO, "viid", CTLFLAG_RD, NULL,
vi->viid, "VI identifer");
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "nrxq", CTLFLAG_RD,
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
&vi->nrxq, 0, "# of rx queues");
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "ntxq", CTLFLAG_RD,
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
&vi->ntxq, 0, "# of tx queues");
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "first_rxq", CTLFLAG_RD,
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
&vi->first_rxq, 0, "index of first rx queue");
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "first_txq", CTLFLAG_RD,
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
&vi->first_txq, 0, "index of first tx queue");
cxgbe(4): new sysctl to display the start of the RSS region for a VI. dev.<ifname>.<inst>.rss_base For example: dev.cc.0.rss_base: 0 dev.cc.1.rss_base: 128 dev.vcc.0.rss_base: 256 dev.vcc.1.rss_base: 384 Sponsored by: Chelsio Communications
2018-10-25 01:20:32 +00:00
SYSCTL_ADD_UINT(ctx, children, OID_AUTO, "rss_base", CTLFLAG_RD, NULL,
vi->rss_base, "start of RSS indirection table");
cxgbe(4): Add sysctl to display the RSS indirection table size for an interface. dev.cxl.<n>.rss_size dev.vcxl.<n>.rss_size MFC after: 3 days
2016-07-08 18:13:23 +00:00
SYSCTL_ADD_UINT(ctx, children, OID_AUTO, "rss_size", CTLFLAG_RD, NULL,
vi->rss_size, "size of RSS indirection table");
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
if (IS_MAIN_VI(vi)) {
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "rsrv_noflowq",
CTLTYPE_INT | CTLFLAG_RW, vi, 0, sysctl_noflowq, "IU",
"Reserve queue 0 for non-flowid packets");
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
#ifdef TCP_OFFLOAD
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (vi->nofldrxq != 0) {
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "nofldrxq", CTLFLAG_RD,
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
&vi->nofldrxq, 0,
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
"# of rx queues for offloaded TCP connections");
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "first_ofld_rxq",
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
CTLFLAG_RD, &vi->first_ofld_rxq, 0,
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
"index of first TOE rx queue");
cxgbe(4): Provide knobs to set the holdoff parameters of TOE rx queues separately from NIC rx queues instead of using the same parameters for both types of queues. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-10-05 07:18:16 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "holdoff_tmr_idx_ofld",
CTLTYPE_INT | CTLFLAG_RW, vi, 0,
sysctl_holdoff_tmr_idx_ofld, "I",
"holdoff timer index for TOE queues");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "holdoff_pktc_idx_ofld",
CTLTYPE_INT | CTLFLAG_RW, vi, 0,
sysctl_holdoff_pktc_idx_ofld, "I",
"holdoff packet counter index for TOE queues");
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
#endif
Fix a few more places to handle ofld tx queues for RATELIMIT. - Drain offload transmit queues when RATELIMIT is enabled but TCP_OFFLOAD is not. - Expose the per-VI nofldtxq and first_ofld_txq sysctls when RATELIMIT is enabled but TCP_OFFLOAD is not. - Clear offload transmit queue stats as part of a 'cxgbetool clearstats' request when RATELIMIT is enabled but TCP_OFFLOAD is not. Reviewed by: np MFC after: 2 weeks Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18966
2019-01-25 20:54:18 +00:00
#if defined(TCP_OFFLOAD) || defined(RATELIMIT)
if (vi->nofldtxq != 0) {
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "nofldtxq", CTLFLAG_RD,
&vi->nofldtxq, 0,
"# of tx queues for TOE/ETHOFLD");
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "first_ofld_txq",
CTLFLAG_RD, &vi->first_ofld_txq, 0,
"index of first TOE/ETHOFLD tx queue");
}
#endif
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
#ifdef DEV_NETMAP
if (vi->nnmrxq != 0) {
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "nnmrxq", CTLFLAG_RD,
&vi->nnmrxq, 0, "# of netmap rx queues");
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "nnmtxq", CTLFLAG_RD,
&vi->nnmtxq, 0, "# of netmap tx queues");
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "first_nm_rxq",
CTLFLAG_RD, &vi->first_nm_rxq, 0,
"index of first netmap rx queue");
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "first_nm_txq",
CTLFLAG_RD, &vi->first_nm_txq, 0,
"index of first netmap tx queue");
}
#endif
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "holdoff_tmr_idx",
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
CTLTYPE_INT | CTLFLAG_RW, vi, 0, sysctl_holdoff_tmr_idx, "I",
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
"holdoff timer index");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "holdoff_pktc_idx",
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
CTLTYPE_INT | CTLFLAG_RW, vi, 0, sysctl_holdoff_pktc_idx, "I",
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
"holdoff packet counter index");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "qsize_rxq",
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
CTLTYPE_INT | CTLFLAG_RW, vi, 0, sysctl_qsize_rxq, "I",
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
"rx queue size");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "qsize_txq",
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
CTLTYPE_INT | CTLFLAG_RW, vi, 0, sysctl_qsize_txq, "I",
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
"tx queue size");
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
}
static void
cxgbe_sysctls(struct port_info *pi)
{
struct sysctl_ctx_list *ctx;
struct sysctl_oid *oid;
cxgbe(4): Provide information about traffic classes in the sysctl mib. Sponsored by: Chelsio Communications
2016-06-07 06:42:35 +00:00
struct sysctl_oid_list *children, *children2;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct adapter *sc = pi->adapter;
cxgbe(4): Provide information about traffic classes in the sysctl mib. Sponsored by: Chelsio Communications
2016-06-07 06:42:35 +00:00
int i;
char name[16];
cxgbe(4): Allow user-configured and driver-configured traffic classes to be used simultaneously. Move sysctl_tc and sysctl_tc_params to t4_sched.c while here. MFC after: 3 weeks Sponsored by: Chelsio Communications
2018-08-06 23:21:13 +00:00
static char *tc_flags = {"\20\1USER\2SYNC\3ASYNC\4ERR"};
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
ctx = device_get_sysctl_ctx(pi->dev);
/*
* dev.cxgbe.X.
*/
oid = device_get_sysctl_tree(pi->dev);
children = SYSCTL_CHILDREN(oid);
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "linkdnrc", CTLTYPE_STRING |
CTLFLAG_RD, pi, 0, sysctl_linkdnrc, "A", "reason why link is down");
if (pi->port_type == FW_PORT_TYPE_BT_XAUI) {
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "temperature",
CTLTYPE_INT | CTLFLAG_RD, pi, 0, sysctl_btphy, "I",
"PHY temperature (in Celsius)");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "fw_version",
CTLTYPE_INT | CTLFLAG_RD, pi, 1, sysctl_btphy, "I",
"PHY firmware version");
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): knobs to enable/disable PAUSE frame based flow control. MFC after: 1 week
2014-09-12 05:25:56 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "pause_settings",
cxgbe(4): Updates to link configuration. - Update struct link_settings and associated shared code. - Add tunables to control FEC and autonegotiation. All ports inherit these values as their initial settings. hw.cxgbe.fec hw.cxgbe.autoneg - Add per-port sysctls to control FEC and autonegotiation. These can be modified at any time. dev.<port>.<n>.fec dev.<port>.<n>.autoneg MFC after: 3 days Sponsored by: Chelsio Communications
2016-12-30 08:59:49 +00:00
CTLTYPE_STRING | CTLFLAG_RW, pi, 0, sysctl_pause_settings, "A",
Add read-only sysctls for all tunables in the cxgbe(4) driver. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18360
2018-11-27 22:02:54 +00:00
"PAUSE settings (bit 0 = rx_pause, 1 = tx_pause, 2 = pause_autoneg)");
cxgbe(4): Updates to link configuration. - Update struct link_settings and associated shared code. - Add tunables to control FEC and autonegotiation. All ports inherit these values as their initial settings. hw.cxgbe.fec hw.cxgbe.autoneg - Add per-port sysctls to control FEC and autonegotiation. These can be modified at any time. dev.<port>.<n>.fec dev.<port>.<n>.autoneg MFC after: 3 days Sponsored by: Chelsio Communications
2016-12-30 08:59:49 +00:00
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "fec",
CTLTYPE_STRING | CTLFLAG_RW, pi, 0, sysctl_fec, "A",
"Forward Error Correction (bit 0 = RS, bit 1 = BASER_RS)");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "autoneg",
CTLTYPE_INT | CTLFLAG_RW, pi, 0, sysctl_autoneg, "I",
"autonegotiation (-1 = not supported)");
cxgbe(4): knobs to enable/disable PAUSE frame based flow control. MFC after: 1 week
2014-09-12 05:25:56 +00:00
cxgbe(4): Add a sysctl to retrieve the maximum speed/bandwidth supported by a port. dev.cxgbe.<n>.max_speed dev.cxl.<n>.max_speed Sponsored by: Chelsio Communications
2016-02-25 01:10:56 +00:00
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "max_speed", CTLFLAG_RD, NULL,
port_top_speed(pi), "max speed (in Gbps)");
cxgbe(4): Read the MPS buffer group map from the firmware as it could be different from hardware defaults. The congestion channel map, which is still fixed, needs to be tracked separately now. Change the congestion setting for TOE rx queues to match the drivers on other OSes while here. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-10-24 05:41:48 +00:00
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "mps_bg_map", CTLFLAG_RD, NULL,
pi->mps_bg_map, "MPS buffer group map");
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "rx_e_chan_map", CTLFLAG_RD,
NULL, pi->rx_e_chan_map, "TP rx e-channel map");
cxgbe(4): Add a sysctl to retrieve the maximum speed/bandwidth supported by a port. dev.cxgbe.<n>.max_speed dev.cxl.<n>.max_speed Sponsored by: Chelsio Communications
2016-02-25 01:10:56 +00:00
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
if (sc->flags & IS_VF)
return;
cxgbe(4): Provide information about traffic classes in the sysctl mib. Sponsored by: Chelsio Communications
2016-06-07 06:42:35 +00:00
/*
* dev.(cxgbe|cxl).X.tc.
*/
oid = SYSCTL_ADD_NODE(ctx, children, OID_AUTO, "tc", CTLFLAG_RD, NULL,
cxgbe(4): Support routines for Tx traffic scheduling. - Create a new file, t4_sched.c, and move all of the code related to traffic management from t4_main.c and t4_sge.c to this file. - Track both Channel Rate Limiter (ch_rl) and Class Rate Limiter (cl_rl) parameters in the PF driver. - Initialize all the cl_rl limiters with somewhat arbitrary default rates and provide routines to update them on the fly. - Provide routines to reserve and release traffic classes. MFC after: 1 month Sponsored by: Chelsio Communications
2017-05-02 20:38:10 +00:00
"Tx scheduler traffic classes (cl_rl)");
cxgbe(4): Allow the driver to specify a burst size when configuring a traffic class for rate limiting. Add experimental knobs that allow the user to specify a default pktsize and burstsize for traffic classes associated with a port: dev.<ifname>.<instance>.tc.pktsize dev.<ifname>.<instance>.tc.burstsize Sponsored by: Chelsio Communications
2018-08-07 22:13:03 +00:00
children2 = SYSCTL_CHILDREN(oid);
SYSCTL_ADD_UINT(ctx, children2, OID_AUTO, "pktsize",
CTLFLAG_RW, &pi->sched_params->pktsize, 0,
"pktsize for per-flow cl-rl (0 means up to the driver )");
SYSCTL_ADD_UINT(ctx, children2, OID_AUTO, "burstsize",
CTLFLAG_RW, &pi->sched_params->burstsize, 0,
"burstsize for per-flow cl-rl (0 means up to the driver)");
cxgbe(4): Provide information about traffic classes in the sysctl mib. Sponsored by: Chelsio Communications
2016-06-07 06:42:35 +00:00
for (i = 0; i < sc->chip_params->nsched_cls; i++) {
cxgbe(4): Support routines for Tx traffic scheduling. - Create a new file, t4_sched.c, and move all of the code related to traffic management from t4_main.c and t4_sge.c to this file. - Track both Channel Rate Limiter (ch_rl) and Class Rate Limiter (cl_rl) parameters in the PF driver. - Initialize all the cl_rl limiters with somewhat arbitrary default rates and provide routines to update them on the fly. - Provide routines to reserve and release traffic classes. MFC after: 1 month Sponsored by: Chelsio Communications
2017-05-02 20:38:10 +00:00
struct tx_cl_rl_params *tc = &pi->sched_params->cl_rl[i];
cxgbe(4): Provide information about traffic classes in the sysctl mib. Sponsored by: Chelsio Communications
2016-06-07 06:42:35 +00:00
snprintf(name, sizeof(name), "%d", i);
children2 = SYSCTL_CHILDREN(SYSCTL_ADD_NODE(ctx,
SYSCTL_CHILDREN(oid), OID_AUTO, name, CTLFLAG_RD, NULL,
"traffic class"));
cxgbe(4): Allow user-configured and driver-configured traffic classes to be used simultaneously. Move sysctl_tc and sysctl_tc_params to t4_sched.c while here. MFC after: 3 weeks Sponsored by: Chelsio Communications
2018-08-06 23:21:13 +00:00
SYSCTL_ADD_PROC(ctx, children2, OID_AUTO, "flags",
CTLTYPE_STRING | CTLFLAG_RD, tc_flags, (uintptr_t)&tc->flags,
sysctl_bitfield_8b, "A", "flags");
cxgbe(4): Provide information about traffic classes in the sysctl mib. Sponsored by: Chelsio Communications
2016-06-07 06:42:35 +00:00
SYSCTL_ADD_UINT(ctx, children2, OID_AUTO, "refcount",
CTLFLAG_RD, &tc->refcount, 0, "references to this class");
SYSCTL_ADD_PROC(ctx, children2, OID_AUTO, "params",
CTLTYPE_STRING | CTLFLAG_RD, sc, (pi->port_id << 16) | i,
sysctl_tc_params, "A", "traffic class parameters");
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
/*
* dev.cxgbe.X.stats.
*/
oid = SYSCTL_ADD_NODE(ctx, children, OID_AUTO, "stats", CTLFLAG_RD,
NULL, "port statistics");
children = SYSCTL_CHILDREN(oid);
cxgbe(4): major tx rework. a) Front load as much work as possible in if_transmit, before any driver lock or software queue has to get involved. b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This is specifically for the tx multiqueue model where one of the if_transmit producer threads becomes the consumer and other producers carry on as usual. mp_ring is implemented as standalone code and it should be possible to use it in any driver with tx multiqueue. It also has: - the ability to enqueue/dequeue multiple items. This might become significant if packet batching is ever implemented. - an abdication mechanism to allow a thread to give up writing tx descriptors and have another if_transmit thread take over. A thread that's writing tx descriptors can end up doing so for an unbounded time period if a) there are other if_transmit threads continuously feeding the sofware queue, and b) the chip keeps up with whatever the thread is throwing at it. - accurate statistics about interesting events even when the stats come at the expense of additional branches/conditional code. The NIC txq lock is uncontested on the fast path at this point. I've left it there for synchronization with the control events (interface up/down, modload/unload). c) Add support for "type 1" coalescing work request in the normal NIC tx path. This work request is optimized for frames with a single item in the DMA gather list. These are very common when forwarding packets. Note that netmap tx in cxgbe already uses these "type 1" work requests. d) Do not request automatic cidx updates every 32 descriptors. Instead, request updates via bits in individual work requests (still every 32 descriptors approximately). Also, request an automatic final update when the queue idles after activity. This means NIC tx reclaim is still performed lazily but it will catch up quickly as soon as the queue idles. This seems to be the best middle ground and I'll probably do something similar for netmap tx as well. e) Implement a faster tx path for WRQs (used by TOE tx and control queues, _not_ by the normal NIC tx). Allow work requests to be written directly to the hardware descriptor ring if room is available. I will convert t4_tom and iw_cxgbe modules to this faster style gradually. MFC after: 2 months
2014-12-31 23:19:16 +00:00
SYSCTL_ADD_UINT(ctx, children, OID_AUTO, "tx_parse_error", CTLFLAG_RD,
&pi->tx_parse_error, 0,
"# of tx packets with invalid length or # of segments");
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
#define SYSCTL_ADD_T4_REG64(pi, name, desc, reg) \
SYSCTL_ADD_OID(ctx, children, OID_AUTO, name, \
Read card capabilities after firmware initialization, instead of setting them up as part of firmware initialization (which the driver gets to do only if it's the master driver). Read the range of tids available for the ETHOFLD functionality if it's enabled. New is_ftid() and is_etid() functions to test whether a tid falls within the range of filter tids or ETHOFLD tids respectively. MFC after: 2 weeks
2013-12-14 03:08:03 +00:00
CTLTYPE_U64 | CTLFLAG_RD, sc, reg, \
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
sysctl_handle_t4_reg64, "QU", desc)
SYSCTL_ADD_T4_REG64(pi, "tx_octets", "# of octets in good frames",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_BYTES_L));
SYSCTL_ADD_T4_REG64(pi, "tx_frames", "total # of good frames",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_FRAMES_L));
SYSCTL_ADD_T4_REG64(pi, "tx_bcast_frames", "# of broadcast frames",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_BCAST_L));
SYSCTL_ADD_T4_REG64(pi, "tx_mcast_frames", "# of multicast frames",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_MCAST_L));
SYSCTL_ADD_T4_REG64(pi, "tx_ucast_frames", "# of unicast frames",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_UCAST_L));
SYSCTL_ADD_T4_REG64(pi, "tx_error_frames", "# of error frames",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_ERROR_L));
SYSCTL_ADD_T4_REG64(pi, "tx_frames_64",
"# of tx frames in this range",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_64B_L));
SYSCTL_ADD_T4_REG64(pi, "tx_frames_65_127",
"# of tx frames in this range",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_65B_127B_L));
SYSCTL_ADD_T4_REG64(pi, "tx_frames_128_255",
"# of tx frames in this range",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_128B_255B_L));
SYSCTL_ADD_T4_REG64(pi, "tx_frames_256_511",
"# of tx frames in this range",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_256B_511B_L));
SYSCTL_ADD_T4_REG64(pi, "tx_frames_512_1023",
"# of tx frames in this range",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_512B_1023B_L));
SYSCTL_ADD_T4_REG64(pi, "tx_frames_1024_1518",
"# of tx frames in this range",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_1024B_1518B_L));
SYSCTL_ADD_T4_REG64(pi, "tx_frames_1519_max",
"# of tx frames in this range",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_1519B_MAX_L));
SYSCTL_ADD_T4_REG64(pi, "tx_drop", "# of dropped tx frames",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_DROP_L));
SYSCTL_ADD_T4_REG64(pi, "tx_pause", "# of pause frames transmitted",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_PAUSE_L));
SYSCTL_ADD_T4_REG64(pi, "tx_ppp0", "# of PPP prio 0 frames transmitted",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_PPP0_L));
SYSCTL_ADD_T4_REG64(pi, "tx_ppp1", "# of PPP prio 1 frames transmitted",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_PPP1_L));
SYSCTL_ADD_T4_REG64(pi, "tx_ppp2", "# of PPP prio 2 frames transmitted",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_PPP2_L));
SYSCTL_ADD_T4_REG64(pi, "tx_ppp3", "# of PPP prio 3 frames transmitted",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_PPP3_L));
SYSCTL_ADD_T4_REG64(pi, "tx_ppp4", "# of PPP prio 4 frames transmitted",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_PPP4_L));
SYSCTL_ADD_T4_REG64(pi, "tx_ppp5", "# of PPP prio 5 frames transmitted",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_PPP5_L));
SYSCTL_ADD_T4_REG64(pi, "tx_ppp6", "# of PPP prio 6 frames transmitted",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_PPP6_L));
SYSCTL_ADD_T4_REG64(pi, "tx_ppp7", "# of PPP prio 7 frames transmitted",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_PPP7_L));
SYSCTL_ADD_T4_REG64(pi, "rx_octets", "# of octets in good frames",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_BYTES_L));
SYSCTL_ADD_T4_REG64(pi, "rx_frames", "total # of good frames",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_FRAMES_L));
SYSCTL_ADD_T4_REG64(pi, "rx_bcast_frames", "# of broadcast frames",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_BCAST_L));
SYSCTL_ADD_T4_REG64(pi, "rx_mcast_frames", "# of multicast frames",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_MCAST_L));
SYSCTL_ADD_T4_REG64(pi, "rx_ucast_frames", "# of unicast frames",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_UCAST_L));
SYSCTL_ADD_T4_REG64(pi, "rx_too_long", "# of frames exceeding MTU",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_MTU_ERROR_L));
SYSCTL_ADD_T4_REG64(pi, "rx_jabber", "# of jabber frames",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_MTU_CRC_ERROR_L));
SYSCTL_ADD_T4_REG64(pi, "rx_fcs_err",
"# of frames received with bad FCS",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_CRC_ERROR_L));
SYSCTL_ADD_T4_REG64(pi, "rx_len_err",
"# of frames received with length error",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_LEN_ERROR_L));
SYSCTL_ADD_T4_REG64(pi, "rx_symbol_err", "symbol errors",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_SYM_ERROR_L));
SYSCTL_ADD_T4_REG64(pi, "rx_runt", "# of short frames received",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_LESS_64B_L));
SYSCTL_ADD_T4_REG64(pi, "rx_frames_64",
"# of rx frames in this range",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_64B_L));
SYSCTL_ADD_T4_REG64(pi, "rx_frames_65_127",
"# of rx frames in this range",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_65B_127B_L));
SYSCTL_ADD_T4_REG64(pi, "rx_frames_128_255",
"# of rx frames in this range",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_128B_255B_L));
SYSCTL_ADD_T4_REG64(pi, "rx_frames_256_511",
"# of rx frames in this range",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_256B_511B_L));
SYSCTL_ADD_T4_REG64(pi, "rx_frames_512_1023",
"# of rx frames in this range",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_512B_1023B_L));
SYSCTL_ADD_T4_REG64(pi, "rx_frames_1024_1518",
"# of rx frames in this range",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_1024B_1518B_L));
SYSCTL_ADD_T4_REG64(pi, "rx_frames_1519_max",
"# of rx frames in this range",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_1519B_MAX_L));
SYSCTL_ADD_T4_REG64(pi, "rx_pause", "# of pause frames received",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_PAUSE_L));
SYSCTL_ADD_T4_REG64(pi, "rx_ppp0", "# of PPP prio 0 frames received",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_PPP0_L));
SYSCTL_ADD_T4_REG64(pi, "rx_ppp1", "# of PPP prio 1 frames received",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_PPP1_L));
SYSCTL_ADD_T4_REG64(pi, "rx_ppp2", "# of PPP prio 2 frames received",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_PPP2_L));
SYSCTL_ADD_T4_REG64(pi, "rx_ppp3", "# of PPP prio 3 frames received",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_PPP3_L));
SYSCTL_ADD_T4_REG64(pi, "rx_ppp4", "# of PPP prio 4 frames received",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_PPP4_L));
SYSCTL_ADD_T4_REG64(pi, "rx_ppp5", "# of PPP prio 5 frames received",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_PPP5_L));
SYSCTL_ADD_T4_REG64(pi, "rx_ppp6", "# of PPP prio 6 frames received",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_PPP6_L));
SYSCTL_ADD_T4_REG64(pi, "rx_ppp7", "# of PPP prio 7 frames received",
PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_PPP7_L));
#undef SYSCTL_ADD_T4_REG64
#define SYSCTL_ADD_T4_PORTSTAT(name, desc) \
SYSCTL_ADD_UQUAD(ctx, children, OID_AUTO, #name, CTLFLAG_RD, \
&pi->stats.name, desc)
sys/dev: minor spelling fixes. Most affect comments, very few have user-visible effects.
2016-05-03 03:41:25 +00:00
/* We get these from port_stats and they may be stale by up to 1s */
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
SYSCTL_ADD_T4_PORTSTAT(rx_ovflow0,
"# drops due to buffer-group 0 overflows");
SYSCTL_ADD_T4_PORTSTAT(rx_ovflow1,
"# drops due to buffer-group 1 overflows");
SYSCTL_ADD_T4_PORTSTAT(rx_ovflow2,
"# drops due to buffer-group 2 overflows");
SYSCTL_ADD_T4_PORTSTAT(rx_ovflow3,
"# drops due to buffer-group 3 overflows");
SYSCTL_ADD_T4_PORTSTAT(rx_trunc0,
"# of buffer-group 0 truncated packets");
SYSCTL_ADD_T4_PORTSTAT(rx_trunc1,
"# of buffer-group 1 truncated packets");
SYSCTL_ADD_T4_PORTSTAT(rx_trunc2,
"# of buffer-group 2 truncated packets");
SYSCTL_ADD_T4_PORTSTAT(rx_trunc3,
"# of buffer-group 3 truncated packets");
#undef SYSCTL_ADD_T4_PORTSTAT
Support for TLS offload of TOE connections on T6 adapters. The TOE engine in Chelsio T6 adapters supports offloading of TLS encryption and TCP segmentation for offloaded connections. Sockets using TLS are required to use a set of custom socket options to upload RX and TX keys to the NIC and to enable RX processing. Currently these socket options are implemented as TCP options in the vendor specific range. A patched OpenSSL library will be made available in a port / package for use with the TLS TOE support. TOE sockets can either offload both transmit and reception of TLS records or just transmit. TLS offload (both RX and TX) is enabled by setting the dev.t6nex.<x>.tls sysctl to 1 and requires TOE to be enabled on the relevant interface. Transmit offload can be used on any "normal" or TLS TOE socket by using the custom socket option to program a transmit key. This permits most TOE sockets to transparently offload TLS when applications use a patched SSL library (e.g. using LD_LIBRARY_PATH to request use of a patched OpenSSL library). Receive offload can only be used with TOE sockets using the TLS mode. The dev.t6nex.0.toe.tls_rx_ports sysctl can be set to a list of TCP port numbers. Any connection with either a local or remote port number in that list will be created as a TLS socket rather than a plain TOE socket. Note that although this sysctl accepts an arbitrary list of port numbers, the sysctl(8) tool is only able to set sysctl nodes to a single value. A TLS socket will hang without receiving data if used by an application that is not using a patched SSL library. Thus, the tls_rx_ports node should be used with care. For a server mostly concerned with offloading TLS transmit, this node is not needed as plain TOE sockets will fall back to software crypto when using an unpatched SSL library. New per-interface statistics nodes are added giving counts of TLS packets and payload bytes (payload bytes do not include TLS headers or authentication tags/MACs) offloaded via the TOE engine, e.g.: dev.cc.0.stats.rx_tls_octets: 149 dev.cc.0.stats.rx_tls_records: 13 dev.cc.0.stats.tx_tls_octets: 26501823 dev.cc.0.stats.tx_tls_records: 1620 TLS transmit work requests are constructed by a new variant of t4_push_frames() called t4_push_tls_records() in tom/t4_tls.c. TLS transmit work requests require a buffer containing IVs. If the IVs are too large to fit into the work request, a separate buffer is allocated when constructing a work request. This buffer is associated with the transmit descriptor and freed when the descriptor is ACKed by the adapter. Received TLS frames use two new CPL messages. The first message is a CPL_TLS_DATA containing the decryped payload of a single TLS record. The handler places the mbuf containing the received payload on an mbufq in the TOE pcb. The second message is a CPL_RX_TLS_CMP message which includes a copy of the TLS header and indicates if there were any errors. The handler for this message places the TLS header into the socket buffer followed by the saved mbuf with the payload data. Both of these handlers are contained in tom/t4_tls.c. A few routines were exposed from t4_cpl_io.c for use by t4_tls.c including send_rx_credits(), a new send_rx_modulate(), and t4_close_conn(). TLS keys for both transmit and receive are stored in onboard memory in the NIC in the "TLS keys" memory region. In some cases a TLS socket can hang with pending data available in the NIC that is not delivered to the host. As a workaround, TLS sockets are more aggressive about sending CPL_RX_DATA_ACK messages anytime that any data is read from a TLS socket. In addition, a fallback timer will periodically send CPL_RX_DATA_ACK messages to the NIC for connections that are still in the handshake phase. Once the connection has finished the handshake and programmed RX keys via the socket option, the timer is stopped. A new function select_ulp_mode() is used to determine what sub-mode a given TOE socket should use (plain TOE, DDP, or TLS). The existing set_tcpddp_ulp_mode() function has been renamed to set_ulp_mode() and handles initialization of TLS-specific state when necessary in addition to DDP-specific state. Since TLS sockets do not receive individual TCP segments but always receive full TLS records, they can receive more data than is available in the current window (e.g. if a 16k TLS record is received but the socket buffer is itself 16k). To cope with this, just drop the window to 0 when this happens, but track the overage and "eat" the overage as it is read from the socket buffer not opening the window (or adding rx_credits) for the overage bytes. Reviewed by: np (earlier version) Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D14529
2018-03-13 23:05:51 +00:00
SYSCTL_ADD_ULONG(ctx, children, OID_AUTO, "tx_tls_records",
CTLFLAG_RD, &pi->tx_tls_records,
"# of TLS records transmitted");
SYSCTL_ADD_ULONG(ctx, children, OID_AUTO, "tx_tls_octets",
CTLFLAG_RD, &pi->tx_tls_octets,
"# of payload octets in transmitted TLS records");
SYSCTL_ADD_ULONG(ctx, children, OID_AUTO, "rx_tls_records",
CTLFLAG_RD, &pi->rx_tls_records,
"# of TLS records received");
SYSCTL_ADD_ULONG(ctx, children, OID_AUTO, "rx_tls_octets",
CTLFLAG_RD, &pi->rx_tls_octets,
"# of payload octets in received TLS records");
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
Display holdoff timers and packet counts as a list of numbers. MFC after: 1 week
2011-03-09 21:07:09 +00:00
static int
sysctl_int_array(SYSCTL_HANDLER_ARGS)
{
cxgbe(4): Do not call sbuf_trim on an sbuf with a drain function. MFC after: 1 week
2015-03-23 23:06:32 +00:00
int rc, *i, space = 0;
Display holdoff timers and packet counts as a list of numbers. MFC after: 1 week
2011-03-09 21:07:09 +00:00
struct sbuf sb;
Set the SBUF_INCLUDENUL flag in sbuf_new_for_sysctl() so that sysctl strings returned to userland include the nulterm byte. Some uses of sbuf_new_for_sysctl() write binary data rather than strings; clear the SBUF_INCLUDENUL flag after calling sbuf_new_for_sysctl() in those cases. (Note that the sbuf code still automatically adds a nulterm byte in sbuf_finish(), but since it's not included in the length it won't get copied to userland along with the binary data.) Remove explicit adding of a nulterm byte in a couple places now that it gets done automatically by the sbuf drain code. PR: 195668
2015-03-14 17:08:28 +00:00
sbuf_new_for_sysctl(&sb, NULL, 64, req);
cxgbe(4): Do not call sbuf_trim on an sbuf with a drain function. MFC after: 1 week
2015-03-23 23:06:32 +00:00
for (i = arg1; arg2; arg2 -= sizeof(int), i++) {
if (space)
sbuf_printf(&sb, " ");
sbuf_printf(&sb, "%d", *i);
space = 1;
}
Set the SBUF_INCLUDENUL flag in sbuf_new_for_sysctl() so that sysctl strings returned to userland include the nulterm byte. Some uses of sbuf_new_for_sysctl() write binary data rather than strings; clear the SBUF_INCLUDENUL flag after calling sbuf_new_for_sysctl() in those cases. (Note that the sbuf code still automatically adds a nulterm byte in sbuf_finish(), but since it's not included in the length it won't get copied to userland along with the binary data.) Remove explicit adding of a nulterm byte in a couple places now that it gets done automatically by the sbuf drain code. PR: 195668
2015-03-14 17:08:28 +00:00
rc = sbuf_finish(&sb);
Display holdoff timers and packet counts as a list of numbers. MFC after: 1 week
2011-03-09 21:07:09 +00:00
sbuf_delete(&sb);
return (rc);
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
static int
cxgbe(4): Break up sysctl_bitfield into 8 bit and 16 bit variants. Have them display the current value of the bitfield rather than the fixed value that was provided when the sysctl node was created. MFC after: 1 week Sponsored by: Chelsio Communications
2018-08-06 21:54:51 +00:00
sysctl_bitfield_8b(SYSCTL_HANDLER_ARGS)
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
{
int rc;
struct sbuf *sb;
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return(rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 128, req);
if (sb == NULL)
return (ENOMEM);
cxgbe(4): Break up sysctl_bitfield into 8 bit and 16 bit variants. Have them display the current value of the bitfield rather than the fixed value that was provided when the sysctl node was created. MFC after: 1 week Sponsored by: Chelsio Communications
2018-08-06 21:54:51 +00:00
sbuf_printf(sb, "%b", *(uint8_t *)(uintptr_t)arg2, (char *)arg1);
rc = sbuf_finish(sb);
sbuf_delete(sb);
return (rc);
}
static int
sysctl_bitfield_16b(SYSCTL_HANDLER_ARGS)
{
int rc;
struct sbuf *sb;
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return(rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 128, req);
if (sb == NULL)
return (ENOMEM);
sbuf_printf(sb, "%b", *(uint16_t *)(uintptr_t)arg2, (char *)arg1);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
rc = sbuf_finish(sb);
sbuf_delete(sb);
return (rc);
}
- Show the reason why link is down if this information is available. - Display the temperature and PHY firmware version of the BT PHY. MFC after: 1 day
2013-07-05 01:53:51 +00:00
static int
sysctl_btphy(SYSCTL_HANDLER_ARGS)
{
struct port_info *pi = arg1;
int op = arg2;
struct adapter *sc = pi->adapter;
u_int v;
int rc;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
rc = begin_synchronized_op(sc, &pi->vi[0], SLEEP_OK | INTR_OK, "t4btt");
- Show the reason why link is down if this information is available. - Display the temperature and PHY firmware version of the BT PHY. MFC after: 1 day
2013-07-05 01:53:51 +00:00
if (rc)
return (rc);
/* XXX: magic numbers */
rc = -t4_mdio_rd(sc, sc->mbox, pi->mdio_addr, 0x1e, op ? 0x20 : 0xc820,
&v);
end_synchronized_op(sc, 0);
if (rc)
return (rc);
if (op == 0)
v /= 256;
rc = sysctl_handle_int(oidp, &v, 0, req);
return (rc);
}
Add a new sysctl, dev.cxgbe.N.rsrv_noflow, and a companion tunable, hw.cxgbe.rsrv_noflow. When set, queue 0 of the port is reserved for TX packets without a flowid. The hash value of packets with a flowid is bumped up by 1. The intent is to provide a private queue for link-level packets like LACP that is unlikely to overflow or suffer deep queue latency. Reviewed by: np Obtained from: Netflix MFC after: 3 days
2014-02-06 18:40:38 +00:00
static int
sysctl_noflowq(SYSCTL_HANDLER_ARGS)
{
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct vi_info *vi = arg1;
Add a new sysctl, dev.cxgbe.N.rsrv_noflow, and a companion tunable, hw.cxgbe.rsrv_noflow. When set, queue 0 of the port is reserved for TX packets without a flowid. The hash value of packets with a flowid is bumped up by 1. The intent is to provide a private queue for link-level packets like LACP that is unlikely to overflow or suffer deep queue latency. Reviewed by: np Obtained from: Netflix MFC after: 3 days
2014-02-06 18:40:38 +00:00
int rc, val;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
val = vi->rsrv_noflowq;
Add a new sysctl, dev.cxgbe.N.rsrv_noflow, and a companion tunable, hw.cxgbe.rsrv_noflow. When set, queue 0 of the port is reserved for TX packets without a flowid. The hash value of packets with a flowid is bumped up by 1. The intent is to provide a private queue for link-level packets like LACP that is unlikely to overflow or suffer deep queue latency. Reviewed by: np Obtained from: Netflix MFC after: 3 days
2014-02-06 18:40:38 +00:00
rc = sysctl_handle_int(oidp, &val, 0, req);
if (rc != 0 || req->newptr == NULL)
return (rc);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if ((val >= 1) && (vi->ntxq > 1))
vi->rsrv_noflowq = 1;
Add a new sysctl, dev.cxgbe.N.rsrv_noflow, and a companion tunable, hw.cxgbe.rsrv_noflow. When set, queue 0 of the port is reserved for TX packets without a flowid. The hash value of packets with a flowid is bumped up by 1. The intent is to provide a private queue for link-level packets like LACP that is unlikely to overflow or suffer deep queue latency. Reviewed by: np Obtained from: Netflix MFC after: 3 days
2014-02-06 18:40:38 +00:00
else
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
vi->rsrv_noflowq = 0;
Add a new sysctl, dev.cxgbe.N.rsrv_noflow, and a companion tunable, hw.cxgbe.rsrv_noflow. When set, queue 0 of the port is reserved for TX packets without a flowid. The hash value of packets with a flowid is bumped up by 1. The intent is to provide a private queue for link-level packets like LACP that is unlikely to overflow or suffer deep queue latency. Reviewed by: np Obtained from: Netflix MFC after: 3 days
2014-02-06 18:40:38 +00:00
return (rc);
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
static int
sysctl_holdoff_tmr_idx(SYSCTL_HANDLER_ARGS)
{
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct vi_info *vi = arg1;
struct adapter *sc = vi->pi->adapter;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
int idx, rc, i;
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
struct sge_rxq *rxq;
uint8_t v;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
idx = vi->tmr_idx;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
rc = sysctl_handle_int(oidp, &idx, 0, req);
if (rc != 0 || req->newptr == NULL)
return (rc);
if (idx < 0 || idx >= SGE_NTIMERS)
return (EINVAL);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
rc = begin_synchronized_op(sc, vi, HOLD_LOCK | SLEEP_OK | INTR_OK,
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
"t4tmr");
if (rc)
return (rc);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
v = V_QINTR_TIMER_IDX(idx) | V_QINTR_CNT_EN(vi->pktc_idx != -1);
for_each_rxq(vi, i, rxq) {
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
#ifdef atomic_store_rel_8
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
atomic_store_rel_8(&rxq->iq.intr_params, v);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
#else
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
rxq->iq.intr_params = v;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
#endif
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
vi->tmr_idx = idx;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
end_synchronized_op(sc, LOCK_HELD);
return (0);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
static int
sysctl_holdoff_pktc_idx(SYSCTL_HANDLER_ARGS)
{
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct vi_info *vi = arg1;
struct adapter *sc = vi->pi->adapter;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
int idx, rc;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
idx = vi->pktc_idx;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
rc = sysctl_handle_int(oidp, &idx, 0, req);
if (rc != 0 || req->newptr == NULL)
return (rc);
if (idx < -1 || idx >= SGE_NCOUNTERS)
return (EINVAL);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
rc = begin_synchronized_op(sc, vi, HOLD_LOCK | SLEEP_OK | INTR_OK,
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
"t4pktc");
if (rc)
return (rc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (vi->flags & VI_INIT_DONE)
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
rc = EBUSY; /* cannot be changed once the queues are created */
else
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
vi->pktc_idx = idx;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
end_synchronized_op(sc, LOCK_HELD);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
return (rc);
}
static int
sysctl_qsize_rxq(SYSCTL_HANDLER_ARGS)
{
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct vi_info *vi = arg1;
struct adapter *sc = vi->pi->adapter;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
int qsize, rc;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
qsize = vi->qsize_rxq;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
rc = sysctl_handle_int(oidp, &qsize, 0, req);
if (rc != 0 || req->newptr == NULL)
return (rc);
if (qsize < 128 || (qsize & 7))
return (EINVAL);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
rc = begin_synchronized_op(sc, vi, HOLD_LOCK | SLEEP_OK | INTR_OK,
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
"t4rxqs");
if (rc)
return (rc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (vi->flags & VI_INIT_DONE)
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
rc = EBUSY; /* cannot be changed once the queues are created */
else
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
vi->qsize_rxq = qsize;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
end_synchronized_op(sc, LOCK_HELD);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
return (rc);
}
static int
sysctl_qsize_txq(SYSCTL_HANDLER_ARGS)
{
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct vi_info *vi = arg1;
struct adapter *sc = vi->pi->adapter;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
int qsize, rc;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
qsize = vi->qsize_txq;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
rc = sysctl_handle_int(oidp, &qsize, 0, req);
if (rc != 0 || req->newptr == NULL)
return (rc);
cxgbe(4): remove buf_ring specific restriction on the txq size. MFC after: 2 months
2015-01-01 09:33:46 +00:00
if (qsize < 128 || qsize > 65536)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
return (EINVAL);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
rc = begin_synchronized_op(sc, vi, HOLD_LOCK | SLEEP_OK | INTR_OK,
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
"t4txqs");
if (rc)
return (rc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (vi->flags & VI_INIT_DONE)
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
rc = EBUSY; /* cannot be changed once the queues are created */
else
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
vi->qsize_txq = qsize;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
end_synchronized_op(sc, LOCK_HELD);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
return (rc);
}
cxgbe(4): knobs to enable/disable PAUSE frame based flow control. MFC after: 1 week
2014-09-12 05:25:56 +00:00
static int
sysctl_pause_settings(SYSCTL_HANDLER_ARGS)
{
struct port_info *pi = arg1;
struct adapter *sc = pi->adapter;
struct link_config *lc = &pi->link_cfg;
int rc;
if (req->newptr == NULL) {
struct sbuf *sb;
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
static char *bits = "\20\1RX\2TX\3AUTO";
cxgbe(4): knobs to enable/disable PAUSE frame based flow control. MFC after: 1 week
2014-09-12 05:25:56 +00:00
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return(rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 128, req);
if (sb == NULL)
return (ENOMEM);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
if (lc->link_ok) {
sbuf_printf(sb, "%b", (lc->fc & (PAUSE_TX | PAUSE_RX)) |
(lc->requested_fc & PAUSE_AUTONEG), bits);
} else {
sbuf_printf(sb, "%b", lc->requested_fc & (PAUSE_TX |
PAUSE_RX | PAUSE_AUTONEG), bits);
}
cxgbe(4): knobs to enable/disable PAUSE frame based flow control. MFC after: 1 week
2014-09-12 05:25:56 +00:00
rc = sbuf_finish(sb);
sbuf_delete(sb);
} else {
char s[2];
int n;
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
s[0] = '0' + (lc->requested_fc & (PAUSE_TX | PAUSE_RX |
PAUSE_AUTONEG));
cxgbe(4): knobs to enable/disable PAUSE frame based flow control. MFC after: 1 week
2014-09-12 05:25:56 +00:00
s[1] = 0;
rc = sysctl_handle_string(oidp, s, sizeof(s), req);
if (rc != 0)
return(rc);
if (s[1] != 0)
return (EINVAL);
if (s[0] < '0' || s[0] > '9')
return (EINVAL); /* not a number */
n = s[0] - '0';
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
if (n & ~(PAUSE_TX | PAUSE_RX | PAUSE_AUTONEG))
cxgbe(4): knobs to enable/disable PAUSE frame based flow control. MFC after: 1 week
2014-09-12 05:25:56 +00:00
return (EINVAL); /* some other bit is set too */
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
rc = begin_synchronized_op(sc, &pi->vi[0], SLEEP_OK | INTR_OK,
"t4PAUSE");
cxgbe(4): knobs to enable/disable PAUSE frame based flow control. MFC after: 1 week
2014-09-12 05:25:56 +00:00
if (rc)
return (rc);
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
PORT_LOCK(pi);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
lc->requested_fc = n;
fixup_link_config(pi);
if (pi->up_vis > 0)
rc = apply_link_config(pi);
set_current_media(pi);
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
PORT_UNLOCK(pi);
cxgbe(4): knobs to enable/disable PAUSE frame based flow control. MFC after: 1 week
2014-09-12 05:25:56 +00:00
end_synchronized_op(sc, 0);
}
return (rc);
}
cxgbe(4): Updates to link configuration. - Update struct link_settings and associated shared code. - Add tunables to control FEC and autonegotiation. All ports inherit these values as their initial settings. hw.cxgbe.fec hw.cxgbe.autoneg - Add per-port sysctls to control FEC and autonegotiation. These can be modified at any time. dev.<port>.<n>.fec dev.<port>.<n>.autoneg MFC after: 3 days Sponsored by: Chelsio Communications
2016-12-30 08:59:49 +00:00
static int
sysctl_fec(SYSCTL_HANDLER_ARGS)
{
struct port_info *pi = arg1;
struct adapter *sc = pi->adapter;
struct link_config *lc = &pi->link_cfg;
int rc;
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
int8_t old;
cxgbe(4): Updates to link configuration. - Update struct link_settings and associated shared code. - Add tunables to control FEC and autonegotiation. All ports inherit these values as their initial settings. hw.cxgbe.fec hw.cxgbe.autoneg - Add per-port sysctls to control FEC and autonegotiation. These can be modified at any time. dev.<port>.<n>.fec dev.<port>.<n>.autoneg MFC after: 3 days Sponsored by: Chelsio Communications
2016-12-30 08:59:49 +00:00
if (req->newptr == NULL) {
struct sbuf *sb;
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
static char *bits = "\20\1RS\2BASE-R\3RSVD1\4RSVD2\5RSVD3\6AUTO";
cxgbe(4): Updates to link configuration. - Update struct link_settings and associated shared code. - Add tunables to control FEC and autonegotiation. All ports inherit these values as their initial settings. hw.cxgbe.fec hw.cxgbe.autoneg - Add per-port sysctls to control FEC and autonegotiation. These can be modified at any time. dev.<port>.<n>.fec dev.<port>.<n>.autoneg MFC after: 3 days Sponsored by: Chelsio Communications
2016-12-30 08:59:49 +00:00
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return(rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 128, req);
if (sb == NULL)
return (ENOMEM);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
/*
* Display the requested_fec when the link is down -- the actual
* FEC makes sense only when the link is up.
*/
if (lc->link_ok) {
sbuf_printf(sb, "%b", (lc->fec & M_FW_PORT_CAP32_FEC) |
(lc->requested_fec & FEC_AUTO), bits);
} else {
sbuf_printf(sb, "%b", lc->requested_fec, bits);
}
cxgbe(4): Updates to link configuration. - Update struct link_settings and associated shared code. - Add tunables to control FEC and autonegotiation. All ports inherit these values as their initial settings. hw.cxgbe.fec hw.cxgbe.autoneg - Add per-port sysctls to control FEC and autonegotiation. These can be modified at any time. dev.<port>.<n>.fec dev.<port>.<n>.autoneg MFC after: 3 days Sponsored by: Chelsio Communications
2016-12-30 08:59:49 +00:00
rc = sbuf_finish(sb);
sbuf_delete(sb);
} else {
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
char s[3];
cxgbe(4): Updates to link configuration. - Update struct link_settings and associated shared code. - Add tunables to control FEC and autonegotiation. All ports inherit these values as their initial settings. hw.cxgbe.fec hw.cxgbe.autoneg - Add per-port sysctls to control FEC and autonegotiation. These can be modified at any time. dev.<port>.<n>.fec dev.<port>.<n>.autoneg MFC after: 3 days Sponsored by: Chelsio Communications
2016-12-30 08:59:49 +00:00
int n;
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
snprintf(s, sizeof(s), "%d",
lc->requested_fec == FEC_AUTO ? -1 :
lc->requested_fec & M_FW_PORT_CAP32_FEC);
cxgbe(4): Updates to link configuration. - Update struct link_settings and associated shared code. - Add tunables to control FEC and autonegotiation. All ports inherit these values as their initial settings. hw.cxgbe.fec hw.cxgbe.autoneg - Add per-port sysctls to control FEC and autonegotiation. These can be modified at any time. dev.<port>.<n>.fec dev.<port>.<n>.autoneg MFC after: 3 days Sponsored by: Chelsio Communications
2016-12-30 08:59:49 +00:00
rc = sysctl_handle_string(oidp, s, sizeof(s), req);
if (rc != 0)
return(rc);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
n = strtol(&s[0], NULL, 0);
if (n < 0 || n & FEC_AUTO)
n = FEC_AUTO;
else {
if (n & ~M_FW_PORT_CAP32_FEC)
return (EINVAL);/* some other bit is set too */
if (!powerof2(n))
return (EINVAL);/* one bit can be set at most */
}
cxgbe(4): Updates to link configuration. - Update struct link_settings and associated shared code. - Add tunables to control FEC and autonegotiation. All ports inherit these values as their initial settings. hw.cxgbe.fec hw.cxgbe.autoneg - Add per-port sysctls to control FEC and autonegotiation. These can be modified at any time. dev.<port>.<n>.fec dev.<port>.<n>.autoneg MFC after: 3 days Sponsored by: Chelsio Communications
2016-12-30 08:59:49 +00:00
rc = begin_synchronized_op(sc, &pi->vi[0], SLEEP_OK | INTR_OK,
"t4fec");
if (rc)
return (rc);
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
PORT_LOCK(pi);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
old = lc->requested_fec;
if (n == FEC_AUTO)
lc->requested_fec = FEC_AUTO;
else if (n == 0)
lc->requested_fec = FEC_NONE;
else {
if ((lc->supported | V_FW_PORT_CAP32_FEC(n)) !=
lc->supported) {
rc = ENOTSUP;
goto done;
cxgbe(4): Various link/media related improvements. - Deal with changes to port_type, and not just port_mod when a transceiver is changed. This fixes hot swapping of transceivers of different types (QSFP+ or QSA or QSFP28 in a QSFP28 port, SFP+ or SFP28 in a SFP28 port, etc.). - Always refresh media information for ifconfig if the port is down. The firmware does not generate tranceiver-change interrupts unless at least one VI is enabled on the physical port. Before this change ifconfig diplayed potentially stale information for ports that were administratively down. - Always recalculate and reapply L1 config on a transceiver change. - Display PAUSE settings in ifconfig. The driver sysctls for this continue to work as well. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-17 00:42:13 +00:00
}
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
lc->requested_fec = n;
cxgbe(4): Updates to link configuration. - Update struct link_settings and associated shared code. - Add tunables to control FEC and autonegotiation. All ports inherit these values as their initial settings. hw.cxgbe.fec hw.cxgbe.autoneg - Add per-port sysctls to control FEC and autonegotiation. These can be modified at any time. dev.<port>.<n>.fec dev.<port>.<n>.autoneg MFC after: 3 days Sponsored by: Chelsio Communications
2016-12-30 08:59:49 +00:00
}
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
fixup_link_config(pi);
if (pi->up_vis > 0) {
rc = apply_link_config(pi);
if (rc != 0) {
lc->requested_fec = old;
if (rc == FW_EPROTO)
rc = ENOTSUP;
}
}
done:
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
PORT_UNLOCK(pi);
cxgbe(4): Updates to link configuration. - Update struct link_settings and associated shared code. - Add tunables to control FEC and autonegotiation. All ports inherit these values as their initial settings. hw.cxgbe.fec hw.cxgbe.autoneg - Add per-port sysctls to control FEC and autonegotiation. These can be modified at any time. dev.<port>.<n>.fec dev.<port>.<n>.autoneg MFC after: 3 days Sponsored by: Chelsio Communications
2016-12-30 08:59:49 +00:00
end_synchronized_op(sc, 0);
}
return (rc);
}
static int
sysctl_autoneg(SYSCTL_HANDLER_ARGS)
{
struct port_info *pi = arg1;
struct adapter *sc = pi->adapter;
struct link_config *lc = &pi->link_cfg;
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
int rc, val;
cxgbe(4): Updates to link configuration. - Update struct link_settings and associated shared code. - Add tunables to control FEC and autonegotiation. All ports inherit these values as their initial settings. hw.cxgbe.fec hw.cxgbe.autoneg - Add per-port sysctls to control FEC and autonegotiation. These can be modified at any time. dev.<port>.<n>.fec dev.<port>.<n>.autoneg MFC after: 3 days Sponsored by: Chelsio Communications
2016-12-30 08:59:49 +00:00
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
if (lc->supported & FW_PORT_CAP32_ANEG)
val = lc->requested_aneg == AUTONEG_DISABLE ? 0 : 1;
cxgbe(4): Updates to link configuration. - Update struct link_settings and associated shared code. - Add tunables to control FEC and autonegotiation. All ports inherit these values as their initial settings. hw.cxgbe.fec hw.cxgbe.autoneg - Add per-port sysctls to control FEC and autonegotiation. These can be modified at any time. dev.<port>.<n>.fec dev.<port>.<n>.autoneg MFC after: 3 days Sponsored by: Chelsio Communications
2016-12-30 08:59:49 +00:00
else
val = -1;
rc = sysctl_handle_int(oidp, &val, 0, req);
if (rc != 0 || req->newptr == NULL)
return (rc);
cxgbe(4): Fixes related to the knob that controls link autonegotiation. - Do not leak the adapter lock in sysctl_autoneg. - Accept only 0 or 1 as valid settings for autonegotiation. - A fixed speed must be requested by the driver when autonegotiation is disabled otherwise the firmware will reject the l1cfg command. Use the top speed supported by the port for now. MFC after: 3 days Sponsored by: Chelsio Communications
2017-05-09 08:08:28 +00:00
if (val == 0)
val = AUTONEG_DISABLE;
else if (val == 1)
val = AUTONEG_ENABLE;
else
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
val = AUTONEG_AUTO;
cxgbe(4): Updates to link configuration. - Update struct link_settings and associated shared code. - Add tunables to control FEC and autonegotiation. All ports inherit these values as their initial settings. hw.cxgbe.fec hw.cxgbe.autoneg - Add per-port sysctls to control FEC and autonegotiation. These can be modified at any time. dev.<port>.<n>.fec dev.<port>.<n>.autoneg MFC after: 3 days Sponsored by: Chelsio Communications
2016-12-30 08:59:49 +00:00
rc = begin_synchronized_op(sc, &pi->vi[0], SLEEP_OK | INTR_OK,
"t4aneg");
if (rc)
return (rc);
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
PORT_LOCK(pi);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
if (val == AUTONEG_ENABLE && !(lc->supported & FW_PORT_CAP32_ANEG)) {
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
rc = ENOTSUP;
goto done;
}
cxgbe(4): Various link/media related improvements. - Deal with changes to port_type, and not just port_mod when a transceiver is changed. This fixes hot swapping of transceivers of different types (QSFP+ or QSA or QSFP28 in a QSFP28 port, SFP+ or SFP28 in a SFP28 port, etc.). - Always refresh media information for ifconfig if the port is down. The firmware does not generate tranceiver-change interrupts unless at least one VI is enabled on the physical port. Before this change ifconfig diplayed potentially stale information for ports that were administratively down. - Always recalculate and reapply L1 config on a transceiver change. - Display PAUSE settings in ifconfig. The driver sysctls for this continue to work as well. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-17 00:42:13 +00:00
lc->requested_aneg = val;
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
fixup_link_config(pi);
if (pi->up_vis > 0)
rc = apply_link_config(pi);
set_current_media(pi);
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
done:
PORT_UNLOCK(pi);
cxgbe(4): Fixes related to the knob that controls link autonegotiation. - Do not leak the adapter lock in sysctl_autoneg. - Accept only 0 or 1 as valid settings for autonegotiation. - A fixed speed must be requested by the driver when autonegotiation is disabled otherwise the firmware will reject the l1cfg command. Use the top speed supported by the port for now. MFC after: 3 days Sponsored by: Chelsio Communications
2017-05-09 08:08:28 +00:00
end_synchronized_op(sc, 0);
cxgbe(4): Updates to link configuration. - Update struct link_settings and associated shared code. - Add tunables to control FEC and autonegotiation. All ports inherit these values as their initial settings. hw.cxgbe.fec hw.cxgbe.autoneg - Add per-port sysctls to control FEC and autonegotiation. These can be modified at any time. dev.<port>.<n>.fec dev.<port>.<n>.autoneg MFC after: 3 days Sponsored by: Chelsio Communications
2016-12-30 08:59:49 +00:00
return (rc);
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
static int
sysctl_handle_t4_reg64(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
int reg = arg2;
uint64_t val;
val = t4_read_reg64(sc, reg);
return (sysctl_handle_64(oidp, &val, 0, req));
}
Display temperature sensor data. Shows -1 if sensor not available on the card. # sysctl dev.t4nex.0.temperature # sysctl dev.t5nex.0.temperature
2013-08-02 18:05:42 +00:00
static int
sysctl_temperature(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
int rc, t;
uint32_t param, val;
rc = begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK, "t4temp");
if (rc)
return (rc);
param = V_FW_PARAMS_MNEM(FW_PARAMS_MNEM_DEV) |
V_FW_PARAMS_PARAM_X(FW_PARAMS_PARAM_DEV_DIAG) |
V_FW_PARAMS_PARAM_Y(FW_PARAM_DEV_DIAG_TMP);
rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 1, &param, &val);
end_synchronized_op(sc, 0);
if (rc)
return (rc);
/* unknown is returned as 0 but we display -1 in that case */
t = val == 0 ? -1 : val;
rc = sysctl_handle_int(oidp, &t, 0, req);
return (rc);
}
cxgbe(4): Add a sysctl to report the chip's microprocessor's load averages. This works with debug or custom firmwares only. sysctl dev.<nexus>.<instance>.loadavg sysctl dev.t6nex.0.loadavg MFC after: 1 month Sponsored by: Chelsio Communications
2018-07-10 03:03:10 +00:00
static int
sysctl_loadavg(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
struct sbuf *sb;
int rc;
uint32_t param, val;
rc = begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK, "t4lavg");
if (rc)
return (rc);
param = V_FW_PARAMS_MNEM(FW_PARAMS_MNEM_DEV) |
V_FW_PARAMS_PARAM_X(FW_PARAMS_PARAM_DEV_LOAD);
rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 1, &param, &val);
end_synchronized_op(sc, 0);
if (rc)
return (rc);
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req);
if (sb == NULL)
return (ENOMEM);
if (val == 0xffffffff) {
/* Only debug and custom firmwares report load averages. */
sbuf_printf(sb, "not available");
} else {
sbuf_printf(sb, "%d %d %d", val & 0xff, (val >> 8) & 0xff,
(val >> 16) & 0xff);
}
rc = sbuf_finish(sb);
sbuf_delete(sb);
return (rc);
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
static int
sysctl_cctrl(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
struct sbuf *sb;
int rc, i;
uint16_t incr[NMTUS][NCCTRL_WIN];
static const char *dec_fac[] = {
"0.5", "0.5625", "0.625", "0.6875", "0.75", "0.8125", "0.875",
"0.9375"
};
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req);
if (sb == NULL)
return (ENOMEM);
t4_read_cong_tbl(sc, incr);
for (i = 0; i < NCCTRL_WIN; ++i) {
sbuf_printf(sb, "%2d: %4u %4u %4u %4u %4u %4u %4u %4u\n", i,
incr[0][i], incr[1][i], incr[2][i], incr[3][i], incr[4][i],
incr[5][i], incr[6][i], incr[7][i]);
sbuf_printf(sb, "%8u %4u %4u %4u %4u %4u %4u %4u %5u %s\n",
incr[8][i], incr[9][i], incr[10][i], incr[11][i],
incr[12][i], incr[13][i], incr[14][i], incr[15][i],
sc->params.a_wnd[i], dec_fac[sc->params.b_wnd[i]]);
}
rc = sbuf_finish(sb);
sbuf_delete(sb);
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
return (rc);
}
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
static const char *qname[CIM_NUM_IBQ + CIM_NUM_OBQ_T5] = {
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
"TP0", "TP1", "ULP", "SGE0", "SGE1", "NC-SI", /* ibq's */
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
"ULP0", "ULP1", "ULP2", "ULP3", "SGE", "NC-SI", /* obq's */
"SGE0-RX", "SGE1-RX" /* additional obq's (T5 onwards) */
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
};
static int
sysctl_cim_ibq_obq(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
struct sbuf *sb;
int rc, i, n, qid = arg2;
uint32_t *buf, *p;
char *qtype;
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
u_int cim_num_obq = sc->chip_params->cim_num_obq;
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
KASSERT(qid >= 0 && qid < CIM_NUM_IBQ + cim_num_obq,
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
("%s: bad qid %d\n", __func__, qid));
if (qid < CIM_NUM_IBQ) {
/* inbound queue */
qtype = "IBQ";
n = 4 * CIM_IBQ_SIZE;
buf = malloc(n * sizeof(uint32_t), M_CXGBE, M_ZERO | M_WAITOK);
rc = t4_read_cim_ibq(sc, qid, buf, n);
} else {
/* outbound queue */
qtype = "OBQ";
qid -= CIM_NUM_IBQ;
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
n = 4 * cim_num_obq * CIM_OBQ_SIZE;
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
buf = malloc(n * sizeof(uint32_t), M_CXGBE, M_ZERO | M_WAITOK);
rc = t4_read_cim_obq(sc, qid, buf, n);
}
if (rc < 0) {
rc = -rc;
goto done;
}
n = rc * sizeof(uint32_t); /* rc has # of words actually read */
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
goto done;
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
sb = sbuf_new_for_sysctl(NULL, NULL, PAGE_SIZE, req);
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
if (sb == NULL) {
rc = ENOMEM;
goto done;
}
sbuf_printf(sb, "%s%d %s", qtype , qid, qname[arg2]);
for (i = 0, p = buf; i < n; i += 16, p += 4)
sbuf_printf(sb, "\n%#06x: %08x %08x %08x %08x", i, p[0], p[1],
p[2], p[3]);
rc = sbuf_finish(sb);
sbuf_delete(sb);
done:
free(buf, M_CXGBE);
return (rc);
}
cxgbe(4): Auto-dump the CIM block's logic analyzer on a TIMER0 interrupt. Sponsored by: Chelsio Communications
2019-02-07 05:40:51 +00:00
static void
sbuf_cim_la4(struct adapter *sc, struct sbuf *sb, uint32_t *buf, uint32_t cfg)
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
{
cxgbe(4): Auto-dump the CIM block's logic analyzer on a TIMER0 interrupt. Sponsored by: Chelsio Communications
2019-02-07 05:40:51 +00:00
uint32_t *p;
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
sbuf_printf(sb, "Status Data PC%s",
cfg & F_UPDBGLACAPTPCONLY ? "" :
" LS0Stat LS0Addr LS0Data");
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
for (p = buf; p <= &buf[sc->params.cim_la_size - 8]; p += 8) {
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
if (cfg & F_UPDBGLACAPTPCONLY) {
sbuf_printf(sb, "\n %02x %08x %08x", p[5] & 0xff,
p[6], p[7]);
sbuf_printf(sb, "\n %02x %02x%06x %02x%06x",
(p[3] >> 8) & 0xff, p[3] & 0xff, p[4] >> 8,
p[4] & 0xff, p[5] >> 8);
sbuf_printf(sb, "\n %02x %x%07x %x%07x",
(p[0] >> 4) & 0xff, p[0] & 0xf, p[1] >> 4,
p[1] & 0xf, p[2] >> 4);
} else {
sbuf_printf(sb,
"\n %02x %x%07x %x%07x %08x %08x "
"%08x%08x%08x%08x",
(p[0] >> 4) & 0xff, p[0] & 0xf, p[1] >> 4,
p[1] & 0xf, p[2] >> 4, p[2] & 0xf, p[3], p[4], p[5],
p[6], p[7]);
}
}
}
cxgbe(4): Auto-dump the CIM block's logic analyzer on a TIMER0 interrupt. Sponsored by: Chelsio Communications
2019-02-07 05:40:51 +00:00
static void
sbuf_cim_la6(struct adapter *sc, struct sbuf *sb, uint32_t *buf, uint32_t cfg)
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
{
cxgbe(4): Auto-dump the CIM block's logic analyzer on a TIMER0 interrupt. Sponsored by: Chelsio Communications
2019-02-07 05:40:51 +00:00
uint32_t *p;
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
sbuf_printf(sb, "Status Inst Data PC%s",
cfg & F_UPDBGLACAPTPCONLY ? "" :
" LS0Stat LS0Addr LS0Data LS1Stat LS1Addr LS1Data");
for (p = buf; p <= &buf[sc->params.cim_la_size - 10]; p += 10) {
if (cfg & F_UPDBGLACAPTPCONLY) {
sbuf_printf(sb, "\n %02x %08x %08x %08x",
p[3] & 0xff, p[2], p[1], p[0]);
sbuf_printf(sb, "\n %02x %02x%06x %02x%06x %02x%06x",
(p[6] >> 8) & 0xff, p[6] & 0xff, p[5] >> 8,
p[5] & 0xff, p[4] >> 8, p[4] & 0xff, p[3] >> 8);
sbuf_printf(sb, "\n %02x %04x%04x %04x%04x %04x%04x",
(p[9] >> 16) & 0xff, p[9] & 0xffff, p[8] >> 16,
p[8] & 0xffff, p[7] >> 16, p[7] & 0xffff,
p[6] >> 16);
} else {
sbuf_printf(sb, "\n %02x %04x%04x %04x%04x %04x%04x "
"%08x %08x %08x %08x %08x %08x",
(p[9] >> 16) & 0xff,
p[9] & 0xffff, p[8] >> 16,
p[8] & 0xffff, p[7] >> 16,
p[7] & 0xffff, p[6] >> 16,
p[2], p[1], p[0], p[5], p[4], p[3]);
}
}
cxgbe(4): Auto-dump the CIM block's logic analyzer on a TIMER0 interrupt. Sponsored by: Chelsio Communications
2019-02-07 05:40:51 +00:00
}
static int
sbuf_cim_la(struct adapter *sc, struct sbuf *sb, int flags)
{
uint32_t cfg, *buf;
int rc;
rc = -t4_cim_read(sc, A_UP_UP_DBG_LA_CFG, 1, &cfg);
if (rc != 0)
return (rc);
MPASS(flags == M_WAITOK || flags == M_NOWAIT);
buf = malloc(sc->params.cim_la_size * sizeof(uint32_t), M_CXGBE,
M_ZERO | flags);
if (buf == NULL)
return (ENOMEM);
rc = -t4_cim_read_la(sc, buf, NULL);
if (rc != 0)
goto done;
if (chip_id(sc) < CHELSIO_T6)
sbuf_cim_la4(sc, sb, buf, cfg);
else
sbuf_cim_la6(sc, sb, buf, cfg);
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
done:
free(buf, M_CXGBE);
return (rc);
}
cxgbe(4): Auto-dump the CIM block's logic analyzer on a TIMER0 interrupt. Sponsored by: Chelsio Communications
2019-02-07 05:40:51 +00:00
static int
sysctl_cim_la(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
struct sbuf *sb;
int rc;
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req);
if (sb == NULL)
return (ENOMEM);
rc = sbuf_cim_la(sc, sb, M_WAITOK);
if (rc == 0)
rc = sbuf_finish(sb);
sbuf_delete(sb);
return (rc);
}
bool
t4_os_dump_cimla(struct adapter *sc, int arg, bool verbose)
{
struct sbuf sb;
int rc;
if (sbuf_new(&sb, NULL, 4096, SBUF_AUTOEXTEND) != &sb)
return (false);
rc = sbuf_cim_la(sc, &sb, M_NOWAIT);
if (rc == 0) {
rc = sbuf_finish(&sb);
if (rc == 0) {
log(LOG_DEBUG, "%s: CIM LA dump follows.\n%s",
device_get_nameunit(sc->dev), sbuf_data(&sb));
}
}
sbuf_delete(&sb);
return (false);
}
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
static int
sysctl_cim_ma_la(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
u_int i;
struct sbuf *sb;
uint32_t *buf, *p;
int rc;
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req);
if (sb == NULL)
return (ENOMEM);
buf = malloc(2 * CIM_MALA_SIZE * 5 * sizeof(uint32_t), M_CXGBE,
M_ZERO | M_WAITOK);
t4_cim_read_ma_la(sc, buf, buf + 5 * CIM_MALA_SIZE);
p = buf;
for (i = 0; i < CIM_MALA_SIZE; i++, p += 5) {
sbuf_printf(sb, "\n%02x%08x%08x%08x%08x", p[4], p[3], p[2],
p[1], p[0]);
}
sbuf_printf(sb, "\n\nCnt ID Tag UE Data RDY VLD");
for (i = 0; i < CIM_MALA_SIZE; i++, p += 5) {
sbuf_printf(sb, "\n%3u %2u %x %u %08x%08x %u %u",
(p[2] >> 10) & 0xff, (p[2] >> 7) & 7,
(p[2] >> 3) & 0xf, (p[2] >> 2) & 1,
(p[1] >> 2) | ((p[2] & 3) << 30),
(p[0] >> 2) | ((p[1] & 3) << 30), (p[0] >> 1) & 1,
p[0] & 1);
}
rc = sbuf_finish(sb);
sbuf_delete(sb);
free(buf, M_CXGBE);
return (rc);
}
static int
sysctl_cim_pif_la(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
u_int i;
struct sbuf *sb;
uint32_t *buf, *p;
int rc;
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req);
if (sb == NULL)
return (ENOMEM);
buf = malloc(2 * CIM_PIFLA_SIZE * 6 * sizeof(uint32_t), M_CXGBE,
M_ZERO | M_WAITOK);
t4_cim_read_pif_la(sc, buf, buf + 6 * CIM_PIFLA_SIZE, NULL, NULL);
p = buf;
sbuf_printf(sb, "Cntl ID DataBE Addr Data");
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
for (i = 0; i < CIM_PIFLA_SIZE; i++, p += 6) {
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
sbuf_printf(sb, "\n %02x %02x %04x %08x %08x%08x%08x%08x",
(p[5] >> 22) & 0xff, (p[5] >> 16) & 0x3f, p[5] & 0xffff,
p[4], p[3], p[2], p[1], p[0]);
}
sbuf_printf(sb, "\n\nCntl ID Data");
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
for (i = 0; i < CIM_PIFLA_SIZE; i++, p += 6) {
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
sbuf_printf(sb, "\n %02x %02x %08x%08x%08x%08x",
(p[4] >> 6) & 0xff, p[4] & 0x3f, p[3], p[2], p[1], p[0]);
}
rc = sbuf_finish(sb);
sbuf_delete(sb);
free(buf, M_CXGBE);
return (rc);
}
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
static int
sysctl_cim_qcfg(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
struct sbuf *sb;
int rc, i;
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
uint16_t base[CIM_NUM_IBQ + CIM_NUM_OBQ_T5];
uint16_t size[CIM_NUM_IBQ + CIM_NUM_OBQ_T5];
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
uint16_t thres[CIM_NUM_IBQ];
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
uint32_t obq_wr[2 * CIM_NUM_OBQ_T5], *wr = obq_wr;
uint32_t stat[4 * (CIM_NUM_IBQ + CIM_NUM_OBQ_T5)], *p = stat;
u_int cim_num_obq, ibq_rdaddr, obq_rdaddr, nq;
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
cim_num_obq = sc->chip_params->cim_num_obq;
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
if (is_t4(sc)) {
ibq_rdaddr = A_UP_IBQ_0_RDADDR;
obq_rdaddr = A_UP_OBQ_0_REALADDR;
} else {
ibq_rdaddr = A_UP_IBQ_0_SHADOW_RDADDR;
obq_rdaddr = A_UP_OBQ_0_SHADOW_REALADDR;
}
nq = CIM_NUM_IBQ + cim_num_obq;
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
rc = -t4_cim_read(sc, ibq_rdaddr, 4 * nq, stat);
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
if (rc == 0)
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
rc = -t4_cim_read(sc, obq_rdaddr, 2 * cim_num_obq, obq_wr);
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
if (rc != 0)
return (rc);
t4_read_cimq_cfg(sc, base, size, thres);
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
sb = sbuf_new_for_sysctl(NULL, NULL, PAGE_SIZE, req);
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
if (sb == NULL)
return (ENOMEM);
cxgbe(4): Adjust whitespace to line up the column titles in cim_qcfg with the values displayed.
2016-10-17 20:57:54 +00:00
sbuf_printf(sb,
" Queue Base Size Thres RdPtr WrPtr SOP EOP Avail");
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
for (i = 0; i < CIM_NUM_IBQ; i++, p += 4)
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
sbuf_printf(sb, "\n%7s %5x %5u %5u %6x %4x %4u %4u %5u",
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
qname[i], base[i], size[i], thres[i], G_IBQRDADDR(p[0]),
G_IBQWRADDR(p[1]), G_QUESOPCNT(p[3]), G_QUEEOPCNT(p[3]),
G_QUEREMFLITS(p[2]) * 16);
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
for ( ; i < nq; i++, p += 4, wr += 2)
sbuf_printf(sb, "\n%7s %5x %5u %12x %4x %4u %4u %5u", qname[i],
cxgbe(4): Add sysctls to extract debug information from the chip: dev.t4nex.X.misc.cim_la logic analyzer dump dev.t4nex.X.misc.cim_qcfg queue configuration dev.t4nex.X.misc.cim_ibq_xxx inbound queues dev.t4nex.X.misc.cim_obq_xxx outbound queues Obtained from: Chelsio MFC after: 1 week
2013-02-21 20:13:15 +00:00
base[i], size[i], G_QUERDADDR(p[0]) & 0x3fff,
wr[0] - base[i], G_QUESOPCNT(p[3]), G_QUEEOPCNT(p[3]),
G_QUEREMFLITS(p[2]) * 16);
rc = sbuf_finish(sb);
sbuf_delete(sb);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
return (rc);
}
static int
sysctl_cpl_stats(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
struct sbuf *sb;
int rc;
struct tp_cpl_stats stats;
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 256, req);
if (sb == NULL)
return (ENOMEM);
cxgbe(4): Rename regwin_lock to reg_lock. It is used to protect access to indirect registers only.
2016-03-08 22:23:30 +00:00
mtx_lock(&sc->reg_lock);
cxgbe(4): Some updates to the common code. - Updated register ranges. - Helper routines for access to TP registers. - Updated routine to read flash parameters. Obtained from: Chelsio Communications MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-26 20:20:58 +00:00
t4_tp_get_cpl_stats(sc, &stats, 0);
cxgbe(4): Rename regwin_lock to reg_lock. It is used to protect access to indirect registers only.
2016-03-08 22:23:30 +00:00
mtx_unlock(&sc->reg_lock);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
if (sc->chip_params->nchan > 2) {
sbuf_printf(sb, " channel 0 channel 1"
" channel 2 channel 3");
sbuf_printf(sb, "\nCPL requests: %10u %10u %10u %10u",
stats.req[0], stats.req[1], stats.req[2], stats.req[3]);
sbuf_printf(sb, "\nCPL responses: %10u %10u %10u %10u",
stats.rsp[0], stats.rsp[1], stats.rsp[2], stats.rsp[3]);
} else {
sbuf_printf(sb, " channel 0 channel 1");
sbuf_printf(sb, "\nCPL requests: %10u %10u",
stats.req[0], stats.req[1]);
sbuf_printf(sb, "\nCPL responses: %10u %10u",
stats.rsp[0], stats.rsp[1]);
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
rc = sbuf_finish(sb);
sbuf_delete(sb);
return (rc);
}
static int
sysctl_ddp_stats(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
struct sbuf *sb;
int rc;
struct tp_usm_stats stats;
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return(rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 256, req);
if (sb == NULL)
return (ENOMEM);
cxgbe(4): Some updates to the common code. - Updated register ranges. - Helper routines for access to TP registers. - Updated routine to read flash parameters. Obtained from: Chelsio Communications MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-26 20:20:58 +00:00
t4_get_usm_stats(sc, &stats, 1);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
sbuf_printf(sb, "Frames: %u\n", stats.frames);
sbuf_printf(sb, "Octets: %ju\n", stats.octets);
sbuf_printf(sb, "Drops: %u", stats.drops);
rc = sbuf_finish(sb);
sbuf_delete(sb);
return (rc);
}
cxgbe(4): Improvements to the code that deals with the firmware's log. - Query the location of the log very early during attach. Refresh the location later after establishing contact with the firmware. - Save the log's location as a flat address in devlog_params. - Use a memory window instead of backdoor access to the EDC/MC to read the log.
2016-03-10 23:17:26 +00:00
static const char * const devlog_level_strings[] = {
Firmware device log. # sysctl dev.t4nex.0.devlog MFC after: mdf's sysctl+sbuf changes are MFC'd
2011-05-31 23:49:13 +00:00
[FW_DEVLOG_LEVEL_EMERG] = "EMERG",
[FW_DEVLOG_LEVEL_CRIT] = "CRIT",
[FW_DEVLOG_LEVEL_ERR] = "ERR",
[FW_DEVLOG_LEVEL_NOTICE] = "NOTICE",
[FW_DEVLOG_LEVEL_INFO] = "INFO",
[FW_DEVLOG_LEVEL_DEBUG] = "DEBUG"
};
cxgbe(4): Improvements to the code that deals with the firmware's log. - Query the location of the log very early during attach. Refresh the location later after establishing contact with the firmware. - Save the log's location as a flat address in devlog_params. - Use a memory window instead of backdoor access to the EDC/MC to read the log.
2016-03-10 23:17:26 +00:00
static const char * const devlog_facility_strings[] = {
Firmware device log. # sysctl dev.t4nex.0.devlog MFC after: mdf's sysctl+sbuf changes are MFC'd
2011-05-31 23:49:13 +00:00
[FW_DEVLOG_FACILITY_CORE] = "CORE",
cxgbe(4): Display CF facility correctly in the device log. MFC after: 3 days
2014-07-15 18:24:41 +00:00
[FW_DEVLOG_FACILITY_CF] = "CF",
Firmware device log. # sysctl dev.t4nex.0.devlog MFC after: mdf's sysctl+sbuf changes are MFC'd
2011-05-31 23:49:13 +00:00
[FW_DEVLOG_FACILITY_SCHED] = "SCHED",
[FW_DEVLOG_FACILITY_TIMER] = "TIMER",
[FW_DEVLOG_FACILITY_RES] = "RES",
[FW_DEVLOG_FACILITY_HW] = "HW",
[FW_DEVLOG_FACILITY_FLR] = "FLR",
[FW_DEVLOG_FACILITY_DMAQ] = "DMAQ",
[FW_DEVLOG_FACILITY_PHY] = "PHY",
[FW_DEVLOG_FACILITY_MAC] = "MAC",
[FW_DEVLOG_FACILITY_PORT] = "PORT",
[FW_DEVLOG_FACILITY_VI] = "VI",
[FW_DEVLOG_FACILITY_FILTER] = "FILTER",
[FW_DEVLOG_FACILITY_ACL] = "ACL",
[FW_DEVLOG_FACILITY_TM] = "TM",
[FW_DEVLOG_FACILITY_QFC] = "QFC",
[FW_DEVLOG_FACILITY_DCB] = "DCB",
[FW_DEVLOG_FACILITY_ETH] = "ETH",
[FW_DEVLOG_FACILITY_OFLD] = "OFLD",
[FW_DEVLOG_FACILITY_RI] = "RI",
[FW_DEVLOG_FACILITY_ISCSI] = "ISCSI",
[FW_DEVLOG_FACILITY_FCOE] = "FCOE",
[FW_DEVLOG_FACILITY_FOISCSI] = "FOISCSI",
cxgbe(4): Improvements to the code that deals with the firmware's log. - Query the location of the log very early during attach. Refresh the location later after establishing contact with the firmware. - Save the log's location as a flat address in devlog_params. - Use a memory window instead of backdoor access to the EDC/MC to read the log.
2016-03-10 23:17:26 +00:00
[FW_DEVLOG_FACILITY_FOFCOE] = "FOFCOE",
[FW_DEVLOG_FACILITY_CHNET] = "CHNET",
Firmware device log. # sysctl dev.t4nex.0.devlog MFC after: mdf's sysctl+sbuf changes are MFC'd
2011-05-31 23:49:13 +00:00
};
static int
cxgbe(4): Auto-dump the device log on a mailbox timeout or when the firmware reports an error in pcie_fw. Sponsored by: Chelsio Communications
2019-02-07 05:06:29 +00:00
sbuf_devlog(struct adapter *sc, struct sbuf *sb, int flags)
Firmware device log. # sysctl dev.t4nex.0.devlog MFC after: mdf's sysctl+sbuf changes are MFC'd
2011-05-31 23:49:13 +00:00
{
cxgbe(4): Auto-dump the device log on a mailbox timeout or when the firmware reports an error in pcie_fw. Sponsored by: Chelsio Communications
2019-02-07 05:06:29 +00:00
int i, j, rc, nentries, first = 0;
Firmware device log. # sysctl dev.t4nex.0.devlog MFC after: mdf's sysctl+sbuf changes are MFC'd
2011-05-31 23:49:13 +00:00
struct devlog_params *dparams = &sc->params.devlog;
struct fw_devlog_e *buf, *e;
uint64_t ftstamp = UINT64_MAX;
cxgbe(4): Improvements to the code that deals with the firmware's log. - Query the location of the log very early during attach. Refresh the location later after establishing contact with the firmware. - Save the log's location as a flat address in devlog_params. - Use a memory window instead of backdoor access to the EDC/MC to read the log.
2016-03-10 23:17:26 +00:00
if (dparams->addr == 0)
return (ENXIO);
Firmware device log. # sysctl dev.t4nex.0.devlog MFC after: mdf's sysctl+sbuf changes are MFC'd
2011-05-31 23:49:13 +00:00
cxgbe(4): Auto-dump the device log on a mailbox timeout or when the firmware reports an error in pcie_fw. Sponsored by: Chelsio Communications
2019-02-07 05:06:29 +00:00
MPASS(flags == M_WAITOK || flags == M_NOWAIT);
buf = malloc(dparams->size, M_CXGBE, M_ZERO | flags);
Firmware device log. # sysctl dev.t4nex.0.devlog MFC after: mdf's sysctl+sbuf changes are MFC'd
2011-05-31 23:49:13 +00:00
if (buf == NULL)
return (ENOMEM);
cxgbe(4): Improvements to the code that deals with the firmware's log. - Query the location of the log very early during attach. Refresh the location later after establishing contact with the firmware. - Save the log's location as a flat address in devlog_params. - Use a memory window instead of backdoor access to the EDC/MC to read the log.
2016-03-10 23:17:26 +00:00
rc = read_via_memwin(sc, 1, dparams->addr, (void *)buf, dparams->size);
Firmware device log. # sysctl dev.t4nex.0.devlog MFC after: mdf's sysctl+sbuf changes are MFC'd
2011-05-31 23:49:13 +00:00
if (rc != 0)
goto done;
cxgbe(4): Improvements to the code that deals with the firmware's log. - Query the location of the log very early during attach. Refresh the location later after establishing contact with the firmware. - Save the log's location as a flat address in devlog_params. - Use a memory window instead of backdoor access to the EDC/MC to read the log.
2016-03-10 23:17:26 +00:00
nentries = dparams->size / sizeof(struct fw_devlog_e);
Firmware device log. # sysctl dev.t4nex.0.devlog MFC after: mdf's sysctl+sbuf changes are MFC'd
2011-05-31 23:49:13 +00:00
for (i = 0; i < nentries; i++) {
e = &buf[i];
if (e->timestamp == 0)
break; /* end */
e->timestamp = be64toh(e->timestamp);
e->seqno = be32toh(e->seqno);
for (j = 0; j < 8; j++)
e->params[j] = be32toh(e->params[j]);
if (e->timestamp < ftstamp) {
ftstamp = e->timestamp;
first = i;
}
}
if (buf[first].timestamp == 0)
goto done; /* nothing in the log */
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
sbuf_printf(sb, "%10s %15s %8s %8s %s\n",
Firmware device log. # sysctl dev.t4nex.0.devlog MFC after: mdf's sysctl+sbuf changes are MFC'd
2011-05-31 23:49:13 +00:00
"Seq#", "Tstamp", "Level", "Facility", "Message");
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
i = first;
do {
e = &buf[i];
if (e->timestamp == 0)
break; /* end */
sbuf_printf(sb, "%10d %15ju %8s %8s ",
e->seqno, e->timestamp,
Use native FreeBSD facilities everywhere except the shared code in common/ MFC after: 1 week
2012-09-13 09:10:10 +00:00
(e->level < nitems(devlog_level_strings) ?
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
devlog_level_strings[e->level] : "UNKNOWN"),
Use native FreeBSD facilities everywhere except the shared code in common/ MFC after: 1 week
2012-09-13 09:10:10 +00:00
(e->facility < nitems(devlog_facility_strings) ?
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
devlog_facility_strings[e->facility] : "UNKNOWN"));
sbuf_printf(sb, e->fmt, e->params[0], e->params[1],
e->params[2], e->params[3], e->params[4],
e->params[5], e->params[6], e->params[7]);
if (++i == nentries)
i = 0;
} while (i != first);
done:
free(buf, M_CXGBE);
return (rc);
}
cxgbe(4): Auto-dump the device log on a mailbox timeout or when the firmware reports an error in pcie_fw. Sponsored by: Chelsio Communications
2019-02-07 05:06:29 +00:00
static int
sysctl_devlog(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
int rc;
struct sbuf *sb;
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req);
if (sb == NULL)
return (ENOMEM);
rc = sbuf_devlog(sc, sb, M_WAITOK);
if (rc == 0)
rc = sbuf_finish(sb);
sbuf_delete(sb);
return (rc);
}
void
t4_os_dump_devlog(struct adapter *sc)
{
int rc;
struct sbuf sb;
if (sbuf_new(&sb, NULL, 4096, SBUF_AUTOEXTEND) != &sb)
return;
rc = sbuf_devlog(sc, &sb, M_NOWAIT);
if (rc == 0) {
rc = sbuf_finish(&sb);
if (rc == 0) {
log(LOG_DEBUG, "%s: device log follows.\n%s",
device_get_nameunit(sc->dev), sbuf_data(&sb));
}
}
sbuf_delete(&sb);
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
static int
sysctl_fcoe_stats(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
struct sbuf *sb;
int rc;
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
struct tp_fcoe_stats stats[MAX_NCHAN];
int i, nchan = sc->chip_params->nchan;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 256, req);
if (sb == NULL)
return (ENOMEM);
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
for (i = 0; i < nchan; i++)
cxgbe(4): Some updates to the common code. - Updated register ranges. - Helper routines for access to TP registers. - Updated routine to read flash parameters. Obtained from: Chelsio Communications MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-26 20:20:58 +00:00
t4_get_fcoe_stats(sc, i, &stats[i], 1);
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
if (nchan > 2) {
sbuf_printf(sb, " channel 0 channel 1"
" channel 2 channel 3");
sbuf_printf(sb, "\noctetsDDP: %16ju %16ju %16ju %16ju",
stats[0].octets_ddp, stats[1].octets_ddp,
stats[2].octets_ddp, stats[3].octets_ddp);
sbuf_printf(sb, "\nframesDDP: %16u %16u %16u %16u",
stats[0].frames_ddp, stats[1].frames_ddp,
stats[2].frames_ddp, stats[3].frames_ddp);
sbuf_printf(sb, "\nframesDrop: %16u %16u %16u %16u",
stats[0].frames_drop, stats[1].frames_drop,
stats[2].frames_drop, stats[3].frames_drop);
} else {
sbuf_printf(sb, " channel 0 channel 1");
sbuf_printf(sb, "\noctetsDDP: %16ju %16ju",
stats[0].octets_ddp, stats[1].octets_ddp);
sbuf_printf(sb, "\nframesDDP: %16u %16u",
stats[0].frames_ddp, stats[1].frames_ddp);
sbuf_printf(sb, "\nframesDrop: %16u %16u",
stats[0].frames_drop, stats[1].frames_drop);
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
rc = sbuf_finish(sb);
sbuf_delete(sb);
return (rc);
}
static int
sysctl_hw_sched(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
struct sbuf *sb;
int rc, i;
unsigned int map, kbps, ipg, mode;
unsigned int pace_tab[NTX_SCHED];
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 256, req);
if (sb == NULL)
return (ENOMEM);
map = t4_read_reg(sc, A_TP_TX_MOD_QUEUE_REQ_MAP);
mode = G_TIMERMODE(t4_read_reg(sc, A_TP_MOD_CONFIG));
t4_read_pace_tbl(sc, pace_tab);
sbuf_printf(sb, "Scheduler Mode Channel Rate (Kbps) "
"Class IPG (0.1 ns) Flow IPG (us)");
for (i = 0; i < NTX_SCHED; ++i, map >>= 2) {
cxgbe(4): Some updates to the common code. - Updated register ranges. - Helper routines for access to TP registers. - Updated routine to read flash parameters. Obtained from: Chelsio Communications MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-26 20:20:58 +00:00
t4_get_tx_sched(sc, i, &kbps, &ipg, 1);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
sbuf_printf(sb, "\n %u %-5s %u ", i,
(mode & (1 << i)) ? "flow" : "class", map & 3);
if (kbps)
sbuf_printf(sb, "%9u ", kbps);
else
sbuf_printf(sb, " disabled ");
if (ipg)
sbuf_printf(sb, "%13u ", ipg);
else
sbuf_printf(sb, " disabled ");
if (pace_tab[i])
sbuf_printf(sb, "%10u", pace_tab[i]);
else
sbuf_printf(sb, " disabled");
}
rc = sbuf_finish(sb);
sbuf_delete(sb);
return (rc);
}
static int
sysctl_lb_stats(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
struct sbuf *sb;
int rc, i, j;
uint64_t *p0, *p1;
struct lb_port_stats s[2];
static const char *stat_name[] = {
"OctetsOK:", "FramesOK:", "BcastFrames:", "McastFrames:",
"UcastFrames:", "ErrorFrames:", "Frames64:", "Frames65To127:",
"Frames128To255:", "Frames256To511:", "Frames512To1023:",
"Frames1024To1518:", "Frames1519ToMax:", "FramesDropped:",
"BG0FramesDropped:", "BG1FramesDropped:", "BG2FramesDropped:",
"BG3FramesDropped:", "BG0FramesTrunc:", "BG1FramesTrunc:",
"BG2FramesTrunc:", "BG3FramesTrunc:"
};
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req);
if (sb == NULL)
return (ENOMEM);
memset(s, 0, sizeof(s));
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
for (i = 0; i < sc->chip_params->nchan; i += 2) {
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
t4_get_lb_stats(sc, i, &s[0]);
t4_get_lb_stats(sc, i + 1, &s[1]);
p0 = &s[0].octets;
p1 = &s[1].octets;
sbuf_printf(sb, "%s Loopback %u"
" Loopback %u", i == 0 ? "" : "\n", i, i + 1);
Use native FreeBSD facilities everywhere except the shared code in common/ MFC after: 1 week
2012-09-13 09:10:10 +00:00
for (j = 0; j < nitems(stat_name); j++)
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
sbuf_printf(sb, "\n%-17s %20ju %20ju", stat_name[j],
*p0++, *p1++);
}
rc = sbuf_finish(sb);
sbuf_delete(sb);
return (rc);
}
Display a string instead of a numeric code in the linkdnrc sysctl. Submitted by: gnn@
2013-07-27 07:43:43 +00:00
static int
sysctl_linkdnrc(SYSCTL_HANDLER_ARGS)
{
int rc = 0;
struct port_info *pi = arg1;
cxgbe(4): Updates to link configuration. - Update struct link_settings and associated shared code. - Add tunables to control FEC and autonegotiation. All ports inherit these values as their initial settings. hw.cxgbe.fec hw.cxgbe.autoneg - Add per-port sysctls to control FEC and autonegotiation. These can be modified at any time. dev.<port>.<n>.fec dev.<port>.<n>.autoneg MFC after: 3 days Sponsored by: Chelsio Communications
2016-12-30 08:59:49 +00:00
struct link_config *lc = &pi->link_cfg;
Display a string instead of a numeric code in the linkdnrc sysctl. Submitted by: gnn@
2013-07-27 07:43:43 +00:00
struct sbuf *sb;
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return(rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 64, req);
if (sb == NULL)
return (ENOMEM);
cxgbe(4): Updates to link configuration. - Update struct link_settings and associated shared code. - Add tunables to control FEC and autonegotiation. All ports inherit these values as their initial settings. hw.cxgbe.fec hw.cxgbe.autoneg - Add per-port sysctls to control FEC and autonegotiation. These can be modified at any time. dev.<port>.<n>.fec dev.<port>.<n>.autoneg MFC after: 3 days Sponsored by: Chelsio Communications
2016-12-30 08:59:49 +00:00
if (lc->link_ok || lc->link_down_rc == 255)
Display a string instead of a numeric code in the linkdnrc sysctl. Submitted by: gnn@
2013-07-27 07:43:43 +00:00
sbuf_printf(sb, "n/a");
else
cxgbe(4): Updates to link configuration. - Update struct link_settings and associated shared code. - Add tunables to control FEC and autonegotiation. All ports inherit these values as their initial settings. hw.cxgbe.fec hw.cxgbe.autoneg - Add per-port sysctls to control FEC and autonegotiation. These can be modified at any time. dev.<port>.<n>.fec dev.<port>.<n>.autoneg MFC after: 3 days Sponsored by: Chelsio Communications
2016-12-30 08:59:49 +00:00
sbuf_printf(sb, "%s", t4_link_down_rc_str(lc->link_down_rc));
Display a string instead of a numeric code in the linkdnrc sysctl. Submitted by: gnn@
2013-07-27 07:43:43 +00:00
rc = sbuf_finish(sb);
sbuf_delete(sb);
return (rc);
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
struct mem_desc {
unsigned int base;
unsigned int limit;
unsigned int idx;
};
static int
mem_desc_cmp(const void *a, const void *b)
{
return ((const struct mem_desc *)a)->base -
((const struct mem_desc *)b)->base;
}
static void
mem_region_show(struct sbuf *sb, const char *name, unsigned int from,
unsigned int to)
{
unsigned int size;
cxgbe(4): Remove a couple of pointless assignments in sysctl_meminfo. Do not display range if start = stop (this is a workaround for some unused regions).
2016-03-16 19:36:11 +00:00
if (from == to)
return;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
size = to - from + 1;
if (size == 0)
return;
/* XXX: need humanize_number(3) in libkern for a more readable 'size' */
sbuf_printf(sb, "%-15s %#x-%#x [%u]\n", name, from, to, size);
}
static int
sysctl_meminfo(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
struct sbuf *sb;
int rc, i, n;
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
uint32_t lo, hi, used, alloc;
static const char *memory[] = {"EDC0:", "EDC1:", "MC:", "MC0:", "MC1:"};
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
static const char *region[] = {
"DBQ contexts:", "IMSG contexts:", "FLM cache:", "TCBs:",
"Pstructs:", "Timers:", "Rx FL:", "Tx FL:", "Pstruct FL:",
"Tx payload:", "Rx payload:", "LE hash:", "iSCSI region:",
"TDDP region:", "TPT region:", "STAG region:", "RQ region:",
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
"RQUDP region:", "PBL region:", "TXPBL region:",
"DBVFIFO region:", "ULPRX state:", "ULPTX state:",
Fetch TLS key parameters from the firmware. The parameters describe how much of the adapter's memory is reserved for storing TLS keys. The 'meminfo' sysctl now lists this region of adapter memory as 'TLS keys' if present. Sponsored by: Chelsio Communications
2018-02-26 21:56:06 +00:00
"On-chip queues:", "TLS keys:",
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
};
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
struct mem_desc avail[4];
Use native FreeBSD facilities everywhere except the shared code in common/ MFC after: 1 week
2012-09-13 09:10:10 +00:00
struct mem_desc mem[nitems(region) + 3]; /* up to 3 holes */
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
struct mem_desc *md = mem;
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req);
if (sb == NULL)
return (ENOMEM);
Use native FreeBSD facilities everywhere except the shared code in common/ MFC after: 1 week
2012-09-13 09:10:10 +00:00
for (i = 0; i < nitems(mem); i++) {
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
mem[i].limit = 0;
mem[i].idx = i;
}
/* Find and sort the populated memory ranges */
i = 0;
lo = t4_read_reg(sc, A_MA_TARGET_MEM_ENABLE);
if (lo & F_EDRAM0_ENABLE) {
hi = t4_read_reg(sc, A_MA_EDRAM0_BAR);
avail[i].base = G_EDRAM0_BASE(hi) << 20;
avail[i].limit = avail[i].base + (G_EDRAM0_SIZE(hi) << 20);
avail[i].idx = 0;
i++;
}
if (lo & F_EDRAM1_ENABLE) {
hi = t4_read_reg(sc, A_MA_EDRAM1_BAR);
avail[i].base = G_EDRAM1_BASE(hi) << 20;
avail[i].limit = avail[i].base + (G_EDRAM1_SIZE(hi) << 20);
avail[i].idx = 1;
i++;
}
if (lo & F_EXT_MEM_ENABLE) {
hi = t4_read_reg(sc, A_MA_EXT_MEMORY_BAR);
avail[i].base = G_EXT_MEM_BASE(hi) << 20;
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
avail[i].limit = avail[i].base +
(G_EXT_MEM_SIZE(hi) << 20);
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
avail[i].idx = is_t5(sc) ? 3 : 2; /* Call it MC0 for T5 */
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
i++;
}
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
if (is_t5(sc) && lo & F_EXT_MEM1_ENABLE) {
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
hi = t4_read_reg(sc, A_MA_EXT_MEMORY1_BAR);
avail[i].base = G_EXT_MEM1_BASE(hi) << 20;
avail[i].limit = avail[i].base +
(G_EXT_MEM1_SIZE(hi) << 20);
avail[i].idx = 4;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
i++;
}
if (!i) /* no memory available */
return 0;
qsort(avail, i, sizeof(struct mem_desc), mem_desc_cmp);
(md++)->base = t4_read_reg(sc, A_SGE_DBQ_CTXT_BADDR);
(md++)->base = t4_read_reg(sc, A_SGE_IMSG_CTXT_BADDR);
(md++)->base = t4_read_reg(sc, A_SGE_FLM_CACHE_BADDR);
(md++)->base = t4_read_reg(sc, A_TP_CMM_TCB_BASE);
(md++)->base = t4_read_reg(sc, A_TP_CMM_MM_BASE);
(md++)->base = t4_read_reg(sc, A_TP_CMM_TIMER_BASE);
(md++)->base = t4_read_reg(sc, A_TP_CMM_MM_RX_FLST_BASE);
(md++)->base = t4_read_reg(sc, A_TP_CMM_MM_TX_FLST_BASE);
(md++)->base = t4_read_reg(sc, A_TP_CMM_MM_PS_FLST_BASE);
/* the next few have explicit upper bounds */
md->base = t4_read_reg(sc, A_TP_PMM_TX_BASE);
md->limit = md->base - 1 +
t4_read_reg(sc, A_TP_PMM_TX_PAGE_SIZE) *
G_PMTXMAXPAGE(t4_read_reg(sc, A_TP_PMM_TX_MAX_PAGE));
md++;
md->base = t4_read_reg(sc, A_TP_PMM_RX_BASE);
md->limit = md->base - 1 +
t4_read_reg(sc, A_TP_PMM_RX_PAGE_SIZE) *
G_PMRXMAXPAGE(t4_read_reg(sc, A_TP_PMM_RX_MAX_PAGE));
md++;
if (t4_read_reg(sc, A_LE_DB_CONFIG) & F_HASHEN) {
cxgbe(4): Remove a couple of pointless assignments in sysctl_meminfo. Do not display range if start = stop (this is a workaround for some unused regions).
2016-03-16 19:36:11 +00:00
if (chip_id(sc) <= CHELSIO_T5)
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
md->base = t4_read_reg(sc, A_LE_DB_HASH_TID_BASE);
cxgbe(4): Remove a couple of pointless assignments in sysctl_meminfo. Do not display range if start = stop (this is a workaround for some unused regions).
2016-03-16 19:36:11 +00:00
else
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
md->base = t4_read_reg(sc, A_LE_DB_HASH_TBL_BASE_ADDR);
md->limit = 0;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
} else {
md->base = 0;
Use native FreeBSD facilities everywhere except the shared code in common/ MFC after: 1 week
2012-09-13 09:10:10 +00:00
md->idx = nitems(region); /* hide it */
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
md++;
#define ulp_region(reg) \
md->base = t4_read_reg(sc, A_ULP_ ## reg ## _LLIMIT);\
(md++)->limit = t4_read_reg(sc, A_ULP_ ## reg ## _ULIMIT)
ulp_region(RX_ISCSI);
ulp_region(RX_TDDP);
ulp_region(TX_TPT);
ulp_region(RX_STAG);
ulp_region(RX_RQ);
ulp_region(RX_RQUDP);
ulp_region(RX_PBL);
ulp_region(TX_PBL);
#undef ulp_region
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
md->base = 0;
md->idx = nitems(region);
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
if (!is_t4(sc)) {
uint32_t size = 0;
uint32_t sge_ctrl = t4_read_reg(sc, A_SGE_CONTROL2);
uint32_t fifo_size = t4_read_reg(sc, A_SGE_DBVFIFO_SIZE);
if (is_t5(sc)) {
if (sge_ctrl & F_VFIFO_ENABLE)
size = G_DBVFIFO_SIZE(fifo_size);
} else
size = G_T6_DBVFIFO_SIZE(fifo_size);
if (size) {
md->base = G_BASEADDR(t4_read_reg(sc,
A_SGE_DBVFIFO_BADDR));
md->limit = md->base + (size << 2) - 1;
}
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
}
md++;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
md->base = t4_read_reg(sc, A_ULP_RX_CTX_BASE);
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
md->limit = 0;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
md++;
md->base = t4_read_reg(sc, A_ULP_TX_ERR_TABLE_BASE);
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
md->limit = 0;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
md++;
md->base = sc->vres.ocq.start;
if (sc->vres.ocq.size)
md->limit = md->base + sc->vres.ocq.size - 1;
else
Use native FreeBSD facilities everywhere except the shared code in common/ MFC after: 1 week
2012-09-13 09:10:10 +00:00
md->idx = nitems(region); /* hide it */
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
md++;
Fetch TLS key parameters from the firmware. The parameters describe how much of the adapter's memory is reserved for storing TLS keys. The 'meminfo' sysctl now lists this region of adapter memory as 'TLS keys' if present. Sponsored by: Chelsio Communications
2018-02-26 21:56:06 +00:00
md->base = sc->vres.key.start;
if (sc->vres.key.size)
md->limit = md->base + sc->vres.key.size - 1;
else
md->idx = nitems(region); /* hide it */
md++;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
/* add any address-space holes, there can be up to 3 */
for (n = 0; n < i - 1; n++)
if (avail[n].limit < avail[n + 1].base)
(md++)->base = avail[n].limit;
if (avail[n].limit)
(md++)->base = avail[n].limit;
n = md - mem;
qsort(mem, n, sizeof(struct mem_desc), mem_desc_cmp);
for (lo = 0; lo < i; lo++)
mem_region_show(sb, memory[avail[lo].idx], avail[lo].base,
avail[lo].limit - 1);
sbuf_printf(sb, "\n");
for (i = 0; i < n; i++) {
Use native FreeBSD facilities everywhere except the shared code in common/ MFC after: 1 week
2012-09-13 09:10:10 +00:00
if (mem[i].idx >= nitems(region))
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
continue; /* skip holes */
if (!mem[i].limit)
mem[i].limit = i < n - 1 ? mem[i + 1].base - 1 : ~0;
mem_region_show(sb, region[mem[i].idx], mem[i].base,
mem[i].limit);
}
sbuf_printf(sb, "\n");
lo = t4_read_reg(sc, A_CIM_SDRAM_BASE_ADDR);
hi = t4_read_reg(sc, A_CIM_SDRAM_ADDR_SIZE) + lo - 1;
mem_region_show(sb, "uP RAM:", lo, hi);
lo = t4_read_reg(sc, A_CIM_EXTMEM2_BASE_ADDR);
hi = t4_read_reg(sc, A_CIM_EXTMEM2_ADDR_SIZE) + lo - 1;
mem_region_show(sb, "uP Extmem2:", lo, hi);
lo = t4_read_reg(sc, A_TP_PMM_RX_MAX_PAGE);
sbuf_printf(sb, "\n%u Rx pages of size %uKiB for %u channels\n",
G_PMRXMAXPAGE(lo),
t4_read_reg(sc, A_TP_PMM_RX_PAGE_SIZE) >> 10,
(lo & F_PMRXNUMCHN) ? 2 : 1);
lo = t4_read_reg(sc, A_TP_PMM_TX_MAX_PAGE);
hi = t4_read_reg(sc, A_TP_PMM_TX_PAGE_SIZE);
sbuf_printf(sb, "%u Tx pages of size %u%ciB for %u channels\n",
G_PMTXMAXPAGE(lo),
hi >= (1 << 20) ? (hi >> 20) : (hi >> 10),
hi >= (1 << 20) ? 'M' : 'K', 1 << G_PMTXNUMCHN(lo));
sbuf_printf(sb, "%u p-structs\n",
t4_read_reg(sc, A_TP_CMM_MM_MAX_PSTRUCT));
for (i = 0; i < 4; i++) {
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
if (chip_id(sc) > CHELSIO_T5)
lo = t4_read_reg(sc, A_MPS_RX_MAC_BG_PG_CNT0 + i * 4);
else
lo = t4_read_reg(sc, A_MPS_RX_PG_RSV0 + i * 4);
if (is_t5(sc)) {
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
used = G_T5_USED(lo);
alloc = G_T5_ALLOC(lo);
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
} else {
used = G_USED(lo);
alloc = G_ALLOC(lo);
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
}
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
/* For T6 these are MAC buffer groups */
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
sbuf_printf(sb, "\nPort %d using %u pages out of %u allocated",
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
i, used, alloc);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
for (i = 0; i < sc->chip_params->nchan; i++) {
if (chip_id(sc) > CHELSIO_T5)
lo = t4_read_reg(sc, A_MPS_RX_LPBK_BG_PG_CNT0 + i * 4);
else
lo = t4_read_reg(sc, A_MPS_RX_PG_RSV4 + i * 4);
if (is_t5(sc)) {
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
used = G_T5_USED(lo);
alloc = G_T5_ALLOC(lo);
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
} else {
used = G_USED(lo);
alloc = G_ALLOC(lo);
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
}
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
/* For T6 these are MAC buffer groups */
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
sbuf_printf(sb,
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
"\nLoopback %d using %u pages out of %u allocated",
i, used, alloc);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
rc = sbuf_finish(sb);
sbuf_delete(sb);
return (rc);
}
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
static inline void
tcamxy2valmask(uint64_t x, uint64_t y, uint8_t *addr, uint64_t *mask)
{
*mask = x | y;
y = htobe64(y);
memcpy(addr, (char *)&y + 2, ETHER_ADDR_LEN);
}
static int
sysctl_mps_tcam(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
struct sbuf *sb;
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
int rc, i;
MPASS(chip_id(sc) <= CHELSIO_T5);
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req);
if (sb == NULL)
return (ENOMEM);
sbuf_printf(sb,
"Idx Ethernet address Mask Vld Ports PF"
" VF Replication P0 P1 P2 P3 ML");
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
for (i = 0; i < sc->chip_params->mps_tcam_size; i++) {
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
uint64_t tcamx, tcamy, mask;
uint32_t cls_lo, cls_hi;
uint8_t addr[ETHER_ADDR_LEN];
tcamy = t4_read_reg64(sc, MPS_CLS_TCAM_Y_L(i));
tcamx = t4_read_reg64(sc, MPS_CLS_TCAM_X_L(i));
if (tcamx & tcamy)
continue;
tcamxy2valmask(tcamx, tcamy, addr, &mask);
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
cls_lo = t4_read_reg(sc, MPS_CLS_SRAM_L(i));
cls_hi = t4_read_reg(sc, MPS_CLS_SRAM_H(i));
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
sbuf_printf(sb, "\n%3u %02x:%02x:%02x:%02x:%02x:%02x %012jx"
" %c %#x%4u%4d", i, addr[0], addr[1], addr[2],
addr[3], addr[4], addr[5], (uintmax_t)mask,
(cls_lo & F_SRAM_VLD) ? 'Y' : 'N',
G_PORTMAP(cls_hi), G_PF(cls_lo),
(cls_lo & F_VF_VALID) ? G_VF(cls_lo) : -1);
if (cls_lo & F_REPLICATE) {
struct fw_ldst_cmd ldst_cmd;
memset(&ldst_cmd, 0, sizeof(ldst_cmd));
ldst_cmd.op_to_addrspace =
htobe32(V_FW_CMD_OP(FW_LDST_CMD) |
F_FW_CMD_REQUEST | F_FW_CMD_READ |
V_FW_LDST_CMD_ADDRSPACE(FW_LDST_ADDRSPC_MPS));
ldst_cmd.cycles_to_len16 = htobe32(FW_LEN16(ldst_cmd));
cxgbe(4): Update T4 and T5 firmwares to 1.14.2.0. Obtained from: Chelsio Communications MFC after: 3 days
2015-07-14 08:02:05 +00:00
ldst_cmd.u.mps.rplc.fid_idx =
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
htobe16(V_FW_LDST_CMD_FID(FW_LDST_MPS_RPLC) |
cxgbe(4): Update T4 and T5 firmwares to 1.14.2.0. Obtained from: Chelsio Communications MFC after: 3 days
2015-07-14 08:02:05 +00:00
V_FW_LDST_CMD_IDX(i));
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
rc = begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK,
"t4mps");
if (rc)
break;
rc = -t4_wr_mbox(sc, sc->mbox, &ldst_cmd,
sizeof(ldst_cmd), &ldst_cmd);
end_synchronized_op(sc, 0);
if (rc != 0) {
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
sbuf_printf(sb, "%36d", rc);
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
rc = 0;
} else {
sbuf_printf(sb, " %08x %08x %08x %08x",
cxgbe(4): Update T4 and T5 firmwares to 1.14.2.0. Obtained from: Chelsio Communications MFC after: 3 days
2015-07-14 08:02:05 +00:00
be32toh(ldst_cmd.u.mps.rplc.rplc127_96),
be32toh(ldst_cmd.u.mps.rplc.rplc95_64),
be32toh(ldst_cmd.u.mps.rplc.rplc63_32),
be32toh(ldst_cmd.u.mps.rplc.rplc31_0));
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
}
} else
sbuf_printf(sb, "%36s", "");
sbuf_printf(sb, "%4u%3u%3u%3u %#3x", G_SRAM_PRIO0(cls_lo),
G_SRAM_PRIO1(cls_lo), G_SRAM_PRIO2(cls_lo),
G_SRAM_PRIO3(cls_lo), (cls_lo >> S_MULTILISTEN0) & 0xf);
}
if (rc)
(void) sbuf_finish(sb);
else
rc = sbuf_finish(sb);
sbuf_delete(sb);
return (rc);
}
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
static int
sysctl_mps_tcam_t6(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
struct sbuf *sb;
int rc, i;
MPASS(chip_id(sc) > CHELSIO_T5);
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req);
if (sb == NULL)
return (ENOMEM);
sbuf_printf(sb, "Idx Ethernet address Mask VNI Mask"
" IVLAN Vld DIP_Hit Lookup Port Vld Ports PF VF"
" Replication"
" P0 P1 P2 P3 ML\n");
for (i = 0; i < sc->chip_params->mps_tcam_size; i++) {
uint8_t dip_hit, vlan_vld, lookup_type, port_num;
uint16_t ivlan;
uint64_t tcamx, tcamy, val, mask;
uint32_t cls_lo, cls_hi, ctl, data2, vnix, vniy;
uint8_t addr[ETHER_ADDR_LEN];
ctl = V_CTLREQID(1) | V_CTLCMDTYPE(0) | V_CTLXYBITSEL(0);
if (i < 256)
ctl |= V_CTLTCAMINDEX(i) | V_CTLTCAMSEL(0);
else
ctl |= V_CTLTCAMINDEX(i - 256) | V_CTLTCAMSEL(1);
t4_write_reg(sc, A_MPS_CLS_TCAM_DATA2_CTL, ctl);
val = t4_read_reg(sc, A_MPS_CLS_TCAM_RDATA1_REQ_ID1);
tcamy = G_DMACH(val) << 32;
tcamy |= t4_read_reg(sc, A_MPS_CLS_TCAM_RDATA0_REQ_ID1);
data2 = t4_read_reg(sc, A_MPS_CLS_TCAM_RDATA2_REQ_ID1);
lookup_type = G_DATALKPTYPE(data2);
port_num = G_DATAPORTNUM(data2);
if (lookup_type && lookup_type != M_DATALKPTYPE) {
/* Inner header VNI */
vniy = ((data2 & F_DATAVIDH2) << 23) |
(G_DATAVIDH1(data2) << 16) | G_VIDL(val);
dip_hit = data2 & F_DATADIPHIT;
vlan_vld = 0;
} else {
vniy = 0;
dip_hit = 0;
vlan_vld = data2 & F_DATAVIDH2;
ivlan = G_VIDL(val);
}
ctl |= V_CTLXYBITSEL(1);
t4_write_reg(sc, A_MPS_CLS_TCAM_DATA2_CTL, ctl);
val = t4_read_reg(sc, A_MPS_CLS_TCAM_RDATA1_REQ_ID1);
tcamx = G_DMACH(val) << 32;
tcamx |= t4_read_reg(sc, A_MPS_CLS_TCAM_RDATA0_REQ_ID1);
data2 = t4_read_reg(sc, A_MPS_CLS_TCAM_RDATA2_REQ_ID1);
if (lookup_type && lookup_type != M_DATALKPTYPE) {
/* Inner header VNI mask */
vnix = ((data2 & F_DATAVIDH2) << 23) |
(G_DATAVIDH1(data2) << 16) | G_VIDL(val);
} else
vnix = 0;
if (tcamx & tcamy)
continue;
tcamxy2valmask(tcamx, tcamy, addr, &mask);
cls_lo = t4_read_reg(sc, MPS_CLS_SRAM_L(i));
cls_hi = t4_read_reg(sc, MPS_CLS_SRAM_H(i));
if (lookup_type && lookup_type != M_DATALKPTYPE) {
sbuf_printf(sb, "\n%3u %02x:%02x:%02x:%02x:%02x:%02x "
"%012jx %06x %06x - - %3c"
" 'I' %4x %3c %#x%4u%4d", i, addr[0],
addr[1], addr[2], addr[3], addr[4], addr[5],
(uintmax_t)mask, vniy, vnix, dip_hit ? 'Y' : 'N',
port_num, cls_lo & F_T6_SRAM_VLD ? 'Y' : 'N',
G_PORTMAP(cls_hi), G_T6_PF(cls_lo),
cls_lo & F_T6_VF_VALID ? G_T6_VF(cls_lo) : -1);
} else {
sbuf_printf(sb, "\n%3u %02x:%02x:%02x:%02x:%02x:%02x "
"%012jx - - ", i, addr[0], addr[1],
addr[2], addr[3], addr[4], addr[5],
(uintmax_t)mask);
if (vlan_vld)
sbuf_printf(sb, "%4u Y ", ivlan);
else
sbuf_printf(sb, " - N ");
sbuf_printf(sb, "- %3c %4x %3c %#x%4u%4d",
lookup_type ? 'I' : 'O', port_num,
cls_lo & F_T6_SRAM_VLD ? 'Y' : 'N',
G_PORTMAP(cls_hi), G_T6_PF(cls_lo),
cls_lo & F_T6_VF_VALID ? G_T6_VF(cls_lo) : -1);
}
if (cls_lo & F_T6_REPLICATE) {
struct fw_ldst_cmd ldst_cmd;
memset(&ldst_cmd, 0, sizeof(ldst_cmd));
ldst_cmd.op_to_addrspace =
htobe32(V_FW_CMD_OP(FW_LDST_CMD) |
F_FW_CMD_REQUEST | F_FW_CMD_READ |
V_FW_LDST_CMD_ADDRSPACE(FW_LDST_ADDRSPC_MPS));
ldst_cmd.cycles_to_len16 = htobe32(FW_LEN16(ldst_cmd));
ldst_cmd.u.mps.rplc.fid_idx =
htobe16(V_FW_LDST_CMD_FID(FW_LDST_MPS_RPLC) |
V_FW_LDST_CMD_IDX(i));
rc = begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK,
"t6mps");
if (rc)
break;
rc = -t4_wr_mbox(sc, sc->mbox, &ldst_cmd,
sizeof(ldst_cmd), &ldst_cmd);
end_synchronized_op(sc, 0);
if (rc != 0) {
sbuf_printf(sb, "%72d", rc);
rc = 0;
} else {
sbuf_printf(sb, " %08x %08x %08x %08x"
" %08x %08x %08x %08x",
be32toh(ldst_cmd.u.mps.rplc.rplc255_224),
be32toh(ldst_cmd.u.mps.rplc.rplc223_192),
be32toh(ldst_cmd.u.mps.rplc.rplc191_160),
be32toh(ldst_cmd.u.mps.rplc.rplc159_128),
be32toh(ldst_cmd.u.mps.rplc.rplc127_96),
be32toh(ldst_cmd.u.mps.rplc.rplc95_64),
be32toh(ldst_cmd.u.mps.rplc.rplc63_32),
be32toh(ldst_cmd.u.mps.rplc.rplc31_0));
}
} else
sbuf_printf(sb, "%72s", "");
sbuf_printf(sb, "%4u%3u%3u%3u %#x",
G_T6_SRAM_PRIO0(cls_lo), G_T6_SRAM_PRIO1(cls_lo),
G_T6_SRAM_PRIO2(cls_lo), G_T6_SRAM_PRIO3(cls_lo),
(cls_lo >> S_T6_MULTILISTEN0) & 0xf);
}
if (rc)
(void) sbuf_finish(sb);
else
rc = sbuf_finish(sb);
sbuf_delete(sb);
return (rc);
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
static int
sysctl_path_mtus(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
struct sbuf *sb;
int rc;
uint16_t mtus[NMTUS];
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 256, req);
if (sb == NULL)
return (ENOMEM);
t4_read_mtu_tbl(sc, mtus, NULL);
sbuf_printf(sb, "%u %u %u %u %u %u %u %u %u %u %u %u %u %u %u %u",
mtus[0], mtus[1], mtus[2], mtus[3], mtus[4], mtus[5], mtus[6],
mtus[7], mtus[8], mtus[9], mtus[10], mtus[11], mtus[12], mtus[13],
mtus[14], mtus[15]);
rc = sbuf_finish(sb);
sbuf_delete(sb);
return (rc);
}
static int
sysctl_pm_stats(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
struct sbuf *sb;
int rc, i;
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
uint32_t tx_cnt[MAX_PM_NSTATS], rx_cnt[MAX_PM_NSTATS];
uint64_t tx_cyc[MAX_PM_NSTATS], rx_cyc[MAX_PM_NSTATS];
static const char *tx_stats[MAX_PM_NSTATS] = {
"Read:", "Write bypass:", "Write mem:", "Bypass + mem:",
"Tx FIFO wait", NULL, "Tx latency"
cxgbe(4): Tidy up the display for payload memory statistics (pm_stats). # sysctl -n dev.t4nex.0.misc.pm_stats # sysctl -n dev.t5nex.0.misc.pm_stats MFC after: 1 week
2013-11-07 00:25:49 +00:00
};
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
static const char *rx_stats[MAX_PM_NSTATS] = {
"Read:", "Write bypass:", "Write mem:", "Flush:",
cxgbe(4): Fix whitespace in the pm_stats display.
2016-10-06 21:25:17 +00:00
"Rx FIFO wait", NULL, "Rx latency"
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
};
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 256, req);
if (sb == NULL)
return (ENOMEM);
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
t4_pmtx_get_stats(sc, tx_cnt, tx_cyc);
t4_pmrx_get_stats(sc, rx_cnt, rx_cyc);
cxgbe(4): Tidy up the display for payload memory statistics (pm_stats). # sysctl -n dev.t4nex.0.misc.pm_stats # sysctl -n dev.t5nex.0.misc.pm_stats MFC after: 1 week
2013-11-07 00:25:49 +00:00
sbuf_printf(sb, " Tx pcmds Tx bytes");
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
for (i = 0; i < 4; i++) {
sbuf_printf(sb, "\n%-13s %10u %20ju", tx_stats[i], tx_cnt[i],
tx_cyc[i]);
}
cxgbe(4): Tidy up the display for payload memory statistics (pm_stats). # sysctl -n dev.t4nex.0.misc.pm_stats # sysctl -n dev.t5nex.0.misc.pm_stats MFC after: 1 week
2013-11-07 00:25:49 +00:00
sbuf_printf(sb, "\n Rx pcmds Rx bytes");
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
for (i = 0; i < 4; i++) {
sbuf_printf(sb, "\n%-13s %10u %20ju", rx_stats[i], rx_cnt[i],
rx_cyc[i]);
}
if (chip_id(sc) > CHELSIO_T5) {
sbuf_printf(sb,
"\n Total wait Total occupancy");
sbuf_printf(sb, "\n%-13s %10u %20ju", tx_stats[i], tx_cnt[i],
tx_cyc[i]);
sbuf_printf(sb, "\n%-13s %10u %20ju", rx_stats[i], rx_cnt[i],
rx_cyc[i]);
i += 2;
MPASS(i < nitems(tx_stats));
sbuf_printf(sb,
"\n Reads Total wait");
sbuf_printf(sb, "\n%-13s %10u %20ju", tx_stats[i], tx_cnt[i],
tx_cyc[i]);
sbuf_printf(sb, "\n%-13s %10u %20ju", rx_stats[i], rx_cnt[i],
rx_cyc[i]);
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
rc = sbuf_finish(sb);
sbuf_delete(sb);
return (rc);
}
static int
sysctl_rdma_stats(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
struct sbuf *sb;
int rc;
struct tp_rdma_stats stats;
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 256, req);
if (sb == NULL)
return (ENOMEM);
cxgbe(4): Rename regwin_lock to reg_lock. It is used to protect access to indirect registers only.
2016-03-08 22:23:30 +00:00
mtx_lock(&sc->reg_lock);
cxgbe(4): Some updates to the common code. - Updated register ranges. - Helper routines for access to TP registers. - Updated routine to read flash parameters. Obtained from: Chelsio Communications MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-26 20:20:58 +00:00
t4_tp_get_rdma_stats(sc, &stats, 0);
cxgbe(4): Rename regwin_lock to reg_lock. It is used to protect access to indirect registers only.
2016-03-08 22:23:30 +00:00
mtx_unlock(&sc->reg_lock);
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
sbuf_printf(sb, "NoRQEModDefferals: %u\n", stats.rqe_dfr_mod);
sbuf_printf(sb, "NoRQEPktDefferals: %u", stats.rqe_dfr_pkt);
rc = sbuf_finish(sb);
sbuf_delete(sb);
return (rc);
}
static int
sysctl_tcp_stats(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
struct sbuf *sb;
int rc;
struct tp_tcp_stats v4, v6;
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 256, req);
if (sb == NULL)
return (ENOMEM);
cxgbe(4): Rename regwin_lock to reg_lock. It is used to protect access to indirect registers only.
2016-03-08 22:23:30 +00:00
mtx_lock(&sc->reg_lock);
cxgbe(4): Some updates to the common code. - Updated register ranges. - Helper routines for access to TP registers. - Updated routine to read flash parameters. Obtained from: Chelsio Communications MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-26 20:20:58 +00:00
t4_tp_get_tcp_stats(sc, &v4, &v6, 0);
cxgbe(4): Rename regwin_lock to reg_lock. It is used to protect access to indirect registers only.
2016-03-08 22:23:30 +00:00
mtx_unlock(&sc->reg_lock);
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
sbuf_printf(sb,
" IP IPv6\n");
sbuf_printf(sb, "OutRsts: %20u %20u\n",
cxgbe(4): First of many changes to reduce diffs with internal shared code: - Rename some CamelCase variables. - s/t4_link_start/t4_link_l1cfg/g - Pull in t4_get_port_type_description. - Move t4_wait_op_done to t4_hw.c. - Flip the order of the RDMA stats. - Remove unsused function t4_iq_start_stop. - Move t4_wait_op_done and t4_wait_op_done_val to t4_hw.c Obtained from: Chelsio Communications
2016-03-03 01:41:53 +00:00
v4.tcp_out_rsts, v6.tcp_out_rsts);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
sbuf_printf(sb, "InSegs: %20ju %20ju\n",
cxgbe(4): First of many changes to reduce diffs with internal shared code: - Rename some CamelCase variables. - s/t4_link_start/t4_link_l1cfg/g - Pull in t4_get_port_type_description. - Move t4_wait_op_done to t4_hw.c. - Flip the order of the RDMA stats. - Remove unsused function t4_iq_start_stop. - Move t4_wait_op_done and t4_wait_op_done_val to t4_hw.c Obtained from: Chelsio Communications
2016-03-03 01:41:53 +00:00
v4.tcp_in_segs, v6.tcp_in_segs);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
sbuf_printf(sb, "OutSegs: %20ju %20ju\n",
cxgbe(4): First of many changes to reduce diffs with internal shared code: - Rename some CamelCase variables. - s/t4_link_start/t4_link_l1cfg/g - Pull in t4_get_port_type_description. - Move t4_wait_op_done to t4_hw.c. - Flip the order of the RDMA stats. - Remove unsused function t4_iq_start_stop. - Move t4_wait_op_done and t4_wait_op_done_val to t4_hw.c Obtained from: Chelsio Communications
2016-03-03 01:41:53 +00:00
v4.tcp_out_segs, v6.tcp_out_segs);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
sbuf_printf(sb, "RetransSegs: %20ju %20ju",
cxgbe(4): First of many changes to reduce diffs with internal shared code: - Rename some CamelCase variables. - s/t4_link_start/t4_link_l1cfg/g - Pull in t4_get_port_type_description. - Move t4_wait_op_done to t4_hw.c. - Flip the order of the RDMA stats. - Remove unsused function t4_iq_start_stop. - Move t4_wait_op_done and t4_wait_op_done_val to t4_hw.c Obtained from: Chelsio Communications
2016-03-03 01:41:53 +00:00
v4.tcp_retrans_segs, v6.tcp_retrans_segs);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
rc = sbuf_finish(sb);
sbuf_delete(sb);
return (rc);
}
static int
sysctl_tids(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
struct sbuf *sb;
int rc;
struct tid_info *t = &sc->tids;
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 256, req);
if (sb == NULL)
return (ENOMEM);
if (t->natids) {
sbuf_printf(sb, "ATID range: 0-%u, in use: %u\n", t->natids - 1,
t->atids_in_use);
}
cxgbe(4): Add support for high priority filters on T6+. They have their own region in the TCAM starting with T6, unlike previous chips where they were in the same region as normal filters. These filters "hit" before anything else in the LE's lookup. The exact order is: a) High priority filters b) TOE's active region (TCAM and/or hash) c) Servers (TOE hw listeners) d) Normal filters MFC after: 1 week Sponsored by: Chelsio Communications
2018-08-09 14:19:47 +00:00
if (t->nhpftids) {
sbuf_printf(sb, "HPFTID range: %u-%u, in use: %u\n",
t->hpftid_base, t->hpftid_end, t->hpftids_in_use);
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
if (t->ntids) {
cxgbe(4): Fix the tid range shown for T6 cards in misc.tids. MFC after: 3 days
2016-12-14 07:36:36 +00:00
sbuf_printf(sb, "TID range: ");
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
if (t4_read_reg(sc, A_LE_DB_CONFIG) & F_HASHEN) {
cxgbe(4): Fix the tid range shown for T6 cards in misc.tids. MFC after: 3 days
2016-12-14 07:36:36 +00:00
uint32_t b, hb;
cxgbe(4): Fix the output of the "tids" sysctl on T6.
2016-09-22 21:19:25 +00:00
cxgbe(4): Fix the tid range shown for T6 cards in misc.tids. MFC after: 3 days
2016-12-14 07:36:36 +00:00
if (chip_id(sc) <= CHELSIO_T5) {
cxgbe(4): Fix the output of the "tids" sysctl on T6.
2016-09-22 21:19:25 +00:00
b = t4_read_reg(sc, A_LE_DB_SERVER_INDEX) / 4;
cxgbe(4): Fix the tid range shown for T6 cards in misc.tids. MFC after: 3 days
2016-12-14 07:36:36 +00:00
hb = t4_read_reg(sc, A_LE_DB_TID_HASHBASE) / 4;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
} else {
cxgbe(4): Fix the tid range shown for T6 cards in misc.tids. MFC after: 3 days
2016-12-14 07:36:36 +00:00
b = t4_read_reg(sc, A_LE_DB_SRVR_START_INDEX);
hb = t4_read_reg(sc, A_T6_LE_DB_HASH_TID_BASE);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
cxgbe(4): Fix the tid range shown for T6 cards in misc.tids. MFC after: 3 days
2016-12-14 07:36:36 +00:00
if (b)
cxgbe(4): Add support for high priority filters on T6+. They have their own region in the TCAM starting with T6, unlike previous chips where they were in the same region as normal filters. These filters "hit" before anything else in the LE's lookup. The exact order is: a) High priority filters b) TOE's active region (TCAM and/or hash) c) Servers (TOE hw listeners) d) Normal filters MFC after: 1 week Sponsored by: Chelsio Communications
2018-08-09 14:19:47 +00:00
sbuf_printf(sb, "%u-%u, ", t->tid_base, b - 1);
cxgbe(4): Fix the tid range shown for T6 cards in misc.tids. MFC after: 3 days
2016-12-14 07:36:36 +00:00
sbuf_printf(sb, "%u-%u", hb, t->ntids - 1);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
} else
cxgbe(4): Add support for high priority filters on T6+. They have their own region in the TCAM starting with T6, unlike previous chips where they were in the same region as normal filters. These filters "hit" before anything else in the LE's lookup. The exact order is: a) High priority filters b) TOE's active region (TCAM and/or hash) c) Servers (TOE hw listeners) d) Normal filters MFC after: 1 week Sponsored by: Chelsio Communications
2018-08-09 14:19:47 +00:00
sbuf_printf(sb, "%u-%u", t->tid_base, t->ntids - 1);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
sbuf_printf(sb, ", in use: %u\n",
atomic_load_acq_int(&t->tids_in_use));
}
if (t->nstids) {
sbuf_printf(sb, "STID range: %u-%u, in use: %u\n", t->stid_base,
t->stid_base + t->nstids - 1, t->stids_in_use);
}
if (t->nftids) {
cxgbe(4): Add support for high priority filters on T6+. They have their own region in the TCAM starting with T6, unlike previous chips where they were in the same region as normal filters. These filters "hit" before anything else in the LE's lookup. The exact order is: a) High priority filters b) TOE's active region (TCAM and/or hash) c) Servers (TOE hw listeners) d) Normal filters MFC after: 1 week Sponsored by: Chelsio Communications
2018-08-09 14:19:47 +00:00
sbuf_printf(sb, "FTID range: %u-%u, in use: %u\n", t->ftid_base,
t->ftid_end, t->ftids_in_use);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
Read card capabilities after firmware initialization, instead of setting them up as part of firmware initialization (which the driver gets to do only if it's the master driver). Read the range of tids available for the ETHOFLD functionality if it's enabled. New is_ftid() and is_etid() functions to test whether a tid falls within the range of filter tids or ETHOFLD tids respectively. MFC after: 2 weeks
2013-12-14 03:08:03 +00:00
if (t->netids) {
cxgbe(4): Implement ifnet callbacks that deal with send tags. An etid (ethoffload tid) is allocated for a send tag and it acquires a reference on the traffic class that matches the send parameters associated with the tag. Sponsored by: Chelsio Communications
2018-05-18 06:09:15 +00:00
sbuf_printf(sb, "ETID range: %u-%u, in use: %u\n", t->etid_base,
t->etid_base + t->netids - 1, t->etids_in_use);
Read card capabilities after firmware initialization, instead of setting them up as part of firmware initialization (which the driver gets to do only if it's the master driver). Read the range of tids available for the ETHOFLD functionality if it's enabled. New is_ftid() and is_etid() functions to test whether a tid falls within the range of filter tids or ETHOFLD tids respectively. MFC after: 2 weeks
2013-12-14 03:08:03 +00:00
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
sbuf_printf(sb, "HW TID usage: %u IP users, %u IPv6 users",
t4_read_reg(sc, A_LE_DB_ACT_CNT_IPV4),
t4_read_reg(sc, A_LE_DB_ACT_CNT_IPV6));
rc = sbuf_finish(sb);
sbuf_delete(sb);
return (rc);
}
static int
sysctl_tp_err_stats(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
struct sbuf *sb;
int rc;
struct tp_err_stats stats;
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 256, req);
if (sb == NULL)
return (ENOMEM);
cxgbe(4): Rename regwin_lock to reg_lock. It is used to protect access to indirect registers only.
2016-03-08 22:23:30 +00:00
mtx_lock(&sc->reg_lock);
cxgbe(4): Some updates to the common code. - Updated register ranges. - Helper routines for access to TP registers. - Updated routine to read flash parameters. Obtained from: Chelsio Communications MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-26 20:20:58 +00:00
t4_tp_get_err_stats(sc, &stats, 0);
cxgbe(4): Rename regwin_lock to reg_lock. It is used to protect access to indirect registers only.
2016-03-08 22:23:30 +00:00
mtx_unlock(&sc->reg_lock);
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
if (sc->chip_params->nchan > 2) {
sbuf_printf(sb, " channel 0 channel 1"
" channel 2 channel 3\n");
sbuf_printf(sb, "macInErrs: %10u %10u %10u %10u\n",
stats.mac_in_errs[0], stats.mac_in_errs[1],
stats.mac_in_errs[2], stats.mac_in_errs[3]);
sbuf_printf(sb, "hdrInErrs: %10u %10u %10u %10u\n",
stats.hdr_in_errs[0], stats.hdr_in_errs[1],
stats.hdr_in_errs[2], stats.hdr_in_errs[3]);
sbuf_printf(sb, "tcpInErrs: %10u %10u %10u %10u\n",
stats.tcp_in_errs[0], stats.tcp_in_errs[1],
stats.tcp_in_errs[2], stats.tcp_in_errs[3]);
sbuf_printf(sb, "tcp6InErrs: %10u %10u %10u %10u\n",
stats.tcp6_in_errs[0], stats.tcp6_in_errs[1],
stats.tcp6_in_errs[2], stats.tcp6_in_errs[3]);
sbuf_printf(sb, "tnlCongDrops: %10u %10u %10u %10u\n",
stats.tnl_cong_drops[0], stats.tnl_cong_drops[1],
stats.tnl_cong_drops[2], stats.tnl_cong_drops[3]);
sbuf_printf(sb, "tnlTxDrops: %10u %10u %10u %10u\n",
stats.tnl_tx_drops[0], stats.tnl_tx_drops[1],
stats.tnl_tx_drops[2], stats.tnl_tx_drops[3]);
sbuf_printf(sb, "ofldVlanDrops: %10u %10u %10u %10u\n",
stats.ofld_vlan_drops[0], stats.ofld_vlan_drops[1],
stats.ofld_vlan_drops[2], stats.ofld_vlan_drops[3]);
sbuf_printf(sb, "ofldChanDrops: %10u %10u %10u %10u\n\n",
stats.ofld_chan_drops[0], stats.ofld_chan_drops[1],
stats.ofld_chan_drops[2], stats.ofld_chan_drops[3]);
} else {
sbuf_printf(sb, " channel 0 channel 1\n");
sbuf_printf(sb, "macInErrs: %10u %10u\n",
stats.mac_in_errs[0], stats.mac_in_errs[1]);
sbuf_printf(sb, "hdrInErrs: %10u %10u\n",
stats.hdr_in_errs[0], stats.hdr_in_errs[1]);
sbuf_printf(sb, "tcpInErrs: %10u %10u\n",
stats.tcp_in_errs[0], stats.tcp_in_errs[1]);
sbuf_printf(sb, "tcp6InErrs: %10u %10u\n",
stats.tcp6_in_errs[0], stats.tcp6_in_errs[1]);
sbuf_printf(sb, "tnlCongDrops: %10u %10u\n",
stats.tnl_cong_drops[0], stats.tnl_cong_drops[1]);
sbuf_printf(sb, "tnlTxDrops: %10u %10u\n",
stats.tnl_tx_drops[0], stats.tnl_tx_drops[1]);
sbuf_printf(sb, "ofldVlanDrops: %10u %10u\n",
stats.ofld_vlan_drops[0], stats.ofld_vlan_drops[1]);
sbuf_printf(sb, "ofldChanDrops: %10u %10u\n\n",
stats.ofld_chan_drops[0], stats.ofld_chan_drops[1]);
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
sbuf_printf(sb, "ofldNoNeigh: %u\nofldCongDefer: %u",
cxgbe(4): First of many changes to reduce diffs with internal shared code: - Rename some CamelCase variables. - s/t4_link_start/t4_link_l1cfg/g - Pull in t4_get_port_type_description. - Move t4_wait_op_done to t4_hw.c. - Flip the order of the RDMA stats. - Remove unsused function t4_iq_start_stop. - Move t4_wait_op_done and t4_wait_op_done_val to t4_hw.c Obtained from: Chelsio Communications
2016-03-03 01:41:53 +00:00
stats.ofld_no_neigh, stats.ofld_cong_defer);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
rc = sbuf_finish(sb);
sbuf_delete(sb);
return (rc);
}
cxgbe(4): Add a sysctl for the event capture mask of the TP block's logic analyzer. dev.t5nex.<n>.misc.tp_la_mask dev.t4nex.<n>.misc.tp_la_mask
2016-03-11 01:54:43 +00:00
static int
sysctl_tp_la_mask(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
struct tp_params *tpp = &sc->params.tp;
u_int mask;
int rc;
mask = tpp->la_mask >> 16;
rc = sysctl_handle_int(oidp, &mask, 0, req);
if (rc != 0 || req->newptr == NULL)
return (rc);
if (mask > 0xffff)
return (EINVAL);
tpp->la_mask = mask << 16;
t4_set_reg_field(sc, A_TP_DBG_LA_CONFIG, 0xffff0000U, tpp->la_mask);
return (0);
}
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
struct field_desc {
const char *name;
u_int start;
u_int width;
};
static void
field_desc_show(struct sbuf *sb, uint64_t v, const struct field_desc *f)
{
char buf[32];
int line_size = 0;
while (f->name) {
uint64_t mask = (1ULL << f->width) - 1;
int len = snprintf(buf, sizeof(buf), "%s: %ju", f->name,
((uintmax_t)v >> f->start) & mask);
if (line_size + len >= 79) {
line_size = 8;
sbuf_printf(sb, "\n ");
}
sbuf_printf(sb, "%s ", buf);
line_size += len + 1;
f++;
}
sbuf_printf(sb, "\n");
}
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
static const struct field_desc tp_la0[] = {
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
{ "RcfOpCodeOut", 60, 4 },
{ "State", 56, 4 },
{ "WcfState", 52, 4 },
{ "RcfOpcSrcOut", 50, 2 },
{ "CRxError", 49, 1 },
{ "ERxError", 48, 1 },
{ "SanityFailed", 47, 1 },
{ "SpuriousMsg", 46, 1 },
{ "FlushInputMsg", 45, 1 },
{ "FlushInputCpl", 44, 1 },
{ "RssUpBit", 43, 1 },
{ "RssFilterHit", 42, 1 },
{ "Tid", 32, 10 },
{ "InitTcb", 31, 1 },
{ "LineNumber", 24, 7 },
{ "Emsg", 23, 1 },
{ "EdataOut", 22, 1 },
{ "Cmsg", 21, 1 },
{ "CdataOut", 20, 1 },
{ "EreadPdu", 19, 1 },
{ "CreadPdu", 18, 1 },
{ "TunnelPkt", 17, 1 },
{ "RcfPeerFin", 16, 1 },
{ "RcfReasonOut", 12, 4 },
{ "TxCchannel", 10, 2 },
{ "RcfTxChannel", 8, 2 },
{ "RxEchannel", 6, 2 },
{ "RcfRxChannel", 5, 1 },
{ "RcfDataOutSrdy", 4, 1 },
{ "RxDvld", 3, 1 },
{ "RxOoDvld", 2, 1 },
{ "RxCongestion", 1, 1 },
{ "TxCongestion", 0, 1 },
{ NULL }
};
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
static const struct field_desc tp_la1[] = {
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
{ "CplCmdIn", 56, 8 },
{ "CplCmdOut", 48, 8 },
{ "ESynOut", 47, 1 },
{ "EAckOut", 46, 1 },
{ "EFinOut", 45, 1 },
{ "ERstOut", 44, 1 },
{ "SynIn", 43, 1 },
{ "AckIn", 42, 1 },
{ "FinIn", 41, 1 },
{ "RstIn", 40, 1 },
{ "DataIn", 39, 1 },
{ "DataInVld", 38, 1 },
{ "PadIn", 37, 1 },
{ "RxBufEmpty", 36, 1 },
{ "RxDdp", 35, 1 },
{ "RxFbCongestion", 34, 1 },
{ "TxFbCongestion", 33, 1 },
{ "TxPktSumSrdy", 32, 1 },
{ "RcfUlpType", 28, 4 },
{ "Eread", 27, 1 },
{ "Ebypass", 26, 1 },
{ "Esave", 25, 1 },
{ "Static0", 24, 1 },
{ "Cread", 23, 1 },
{ "Cbypass", 22, 1 },
{ "Csave", 21, 1 },
{ "CPktOut", 20, 1 },
{ "RxPagePoolFull", 18, 2 },
{ "RxLpbkPkt", 17, 1 },
{ "TxLpbkPkt", 16, 1 },
{ "RxVfValid", 15, 1 },
{ "SynLearned", 14, 1 },
{ "SetDelEntry", 13, 1 },
{ "SetInvEntry", 12, 1 },
{ "CpcmdDvld", 11, 1 },
{ "CpcmdSave", 10, 1 },
{ "RxPstructsFull", 8, 2 },
{ "EpcmdDvld", 7, 1 },
{ "EpcmdFlush", 6, 1 },
{ "EpcmdTrimPrefix", 5, 1 },
{ "EpcmdTrimPostfix", 4, 1 },
{ "ERssIp4Pkt", 3, 1 },
{ "ERssIp6Pkt", 2, 1 },
{ "ERssTcpUdpPkt", 1, 1 },
{ "ERssFceFipPkt", 0, 1 },
{ NULL }
};
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
static const struct field_desc tp_la2[] = {
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
{ "CplCmdIn", 56, 8 },
{ "MpsVfVld", 55, 1 },
{ "MpsPf", 52, 3 },
{ "MpsVf", 44, 8 },
{ "SynIn", 43, 1 },
{ "AckIn", 42, 1 },
{ "FinIn", 41, 1 },
{ "RstIn", 40, 1 },
{ "DataIn", 39, 1 },
{ "DataInVld", 38, 1 },
{ "PadIn", 37, 1 },
{ "RxBufEmpty", 36, 1 },
{ "RxDdp", 35, 1 },
{ "RxFbCongestion", 34, 1 },
{ "TxFbCongestion", 33, 1 },
{ "TxPktSumSrdy", 32, 1 },
{ "RcfUlpType", 28, 4 },
{ "Eread", 27, 1 },
{ "Ebypass", 26, 1 },
{ "Esave", 25, 1 },
{ "Static0", 24, 1 },
{ "Cread", 23, 1 },
{ "Cbypass", 22, 1 },
{ "Csave", 21, 1 },
{ "CPktOut", 20, 1 },
{ "RxPagePoolFull", 18, 2 },
{ "RxLpbkPkt", 17, 1 },
{ "TxLpbkPkt", 16, 1 },
{ "RxVfValid", 15, 1 },
{ "SynLearned", 14, 1 },
{ "SetDelEntry", 13, 1 },
{ "SetInvEntry", 12, 1 },
{ "CpcmdDvld", 11, 1 },
{ "CpcmdSave", 10, 1 },
{ "RxPstructsFull", 8, 2 },
{ "EpcmdDvld", 7, 1 },
{ "EpcmdFlush", 6, 1 },
{ "EpcmdTrimPrefix", 5, 1 },
{ "EpcmdTrimPostfix", 4, 1 },
{ "ERssIp4Pkt", 3, 1 },
{ "ERssIp6Pkt", 2, 1 },
{ "ERssTcpUdpPkt", 1, 1 },
{ "ERssFceFipPkt", 0, 1 },
{ NULL }
};
static void
tp_la_show(struct sbuf *sb, uint64_t *p, int idx)
{
field_desc_show(sb, *p, tp_la0);
}
static void
tp_la_show2(struct sbuf *sb, uint64_t *p, int idx)
{
if (idx)
sbuf_printf(sb, "\n");
field_desc_show(sb, p[0], tp_la0);
if (idx < (TPLA_SIZE / 2 - 1) || p[1] != ~0ULL)
field_desc_show(sb, p[1], tp_la0);
}
static void
tp_la_show3(struct sbuf *sb, uint64_t *p, int idx)
{
if (idx)
sbuf_printf(sb, "\n");
field_desc_show(sb, p[0], tp_la0);
if (idx < (TPLA_SIZE / 2 - 1) || p[1] != ~0ULL)
field_desc_show(sb, p[1], (p[0] & (1 << 17)) ? tp_la2 : tp_la1);
}
static int
sysctl_tp_la(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
struct sbuf *sb;
uint64_t *buf, *p;
int rc;
u_int i, inc;
void (*show_func)(struct sbuf *, uint64_t *, int);
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req);
if (sb == NULL)
return (ENOMEM);
buf = malloc(TPLA_SIZE * sizeof(uint64_t), M_CXGBE, M_ZERO | M_WAITOK);
t4_tp_read_la(sc, buf, NULL);
p = buf;
switch (G_DBGLAMODE(t4_read_reg(sc, A_TP_DBG_LA_CONFIG))) {
case 2:
inc = 2;
show_func = tp_la_show2;
break;
case 3:
inc = 2;
show_func = tp_la_show3;
break;
default:
inc = 1;
show_func = tp_la_show;
}
for (i = 0; i < TPLA_SIZE / inc; i++, p += inc)
(*show_func)(sb, p, i);
rc = sbuf_finish(sb);
sbuf_delete(sb);
free(buf, M_CXGBE);
return (rc);
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
static int
sysctl_tx_rate(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
struct sbuf *sb;
int rc;
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
u64 nrate[MAX_NCHAN], orate[MAX_NCHAN];
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
Firmware device log. # sysctl dev.t4nex.0.devlog MFC after: mdf's sysctl+sbuf changes are MFC'd
2011-05-31 23:49:13 +00:00
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
sb = sbuf_new_for_sysctl(NULL, NULL, 256, req);
if (sb == NULL)
return (ENOMEM);
Firmware device log. # sysctl dev.t4nex.0.devlog MFC after: mdf's sysctl+sbuf changes are MFC'd
2011-05-31 23:49:13 +00:00
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
t4_get_chan_txrate(sc, nrate, orate);
cxgbe(4): Very basic T6 awareness. This is part of ongoing work to update to the latest internal shared code. - Add a chip_params structure to keep track of hardware constants for all generations of Terminators handled by cxgbe. - Update t4_hw_pci_read_cfg4 to work with T6. - Update the hardware debug sysctls (hidden within dev.<tNnex>.<n>.misc.*) to work with T6. Most of the changes are in the decoders for the CIM logic analyzer and the MPS TCAM. - Acquire the regwin lock around indirect register accesses. Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-04 13:11:13 +00:00
if (sc->chip_params->nchan > 2) {
sbuf_printf(sb, " channel 0 channel 1"
" channel 2 channel 3\n");
sbuf_printf(sb, "NIC B/s: %10ju %10ju %10ju %10ju\n",
nrate[0], nrate[1], nrate[2], nrate[3]);
sbuf_printf(sb, "Offload B/s: %10ju %10ju %10ju %10ju",
orate[0], orate[1], orate[2], orate[3]);
} else {
sbuf_printf(sb, " channel 0 channel 1\n");
sbuf_printf(sb, "NIC B/s: %10ju %10ju\n",
nrate[0], nrate[1]);
sbuf_printf(sb, "Offload B/s: %10ju %10ju",
orate[0], orate[1]);
}
Firmware device log. # sysctl dev.t4nex.0.devlog MFC after: mdf's sysctl+sbuf changes are MFC'd
2011-05-31 23:49:13 +00:00
rc = sbuf_finish(sb);
sbuf_delete(sb);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
Firmware device log. # sysctl dev.t4nex.0.devlog MFC after: mdf's sysctl+sbuf changes are MFC'd
2011-05-31 23:49:13 +00:00
return (rc);
}
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
cxgbe(4): Some more debug sysctls. These work on both T4 and T5 based cards. dev.t5nex.0.misc.cim_ma_la: CIM MA logic analyzer dev.t5nex.0.misc.cim_pif_la: CIM PIF logic analyzer dev.t5nex.0.misc.mps_tcam: MPS TCAM entries dev.t5nex.0.misc.tp_la: TP logic analyzer dev.t5nex.0.misc.ulprx_la: ULPRX logic analyzer Obtained from: Chelsio MFC after: 1 week
2013-06-01 02:07:37 +00:00
static int
sysctl_ulprx_la(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
struct sbuf *sb;
uint32_t *buf, *p;
int rc, i;
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req);
if (sb == NULL)
return (ENOMEM);
buf = malloc(ULPRX_LA_SIZE * 8 * sizeof(uint32_t), M_CXGBE,
M_ZERO | M_WAITOK);
t4_ulprx_read_la(sc, buf);
p = buf;
sbuf_printf(sb, " Pcmd Type Message"
" Data");
for (i = 0; i < ULPRX_LA_SIZE; i++, p += 8) {
sbuf_printf(sb, "\n%08x%08x %4x %08x %08x%08x%08x%08x",
p[1], p[0], p[2], p[3], p[7], p[6], p[5], p[4]);
}
rc = sbuf_finish(sb);
sbuf_delete(sb);
free(buf, M_CXGBE);
return (rc);
}
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
static int
Cosmetic change (s/wrwc/wcwr/;s/WRWC/WCWR/). MFC after: 3 days.
2013-04-11 22:49:29 +00:00
sysctl_wcwr_stats(SYSCTL_HANDLER_ARGS)
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
{
struct adapter *sc = arg1;
struct sbuf *sb;
int rc, v;
cxgbe(4): Deal with the slightly different SGE_STAT_CFG in T6. Sponsored by: Chelsio Communications
2016-09-11 17:57:53 +00:00
MPASS(chip_id(sc) >= CHELSIO_T5);
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req);
if (sb == NULL)
return (ENOMEM);
v = t4_read_reg(sc, A_SGE_STAT_CFG);
if (G_STATSOURCE_T5(v) == 7) {
cxgbe(4): Deal with the slightly different SGE_STAT_CFG in T6. Sponsored by: Chelsio Communications
2016-09-11 17:57:53 +00:00
int mode;
mode = is_t5(sc) ? G_STATMODE(v) : G_T6_STATMODE(v);
if (mode == 0) {
Get rid of a couple of stray \n's. MFC after: 3 days.
2013-04-11 21:17:49 +00:00
sbuf_printf(sb, "total %d, incomplete %d",
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
t4_read_reg(sc, A_SGE_STAT_TOTAL),
t4_read_reg(sc, A_SGE_STAT_MATCH));
cxgbe(4): Deal with the slightly different SGE_STAT_CFG in T6. Sponsored by: Chelsio Communications
2016-09-11 17:57:53 +00:00
} else if (mode == 1) {
Get rid of a couple of stray \n's. MFC after: 3 days.
2013-04-11 21:17:49 +00:00
sbuf_printf(sb, "total %d, data overflow %d",
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
t4_read_reg(sc, A_SGE_STAT_TOTAL),
t4_read_reg(sc, A_SGE_STAT_MATCH));
cxgbe(4): Deal with the slightly different SGE_STAT_CFG in T6. Sponsored by: Chelsio Communications
2016-09-11 17:57:53 +00:00
} else {
sbuf_printf(sb, "unknown mode %d", mode);
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
}
}
rc = sbuf_finish(sb);
sbuf_delete(sb);
return (rc);
}
cxgbe(4): Provide information about traffic classes in the sysctl mib. Sponsored by: Chelsio Communications
2016-06-07 06:42:35 +00:00
cxgbe(4): sysctls to display the local and intr CPUs for the adapter. The driver assumes the list can change (even though it does't right now) and queries it every time the sysctl runs. sysctl dev.<nexus>.<inst>.local_cpus sysctl dev.<nexus>.<inst>.intr_cpus sysctl dev.t6nex.0.local_cpus sysctl dev.t6nex.0.intr_cpus Sponsored by: Chelsio Communications
2018-06-15 18:04:44 +00:00
static int
sysctl_cpus(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
enum cpu_sets op = arg2;
cpuset_t cpuset;
struct sbuf *sb;
int i, rc;
MPASS(op == LOCAL_CPUS || op == INTR_CPUS);
CPU_ZERO(&cpuset);
rc = bus_get_cpus(sc->dev, op, sizeof(cpuset), &cpuset);
if (rc != 0)
return (rc);
rc = sysctl_wire_old_buffer(req, 0);
if (rc != 0)
return (rc);
sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req);
if (sb == NULL)
return (ENOMEM);
CPU_FOREACH(i)
sbuf_printf(sb, "%d ", i);
rc = sbuf_finish(sb);
sbuf_delete(sb);
return (rc);
}
Fix the following gcc warnings on sparc64, when TCP_OFFLOAD is not defined: sys/dev/cxgbe/t4_main.c:7474: warning: 'sysctl_tp_tick' defined but not used sys/dev/cxgbe/t4_main.c:7505: warning: 'sysctl_tp_dack_timer' defined but not used sys/dev/cxgbe/t4_main.c:7519: warning: 'sysctl_tp_timer' defined but not used This just adds a bunch of #ifdef TCP_OFFLOAD in the right places. Reviewed by: np Differential Revision: https://reviews.freebsd.org/D5620
2016-03-12 18:38:51 +00:00
#ifdef TCP_OFFLOAD
Support for TLS offload of TOE connections on T6 adapters. The TOE engine in Chelsio T6 adapters supports offloading of TLS encryption and TCP segmentation for offloaded connections. Sockets using TLS are required to use a set of custom socket options to upload RX and TX keys to the NIC and to enable RX processing. Currently these socket options are implemented as TCP options in the vendor specific range. A patched OpenSSL library will be made available in a port / package for use with the TLS TOE support. TOE sockets can either offload both transmit and reception of TLS records or just transmit. TLS offload (both RX and TX) is enabled by setting the dev.t6nex.<x>.tls sysctl to 1 and requires TOE to be enabled on the relevant interface. Transmit offload can be used on any "normal" or TLS TOE socket by using the custom socket option to program a transmit key. This permits most TOE sockets to transparently offload TLS when applications use a patched SSL library (e.g. using LD_LIBRARY_PATH to request use of a patched OpenSSL library). Receive offload can only be used with TOE sockets using the TLS mode. The dev.t6nex.0.toe.tls_rx_ports sysctl can be set to a list of TCP port numbers. Any connection with either a local or remote port number in that list will be created as a TLS socket rather than a plain TOE socket. Note that although this sysctl accepts an arbitrary list of port numbers, the sysctl(8) tool is only able to set sysctl nodes to a single value. A TLS socket will hang without receiving data if used by an application that is not using a patched SSL library. Thus, the tls_rx_ports node should be used with care. For a server mostly concerned with offloading TLS transmit, this node is not needed as plain TOE sockets will fall back to software crypto when using an unpatched SSL library. New per-interface statistics nodes are added giving counts of TLS packets and payload bytes (payload bytes do not include TLS headers or authentication tags/MACs) offloaded via the TOE engine, e.g.: dev.cc.0.stats.rx_tls_octets: 149 dev.cc.0.stats.rx_tls_records: 13 dev.cc.0.stats.tx_tls_octets: 26501823 dev.cc.0.stats.tx_tls_records: 1620 TLS transmit work requests are constructed by a new variant of t4_push_frames() called t4_push_tls_records() in tom/t4_tls.c. TLS transmit work requests require a buffer containing IVs. If the IVs are too large to fit into the work request, a separate buffer is allocated when constructing a work request. This buffer is associated with the transmit descriptor and freed when the descriptor is ACKed by the adapter. Received TLS frames use two new CPL messages. The first message is a CPL_TLS_DATA containing the decryped payload of a single TLS record. The handler places the mbuf containing the received payload on an mbufq in the TOE pcb. The second message is a CPL_RX_TLS_CMP message which includes a copy of the TLS header and indicates if there were any errors. The handler for this message places the TLS header into the socket buffer followed by the saved mbuf with the payload data. Both of these handlers are contained in tom/t4_tls.c. A few routines were exposed from t4_cpl_io.c for use by t4_tls.c including send_rx_credits(), a new send_rx_modulate(), and t4_close_conn(). TLS keys for both transmit and receive are stored in onboard memory in the NIC in the "TLS keys" memory region. In some cases a TLS socket can hang with pending data available in the NIC that is not delivered to the host. As a workaround, TLS sockets are more aggressive about sending CPL_RX_DATA_ACK messages anytime that any data is read from a TLS socket. In addition, a fallback timer will periodically send CPL_RX_DATA_ACK messages to the NIC for connections that are still in the handshake phase. Once the connection has finished the handshake and programmed RX keys via the socket option, the timer is stopped. A new function select_ulp_mode() is used to determine what sub-mode a given TOE socket should use (plain TOE, DDP, or TLS). The existing set_tcpddp_ulp_mode() function has been renamed to set_ulp_mode() and handles initialization of TLS-specific state when necessary in addition to DDP-specific state. Since TLS sockets do not receive individual TCP segments but always receive full TLS records, they can receive more data than is available in the current window (e.g. if a 16k TLS record is received but the socket buffer is itself 16k). To cope with this, just drop the window to 0 when this happens, but track the overage and "eat" the overage as it is read from the socket buffer not opening the window (or adding rx_credits) for the overage bytes. Reviewed by: np (earlier version) Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D14529
2018-03-13 23:05:51 +00:00
static int
sysctl_tls_rx_ports(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
int *old_ports, *new_ports;
int i, new_count, rc;
if (req->newptr == NULL && req->oldptr == NULL)
return (SYSCTL_OUT(req, NULL, imax(sc->tt.num_tls_rx_ports, 1) *
sizeof(sc->tt.tls_rx_ports[0])));
rc = begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK, "t4tlsrx");
if (rc)
return (rc);
if (sc->tt.num_tls_rx_ports == 0) {
i = -1;
rc = SYSCTL_OUT(req, &i, sizeof(i));
} else
rc = SYSCTL_OUT(req, sc->tt.tls_rx_ports,
sc->tt.num_tls_rx_ports * sizeof(sc->tt.tls_rx_ports[0]));
if (rc == 0 && req->newptr != NULL) {
new_count = req->newlen / sizeof(new_ports[0]);
new_ports = malloc(new_count * sizeof(new_ports[0]), M_CXGBE,
M_WAITOK);
rc = SYSCTL_IN(req, new_ports, new_count *
sizeof(new_ports[0]));
if (rc)
goto err;
/* Allow setting to a single '-1' to clear the list. */
if (new_count == 1 && new_ports[0] == -1) {
ADAPTER_LOCK(sc);
old_ports = sc->tt.tls_rx_ports;
sc->tt.tls_rx_ports = NULL;
sc->tt.num_tls_rx_ports = 0;
ADAPTER_UNLOCK(sc);
free(old_ports, M_CXGBE);
} else {
for (i = 0; i < new_count; i++) {
if (new_ports[i] < 1 ||
new_ports[i] > IPPORT_MAX) {
rc = EINVAL;
goto err;
}
}
ADAPTER_LOCK(sc);
old_ports = sc->tt.tls_rx_ports;
sc->tt.tls_rx_ports = new_ports;
sc->tt.num_tls_rx_ports = new_count;
ADAPTER_UNLOCK(sc);
free(old_ports, M_CXGBE);
new_ports = NULL;
}
err:
free(new_ports, M_CXGBE);
}
end_synchronized_op(sc, 0);
return (rc);
}
cxgbe(4): sysctls to display the TOE's TCP timers. cask:~# sysctl -d dev.t5nex.0.toe dev.t5nex.0.toe.finwait2_timer: FINWAIT2 timer (us) dev.t5nex.0.toe.initial_srtt: Initial SRTT (us) dev.t5nex.0.toe.keepalive_intvl: Keepidle interval (us) dev.t5nex.0.toe.keepalive_idle: Keepidle idle timer (us) dev.t5nex.0.toe.persist_max: Persist timer max (us) dev.t5nex.0.toe.persist_min: Persist timer min (us) dev.t5nex.0.toe.rexmt_max: Retransmit max (us) dev.t5nex.0.toe.rexmt_min: Retransmit min (us) dev.t5nex.0.toe.dack_timer: DACK timer (us) dev.t5nex.0.toe.dack_tick: DACK tick (us) dev.t5nex.0.toe.timestamp_tick: TCP timestamp tick (us) dev.t5nex.0.toe.timer_tick: TP timer tick (us) ... cask:~# sysctl dev.t5nex.0.toe dev.t5nex.0.toe.finwait2_timer: 9765440 dev.t5nex.0.toe.initial_srtt: 244128 dev.t5nex.0.toe.keepalive_intvl: 73240800 dev.t5nex.0.toe.keepalive_idle: 7031116800 dev.t5nex.0.toe.persist_max: 9765440 dev.t5nex.0.toe.persist_min: 976544 dev.t5nex.0.toe.rexmt_max: 9765440 dev.t5nex.0.toe.rexmt_min: 244128 dev.t5nex.0.toe.dack_timer: 19520 dev.t5nex.0.toe.dack_tick: 32.768 dev.t5nex.0.toe.timestamp_tick: 1048.576 dev.t5nex.0.toe.timer_tick: 32.768 ...
2016-03-11 23:24:04 +00:00
static void
unit_conv(char *buf, size_t len, u_int val, u_int factor)
{
u_int rem = val % factor;
if (rem == 0)
snprintf(buf, len, "%u", val / factor);
else {
while (rem % 10 == 0)
rem /= 10;
snprintf(buf, len, "%u.%u", val / factor, rem);
}
}
static int
sysctl_tp_tick(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
char buf[16];
u_int res, re;
u_int cclk_ps = 1000000000 / sc->params.vpd.cclk;
res = t4_read_reg(sc, A_TP_TIMER_RESOLUTION);
switch (arg2) {
case 0:
/* timer_tick */
re = G_TIMERRESOLUTION(res);
break;
case 1:
/* TCP timestamp tick */
re = G_TIMESTAMPRESOLUTION(res);
break;
case 2:
/* DACK tick */
re = G_DELAYEDACKRESOLUTION(res);
break;
default:
return (EDOOFUS);
}
unit_conv(buf, sizeof(buf), (cclk_ps << re), 1000000);
return (sysctl_handle_string(oidp, buf, sizeof(buf), req));
}
static int
sysctl_tp_dack_timer(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
u_int res, dack_re, v;
u_int cclk_ps = 1000000000 / sc->params.vpd.cclk;
res = t4_read_reg(sc, A_TP_TIMER_RESOLUTION);
dack_re = G_DELAYEDACKRESOLUTION(res);
v = ((cclk_ps << dack_re) / 1000000) * t4_read_reg(sc, A_TP_DACK_TIMER);
return (sysctl_handle_int(oidp, &v, 0, req));
}
static int
sysctl_tp_timer(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
int reg = arg2;
u_int tre;
u_long tp_tick_us, v;
u_int cclk_ps = 1000000000 / sc->params.vpd.cclk;
MPASS(reg == A_TP_RXT_MIN || reg == A_TP_RXT_MAX ||
cxgbe(4): Fix an always-true assertion (reported by PVS-Studio). sys/dev/cxgbe/t4_main.c: PVS-Studio: Expression is Always True (CWE-571) (3) PR: 217746 Submitted by: Svyatoslav (razmyslov at viva64 com) MFC after: 1 week
2017-03-13 17:16:29 +00:00
reg == A_TP_PERS_MIN || reg == A_TP_PERS_MAX ||
reg == A_TP_KEEP_IDLE || reg == A_TP_KEEP_INTVL ||
reg == A_TP_INIT_SRTT || reg == A_TP_FINWAIT2_TIMER);
cxgbe(4): sysctls to display the TOE's TCP timers. cask:~# sysctl -d dev.t5nex.0.toe dev.t5nex.0.toe.finwait2_timer: FINWAIT2 timer (us) dev.t5nex.0.toe.initial_srtt: Initial SRTT (us) dev.t5nex.0.toe.keepalive_intvl: Keepidle interval (us) dev.t5nex.0.toe.keepalive_idle: Keepidle idle timer (us) dev.t5nex.0.toe.persist_max: Persist timer max (us) dev.t5nex.0.toe.persist_min: Persist timer min (us) dev.t5nex.0.toe.rexmt_max: Retransmit max (us) dev.t5nex.0.toe.rexmt_min: Retransmit min (us) dev.t5nex.0.toe.dack_timer: DACK timer (us) dev.t5nex.0.toe.dack_tick: DACK tick (us) dev.t5nex.0.toe.timestamp_tick: TCP timestamp tick (us) dev.t5nex.0.toe.timer_tick: TP timer tick (us) ... cask:~# sysctl dev.t5nex.0.toe dev.t5nex.0.toe.finwait2_timer: 9765440 dev.t5nex.0.toe.initial_srtt: 244128 dev.t5nex.0.toe.keepalive_intvl: 73240800 dev.t5nex.0.toe.keepalive_idle: 7031116800 dev.t5nex.0.toe.persist_max: 9765440 dev.t5nex.0.toe.persist_min: 976544 dev.t5nex.0.toe.rexmt_max: 9765440 dev.t5nex.0.toe.rexmt_min: 244128 dev.t5nex.0.toe.dack_timer: 19520 dev.t5nex.0.toe.dack_tick: 32.768 dev.t5nex.0.toe.timestamp_tick: 1048.576 dev.t5nex.0.toe.timer_tick: 32.768 ...
2016-03-11 23:24:04 +00:00
tre = G_TIMERRESOLUTION(t4_read_reg(sc, A_TP_TIMER_RESOLUTION));
tp_tick_us = (cclk_ps << tre) / 1000000;
if (reg == A_TP_INIT_SRTT)
v = tp_tick_us * G_INITSRTT(t4_read_reg(sc, reg));
else
v = tp_tick_us * t4_read_reg(sc, reg);
return (sysctl_handle_long(oidp, &v, 0, req));
}
cxgbe(4): Display some more TOE parameters related to retransmission and keepalive in the sysctl MIB. Provide tunables to change some of these parameters. These are supposed to be setup by the firmware so these tunables are for experimentation only. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-24 19:17:13 +00:00
/*
* All fields in TP_SHIFT_CNT are 4b and the starting location of the field is
* passed to this function.
*/
static int
sysctl_tp_shift_cnt(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
int idx = arg2;
u_int v;
MPASS(idx >= 0 && idx <= 24);
v = (t4_read_reg(sc, A_TP_SHIFT_CNT) >> idx) & 0xf;
return (sysctl_handle_int(oidp, &v, 0, req));
}
static int
sysctl_tp_backoff(SYSCTL_HANDLER_ARGS)
{
struct adapter *sc = arg1;
int idx = arg2;
u_int shift, v, r;
MPASS(idx >= 0 && idx < 16);
r = A_TP_TCP_BACKOFF_REG0 + (idx & ~3);
shift = (idx & 3) << 3;
v = (t4_read_reg(sc, r) >> shift) & M_TIMERBACKOFFINDEX0;
return (sysctl_handle_int(oidp, &v, 0, req));
}
cxgbe(4): Provide knobs to set the holdoff parameters of TOE rx queues separately from NIC rx queues instead of using the same parameters for both types of queues. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-10-05 07:18:16 +00:00
static int
sysctl_holdoff_tmr_idx_ofld(SYSCTL_HANDLER_ARGS)
{
struct vi_info *vi = arg1;
struct adapter *sc = vi->pi->adapter;
int idx, rc, i;
struct sge_ofld_rxq *ofld_rxq;
uint8_t v;
idx = vi->ofld_tmr_idx;
rc = sysctl_handle_int(oidp, &idx, 0, req);
if (rc != 0 || req->newptr == NULL)
return (rc);
if (idx < 0 || idx >= SGE_NTIMERS)
return (EINVAL);
rc = begin_synchronized_op(sc, vi, HOLD_LOCK | SLEEP_OK | INTR_OK,
"t4otmr");
if (rc)
return (rc);
v = V_QINTR_TIMER_IDX(idx) | V_QINTR_CNT_EN(vi->ofld_pktc_idx != -1);
for_each_ofld_rxq(vi, i, ofld_rxq) {
#ifdef atomic_store_rel_8
atomic_store_rel_8(&ofld_rxq->iq.intr_params, v);
#else
ofld_rxq->iq.intr_params = v;
#endif
}
vi->ofld_tmr_idx = idx;
end_synchronized_op(sc, LOCK_HELD);
return (0);
}
static int
sysctl_holdoff_pktc_idx_ofld(SYSCTL_HANDLER_ARGS)
{
struct vi_info *vi = arg1;
struct adapter *sc = vi->pi->adapter;
int idx, rc;
idx = vi->ofld_pktc_idx;
rc = sysctl_handle_int(oidp, &idx, 0, req);
if (rc != 0 || req->newptr == NULL)
return (rc);
if (idx < -1 || idx >= SGE_NCOUNTERS)
return (EINVAL);
rc = begin_synchronized_op(sc, vi, HOLD_LOCK | SLEEP_OK | INTR_OK,
"t4opktc");
if (rc)
return (rc);
if (vi->flags & VI_INIT_DONE)
rc = EBUSY; /* cannot be changed once the queues are created */
else
vi->ofld_pktc_idx = idx;
end_synchronized_op(sc, LOCK_HELD);
return (rc);
}
Fix the following gcc warnings on sparc64, when TCP_OFFLOAD is not defined: sys/dev/cxgbe/t4_main.c:7474: warning: 'sysctl_tp_tick' defined but not used sys/dev/cxgbe/t4_main.c:7505: warning: 'sysctl_tp_dack_timer' defined but not used sys/dev/cxgbe/t4_main.c:7519: warning: 'sysctl_tp_timer' defined but not used This just adds a bunch of #ifdef TCP_OFFLOAD in the right places. Reviewed by: np Differential Revision: https://reviews.freebsd.org/D5620
2016-03-12 18:38:51 +00:00
#endif
cxgbe(4): sysctls to display the TOE's TCP timers. cask:~# sysctl -d dev.t5nex.0.toe dev.t5nex.0.toe.finwait2_timer: FINWAIT2 timer (us) dev.t5nex.0.toe.initial_srtt: Initial SRTT (us) dev.t5nex.0.toe.keepalive_intvl: Keepidle interval (us) dev.t5nex.0.toe.keepalive_idle: Keepidle idle timer (us) dev.t5nex.0.toe.persist_max: Persist timer max (us) dev.t5nex.0.toe.persist_min: Persist timer min (us) dev.t5nex.0.toe.rexmt_max: Retransmit max (us) dev.t5nex.0.toe.rexmt_min: Retransmit min (us) dev.t5nex.0.toe.dack_timer: DACK timer (us) dev.t5nex.0.toe.dack_tick: DACK tick (us) dev.t5nex.0.toe.timestamp_tick: TCP timestamp tick (us) dev.t5nex.0.toe.timer_tick: TP timer tick (us) ... cask:~# sysctl dev.t5nex.0.toe dev.t5nex.0.toe.finwait2_timer: 9765440 dev.t5nex.0.toe.initial_srtt: 244128 dev.t5nex.0.toe.keepalive_intvl: 73240800 dev.t5nex.0.toe.keepalive_idle: 7031116800 dev.t5nex.0.toe.persist_max: 9765440 dev.t5nex.0.toe.persist_min: 976544 dev.t5nex.0.toe.rexmt_max: 9765440 dev.t5nex.0.toe.rexmt_min: 244128 dev.t5nex.0.toe.dack_timer: 19520 dev.t5nex.0.toe.dack_tick: 32.768 dev.t5nex.0.toe.timestamp_tick: 1048.576 dev.t5nex.0.toe.timer_tick: 32.768 ...
2016-03-11 23:24:04 +00:00
- driver ioctl to get SGE context for any given queue. - sysctls to display the context id, cidx, and pidx of all kinds of queues. MFC after: 3 days
2011-06-11 04:50:54 +00:00
static int
get_sge_context(struct adapter *sc, struct t4_sge_context *cntxt)
{
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
int rc;
- driver ioctl to get SGE context for any given queue. - sysctls to display the context id, cidx, and pidx of all kinds of queues. MFC after: 3 days
2011-06-11 04:50:54 +00:00
if (cntxt->cid > M_CTXTQID)
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
return (EINVAL);
- driver ioctl to get SGE context for any given queue. - sysctls to display the context id, cidx, and pidx of all kinds of queues. MFC after: 3 days
2011-06-11 04:50:54 +00:00
if (cntxt->mem_id != CTXT_EGRESS && cntxt->mem_id != CTXT_INGRESS &&
cntxt->mem_id != CTXT_FLM && cntxt->mem_id != CTXT_CNM)
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
return (EINVAL);
- driver ioctl to get SGE context for any given queue. - sysctls to display the context id, cidx, and pidx of all kinds of queues. MFC after: 3 days
2011-06-11 04:50:54 +00:00
Do not hold locks around hardware context reads. MFC after: 3 days
2013-02-09 00:35:28 +00:00
rc = begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK, "t4ctxt");
if (rc)
return (rc);
- driver ioctl to get SGE context for any given queue. - sysctls to display the context id, cidx, and pidx of all kinds of queues. MFC after: 3 days
2011-06-11 04:50:54 +00:00
if (sc->flags & FW_OK) {
Do not hold locks around hardware context reads. MFC after: 3 days
2013-02-09 00:35:28 +00:00
rc = -t4_sge_ctxt_rd(sc, sc->mbox, cntxt->cid, cntxt->mem_id,
&cntxt->data[0]);
if (rc == 0)
goto done;
- driver ioctl to get SGE context for any given queue. - sysctls to display the context id, cidx, and pidx of all kinds of queues. MFC after: 3 days
2011-06-11 04:50:54 +00:00
}
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
/*
* Read via firmware failed or wasn't even attempted. Read directly via
* the backdoor.
*/
Do not hold locks around hardware context reads. MFC after: 3 days
2013-02-09 00:35:28 +00:00
rc = -t4_sge_ctxt_rd_bd(sc, cntxt->cid, cntxt->mem_id, &cntxt->data[0]);
done:
end_synchronized_op(sc, 0);
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
return (rc);
}
static int
load_fw(struct adapter *sc, struct t4_data *fw)
{
int rc;
uint8_t *fw_data;
- driver ioctl to get SGE context for any given queue. - sysctls to display the context id, cidx, and pidx of all kinds of queues. MFC after: 3 days
2011-06-11 04:50:54 +00:00
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
rc = begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK, "t4ldfw");
if (rc)
return (rc);
cxgbe(4): Add two new debug flags -- one to allow manual firmware install after full initialization, and another to disable the TCB cache (T6+). The latter works as a tunable only. Note that debug_flags are for debugging only and should not be set normally. MFC after: 1 week Sponsored by: Chelsio Communications
2017-08-30 23:41:04 +00:00
/*
* The firmware, with the sole exception of the memory parity error
* handler, runs from memory and not flash. It is almost always safe to
* install a new firmware on a running system. Just set bit 1 in
* hw.cxgbe.dflags or dev.<nexus>.<n>.dflags first.
*/
if (sc->flags & FULL_INIT_DONE &&
(sc->debug_flags & DF_LOAD_FW_ANYTIME) == 0) {
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
rc = EBUSY;
goto done;
- driver ioctl to get SGE context for any given queue. - sysctls to display the context id, cidx, and pidx of all kinds of queues. MFC after: 3 days
2011-06-11 04:50:54 +00:00
}
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
fw_data = malloc(fw->len, M_CXGBE, M_WAITOK);
if (fw_data == NULL) {
rc = ENOMEM;
goto done;
}
rc = copyin(fw->data, fw_data, fw->len);
if (rc == 0)
rc = -t4_load_fw(sc, fw_data, fw->len);
free(fw_data, M_CXGBE);
done:
end_synchronized_op(sc, 0);
- driver ioctl to get SGE context for any given queue. - sysctls to display the context id, cidx, and pidx of all kinds of queues. MFC after: 3 days
2011-06-11 04:50:54 +00:00
return (rc);
}
cxgbe(4): Add an ioctl to copy a firmware config file to the card's flash.
2016-10-07 19:02:39 +00:00
static int
load_cfg(struct adapter *sc, struct t4_data *cfg)
{
int rc;
uint8_t *cfg_data = NULL;
rc = begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK, "t4ldcf");
if (rc)
return (rc);
if (cfg->len == 0) {
/* clear */
rc = -t4_load_cfg(sc, NULL, 0);
goto done;
}
cfg_data = malloc(cfg->len, M_CXGBE, M_WAITOK);
if (cfg_data == NULL) {
rc = ENOMEM;
goto done;
}
rc = copyin(cfg->data, cfg_data, cfg->len);
if (rc == 0)
rc = -t4_load_cfg(sc, cfg_data, cfg->len);
free(cfg_data, M_CXGBE);
done:
end_synchronized_op(sc, 0);
return (rc);
}
cxgbe(4): New ioctls to flash bootrom and boot config to the card. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-18 00:50:58 +00:00
static int
load_boot(struct adapter *sc, struct t4_bootrom *br)
{
int rc;
uint8_t *br_data = NULL;
u_int offset;
if (br->len > 1024 * 1024)
return (EFBIG);
if (br->pf_offset == 0) {
/* pfidx */
if (br->pfidx_addr > 7)
return (EINVAL);
offset = G_OFFSET(t4_read_reg(sc, PF_REG(br->pfidx_addr,
A_PCIE_PF_EXPROM_OFST)));
} else if (br->pf_offset == 1) {
/* offset */
offset = G_OFFSET(br->pfidx_addr);
} else {
return (EINVAL);
}
rc = begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK, "t4ldbr");
if (rc)
return (rc);
if (br->len == 0) {
/* clear */
rc = -t4_load_boot(sc, NULL, offset, 0);
goto done;
}
br_data = malloc(br->len, M_CXGBE, M_WAITOK);
if (br_data == NULL) {
rc = ENOMEM;
goto done;
}
rc = copyin(br->data, br_data, br->len);
if (rc == 0)
rc = -t4_load_boot(sc, br_data, offset, br->len);
free(br_data, M_CXGBE);
done:
end_synchronized_op(sc, 0);
return (rc);
}
static int
load_bootcfg(struct adapter *sc, struct t4_data *bc)
{
int rc;
uint8_t *bc_data = NULL;
rc = begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK, "t4ldcf");
if (rc)
return (rc);
if (bc->len == 0) {
/* clear */
rc = -t4_load_bootcfg(sc, NULL, 0);
goto done;
}
bc_data = malloc(bc->len, M_CXGBE, M_WAITOK);
if (bc_data == NULL) {
rc = ENOMEM;
goto done;
}
rc = copyin(bc->data, bc_data, bc->len);
if (rc == 0)
rc = -t4_load_bootcfg(sc, bc_data, bc->len);
free(bc_data, M_CXGBE);
done:
end_synchronized_op(sc, 0);
return (rc);
}
cxgbe(4): Initial import of the "collect" component of Chelsio unified debug (cudbg) code, hooked up to the main driver via an ioctl. The ioctl can be used to collect the chip's internal state in a compressed dump file. These dumps can be decoded with the "view" component of cudbg. Obtained from: Chelsio Communications MFC after: 2 months Sponsored by: Chelsio Communications
2017-08-03 14:43:30 +00:00
static int
cudbg_dump(struct adapter *sc, struct t4_cudbg_dump *dump)
{
int rc;
struct cudbg_init *cudbg;
void *handle, *buf;
/* buf is large, don't block if no memory is available */
cxgbe(4): Zero out the memory allocated for the debug dump. cudbg_collect seems to expect it this way. MFC after: 1 week Sponsored by: Chelsio Communications
2017-08-30 18:46:38 +00:00
buf = malloc(dump->len, M_CXGBE, M_NOWAIT | M_ZERO);
cxgbe(4): Initial import of the "collect" component of Chelsio unified debug (cudbg) code, hooked up to the main driver via an ioctl. The ioctl can be used to collect the chip's internal state in a compressed dump file. These dumps can be decoded with the "view" component of cudbg. Obtained from: Chelsio Communications MFC after: 2 months Sponsored by: Chelsio Communications
2017-08-03 14:43:30 +00:00
if (buf == NULL)
return (ENOMEM);
handle = cudbg_alloc_handle();
if (handle == NULL) {
rc = ENOMEM;
goto done;
}
cudbg = cudbg_get_init(handle);
cudbg->adap = sc;
cudbg->print = (cudbg_print_cb)printf;
#ifndef notyet
device_printf(sc->dev, "%s: wr_flash %u, len %u, data %p.\n",
__func__, dump->wr_flash, dump->len, dump->data);
#endif
if (dump->wr_flash)
cudbg->use_flash = 1;
MPASS(sizeof(cudbg->dbg_bitmap) == sizeof(dump->bitmap));
memcpy(cudbg->dbg_bitmap, dump->bitmap, sizeof(cudbg->dbg_bitmap));
rc = cudbg_collect(handle, buf, &dump->len);
if (rc != 0)
goto done;
rc = copyout(buf, dump->data, dump->len);
done:
cudbg_free_handle(handle);
free(buf, M_CXGBE);
return (rc);
}
cxgbe(4): Add support for Connection Offload Policy (aka COP). COP allows fine-grained control on whether to offload a TCP connection using t4_tom, and what settings to apply to a connection selected for offload. t4_tom must still be loaded and IFCAP_TOE must still be enabled for full TCP offload to take place on an interface. The difference is that IFCAP_TOE used to be the only knob and would enable TOE for all new connections on the inteface, but now the driver will also consult the COP, if any, before offloading to the hardware TOE. A policy is a plain text file with any number of rules, one per line. Each rule has a "match" part consisting of a socket-type (L = listen, A = active open, P = passive open, D = don't care) and a pcap-filter(7) expression, and a "settings" part that specifies whether to offload the connection or not and the parameters to use if so. The general format of a rule is: [socket-type] expr => settings Example. See cxgbetool(8) for more information. [L] ip && port http => offload [L] port 443 => !offload [L] port ssh => offload [P] src net 192.168/16 && dst port ssh => offload !nagle !timestamp cong newreno [P] dst port ssh => offload !nagle ecn cong tahoe [P] dst port http => offload [A] dst port 443 => offload tls [A] dst net 192.168/16 => offload !timestamp cong highspeed The driver processes the rules for each new listen, active open, or passive open and stops at the first match. There is an implicit rule at the end of every policy that prohibits offload when no rule in the policy matches: [D] all => !offload This is a reworked and expanded version of a patch submitted by Krishnamraju Eraparaju @ Chelsio. Sponsored by: Chelsio Communications
2018-04-14 19:07:56 +00:00
static void
free_offload_policy(struct t4_offload_policy *op)
{
struct offload_rule *r;
int i;
if (op == NULL)
return;
r = &op->rule[0];
for (i = 0; i < op->nrules; i++, r++) {
free(r->bpf_prog.bf_insns, M_CXGBE);
}
free(op->rule, M_CXGBE);
free(op, M_CXGBE);
}
static int
set_offload_policy(struct adapter *sc, struct t4_offload_policy *uop)
{
int i, rc, len;
struct t4_offload_policy *op, *old;
struct bpf_program *bf;
const struct offload_settings *s;
struct offload_rule *r;
void *u;
if (!is_offload(sc))
return (ENODEV);
if (uop->nrules == 0) {
/* Delete installed policies. */
op = NULL;
goto set_policy;
Fix two errors reported by PVS Studio: V646 Consider inspecting the application's logic. It's possible that 'else' keyword is missing. Reviewed by: gallatin, np, pfg Approved by: pfg Differential Revision: https://reviews.freebsd.org/D20396
2019-05-26 12:41:03 +00:00
} else if (uop->nrules > 256) { /* arbitrary */
cxgbe(4): Add support for Connection Offload Policy (aka COP). COP allows fine-grained control on whether to offload a TCP connection using t4_tom, and what settings to apply to a connection selected for offload. t4_tom must still be loaded and IFCAP_TOE must still be enabled for full TCP offload to take place on an interface. The difference is that IFCAP_TOE used to be the only knob and would enable TOE for all new connections on the inteface, but now the driver will also consult the COP, if any, before offloading to the hardware TOE. A policy is a plain text file with any number of rules, one per line. Each rule has a "match" part consisting of a socket-type (L = listen, A = active open, P = passive open, D = don't care) and a pcap-filter(7) expression, and a "settings" part that specifies whether to offload the connection or not and the parameters to use if so. The general format of a rule is: [socket-type] expr => settings Example. See cxgbetool(8) for more information. [L] ip && port http => offload [L] port 443 => !offload [L] port ssh => offload [P] src net 192.168/16 && dst port ssh => offload !nagle !timestamp cong newreno [P] dst port ssh => offload !nagle ecn cong tahoe [P] dst port http => offload [A] dst port 443 => offload tls [A] dst net 192.168/16 => offload !timestamp cong highspeed The driver processes the rules for each new listen, active open, or passive open and stops at the first match. There is an implicit rule at the end of every policy that prohibits offload when no rule in the policy matches: [D] all => !offload This is a reworked and expanded version of a patch submitted by Krishnamraju Eraparaju @ Chelsio. Sponsored by: Chelsio Communications
2018-04-14 19:07:56 +00:00
return (E2BIG);
}
/* Copy userspace offload policy to kernel */
op = malloc(sizeof(*op), M_CXGBE, M_ZERO | M_WAITOK);
op->nrules = uop->nrules;
len = op->nrules * sizeof(struct offload_rule);
op->rule = malloc(len, M_CXGBE, M_ZERO | M_WAITOK);
rc = copyin(uop->rule, op->rule, len);
if (rc) {
free(op->rule, M_CXGBE);
free(op, M_CXGBE);
return (rc);
}
r = &op->rule[0];
for (i = 0; i < op->nrules; i++, r++) {
/* Validate open_type */
if (r->open_type != OPEN_TYPE_LISTEN &&
r->open_type != OPEN_TYPE_ACTIVE &&
r->open_type != OPEN_TYPE_PASSIVE &&
r->open_type != OPEN_TYPE_DONTCARE) {
error:
/*
* Rules 0 to i have malloc'd filters that need to be
* freed. Rules i+1 to nrules have userspace pointers
* and should be left alone.
*/
op->nrules = i;
free_offload_policy(op);
return (rc);
}
/* Validate settings */
s = &r->settings;
if ((s->offload != 0 && s->offload != 1) ||
s->cong_algo < -1 || s->cong_algo > CONG_ALG_HIGHSPEED ||
s->sched_class < -1 ||
s->sched_class >= sc->chip_params->nsched_cls) {
rc = EINVAL;
goto error;
}
bf = &r->bpf_prog;
u = bf->bf_insns; /* userspace ptr */
bf->bf_insns = NULL;
if (bf->bf_len == 0) {
/* legal, matches everything */
continue;
}
len = bf->bf_len * sizeof(*bf->bf_insns);
bf->bf_insns = malloc(len, M_CXGBE, M_ZERO | M_WAITOK);
rc = copyin(u, bf->bf_insns, len);
if (rc != 0)
goto error;
if (!bpf_validate(bf->bf_insns, bf->bf_len)) {
rc = EINVAL;
goto error;
}
}
set_policy:
rw_wlock(&sc->policy_lock);
old = sc->policy;
sc->policy = op;
rw_wunlock(&sc->policy_lock);
free_offload_policy(old);
return (0);
}
cxgbe(4): Add general purpose routines that offer safe access to the chip's memory windows. Convert existing users of these windows to the new routines.
2016-03-10 06:15:31 +00:00
#define MAX_READ_BUF_SIZE (128 * 1024)
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
static int
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
read_card_mem(struct adapter *sc, int win, struct t4_mem_range *mr)
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
{
cxgbe(4): Add general purpose routines that offer safe access to the chip's memory windows. Convert existing users of these windows to the new routines.
2016-03-10 06:15:31 +00:00
uint32_t addr, remaining, n;
uint32_t *buf;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
int rc;
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
uint8_t *dst;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
rc = validate_mem_range(sc, mr->addr, mr->len);
if (rc != 0)
return (rc);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): Add general purpose routines that offer safe access to the chip's memory windows. Convert existing users of these windows to the new routines.
2016-03-10 06:15:31 +00:00
buf = malloc(min(mr->len, MAX_READ_BUF_SIZE), M_CXGBE, M_WAITOK);
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
addr = mr->addr;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
remaining = mr->len;
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
dst = (void *)mr->data;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
while (remaining) {
cxgbe(4): Add general purpose routines that offer safe access to the chip's memory windows. Convert existing users of these windows to the new routines.
2016-03-10 06:15:31 +00:00
n = min(remaining, MAX_READ_BUF_SIZE);
read_via_memwin(sc, 2, addr, buf, n);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
rc = copyout(buf, dst, n);
if (rc != 0)
break;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
dst += n;
remaining -= n;
addr += n;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
free(buf, M_CXGBE);
return (rc);
}
cxgbe(4): Add general purpose routines that offer safe access to the chip's memory windows. Convert existing users of these windows to the new routines.
2016-03-10 06:15:31 +00:00
#undef MAX_READ_BUF_SIZE
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
Add a driver ioctl to read a byte from any device on a port's i2c bus. This lets userspace read arbitrary information from the SFP+ modules etc. on this bus. Reading multiple bytes in the same transaction isn't possible right now. I'll update the driver once the chip's firmware supports this. MFC after: 3 days
2012-10-10 17:13:46 +00:00
static int
read_i2c(struct adapter *sc, struct t4_i2c_data *i2cd)
{
int rc;
if (i2cd->len == 0 || i2cd->port_id >= sc->params.nports)
return (EINVAL);
Allow multi-byte reads in the private CHELSIO_T4_GET_I2C ioctl. The firmware allows up to 48B to be read this way but the driver limits itself to 8B at a time to remain compatible with old cxgbetool binaries. MFC after: 1 week
2014-07-15 01:03:29 +00:00
if (i2cd->len > sizeof(i2cd->data))
return (EFBIG);
Add a driver ioctl to read a byte from any device on a port's i2c bus. This lets userspace read arbitrary information from the SFP+ modules etc. on this bus. Reading multiple bytes in the same transaction isn't possible right now. I'll update the driver once the chip's firmware supports this. MFC after: 3 days
2012-10-10 17:13:46 +00:00
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
rc = begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK, "t4i2crd");
if (rc)
return (rc);
Add a driver ioctl to read a byte from any device on a port's i2c bus. This lets userspace read arbitrary information from the SFP+ modules etc. on this bus. Reading multiple bytes in the same transaction isn't possible right now. I'll update the driver once the chip's firmware supports this. MFC after: 3 days
2012-10-10 17:13:46 +00:00
rc = -t4_i2c_rd(sc, sc->mbox, i2cd->port_id, i2cd->dev_addr,
Allow multi-byte reads in the private CHELSIO_T4_GET_I2C ioctl. The firmware allows up to 48B to be read this way but the driver limits itself to 8B at a time to remain compatible with old cxgbetool binaries. MFC after: 1 week
2014-07-15 01:03:29 +00:00
i2cd->offset, i2cd->len, &i2cd->data[0]);
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
end_synchronized_op(sc, 0);
Add a driver ioctl to read a byte from any device on a port's i2c bus. This lets userspace read arbitrary information from the SFP+ modules etc. on this bus. Reading multiple bytes in the same transaction isn't possible right now. I'll update the driver once the chip's firmware supports this. MFC after: 3 days
2012-10-10 17:13:46 +00:00
return (rc);
}
cxgbe(4): Clear the freelist statistics in the clearstats ioctl. Move all clearstats code into its own function while here. MFC after: 1 week Sponsored by: Chelsio Communications
2019-07-09 22:24:22 +00:00
static int
clear_stats(struct adapter *sc, u_int port_id)
{
int i, v, bg_map;
struct port_info *pi;
struct vi_info *vi;
struct sge_rxq *rxq;
struct sge_txq *txq;
struct sge_wrq *wrq;
#ifdef TCP_OFFLOAD
struct sge_ofld_rxq *ofld_rxq;
#endif
if (port_id >= sc->params.nports)
return (EINVAL);
pi = sc->port[port_id];
if (pi == NULL)
return (EIO);
/* MAC stats */
t4_clr_port_stats(sc, pi->tx_chan);
pi->tx_parse_error = 0;
pi->tnl_cong_drops = 0;
mtx_lock(&sc->reg_lock);
for_each_vi(pi, v, vi) {
if (vi->flags & VI_INIT_DONE)
t4_clr_vi_stats(sc, vi->vin);
}
bg_map = pi->mps_bg_map;
v = 0; /* reuse */
while (bg_map) {
i = ffs(bg_map) - 1;
t4_write_indirect(sc, A_TP_MIB_INDEX, A_TP_MIB_DATA, &v,
1, A_TP_MIB_TNL_CNG_DROP_0 + i);
bg_map &= ~(1 << i);
}
mtx_unlock(&sc->reg_lock);
/*
* Since this command accepts a port, clear stats for
* all VIs on this port.
*/
for_each_vi(pi, v, vi) {
if (vi->flags & VI_INIT_DONE) {
for_each_rxq(vi, i, rxq) {
#if defined(INET) || defined(INET6)
rxq->lro.lro_queued = 0;
rxq->lro.lro_flushed = 0;
#endif
rxq->rxcsum = 0;
rxq->vlan_extraction = 0;
rxq->fl.mbuf_allocated = 0;
rxq->fl.mbuf_inlined = 0;
rxq->fl.cl_allocated = 0;
rxq->fl.cl_recycled = 0;
rxq->fl.cl_fast_recycled = 0;
}
for_each_txq(vi, i, txq) {
txq->txcsum = 0;
txq->tso_wrs = 0;
txq->vlan_insertion = 0;
txq->imm_wrs = 0;
txq->sgl_wrs = 0;
txq->txpkt_wrs = 0;
txq->txpkts0_wrs = 0;
txq->txpkts1_wrs = 0;
txq->txpkts0_pkts = 0;
txq->txpkts1_pkts = 0;
txq->raw_wrs = 0;
mp_ring_reset_stats(txq->r);
}
#if defined(TCP_OFFLOAD) || defined(RATELIMIT)
for_each_ofld_txq(vi, i, wrq) {
wrq->tx_wrs_direct = 0;
wrq->tx_wrs_copied = 0;
}
#endif
#ifdef TCP_OFFLOAD
for_each_ofld_rxq(vi, i, ofld_rxq) {
ofld_rxq->fl.mbuf_allocated = 0;
ofld_rxq->fl.mbuf_inlined = 0;
ofld_rxq->fl.cl_allocated = 0;
ofld_rxq->fl.cl_recycled = 0;
ofld_rxq->fl.cl_fast_recycled = 0;
}
#endif
if (IS_MAIN_VI(vi)) {
wrq = &sc->sge.ctrlq[pi->port_id];
wrq->tx_wrs_direct = 0;
wrq->tx_wrs_copied = 0;
}
}
}
return (0);
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
int
t4_os_find_pci_capability(struct adapter *sc, int cap)
{
Simplify t4_os_find_pci_capability. MFC after: 3 days
2011-05-19 19:37:41 +00:00
int i;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
Simplify t4_os_find_pci_capability. MFC after: 3 days
2011-05-19 19:37:41 +00:00
return (pci_find_cap(sc->dev, cap, &i) == 0 ? i : 0);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
int
t4_os_pci_save_state(struct adapter *sc)
{
device_t dev;
struct pci_devinfo *dinfo;
dev = sc->dev;
dinfo = device_get_ivars(dev);
pci_cfg_save(dev, dinfo, 0);
return (0);
}
int
t4_os_pci_restore_state(struct adapter *sc)
{
device_t dev;
struct pci_devinfo *dinfo;
dev = sc->dev;
dinfo = device_get_ivars(dev);
pci_cfg_restore(dev, dinfo);
return (0);
}
Be sure to stay within the bounds of the mod_str array when displaying the transceiver type.
2011-03-05 04:19:38 +00:00
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
void
cxgbe(4): Save the last reported link parameters and compare them with the current state to determine whether to generate a link-state change notification. This fixes a bug introduced in r321063 that caused the driver to sometimes skip these notifications. Reported by: Jason Eggleston @ LLNW MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-12 14:02:19 +00:00
t4_os_portmod_changed(struct port_info *pi)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
{
cxgbe(4): Maintain one ifmedia per physical port instead of one per Virtual Interface (VI). All autonomous VIs that share a port share the same media. MFC after: 1 week Sponsored by: Chelsio Communications
2017-08-28 21:44:25 +00:00
struct adapter *sc = pi->adapter;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct vi_info *vi;
struct ifnet *ifp;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
static const char *mod_str[] = {
Update header and related code for firmware 1.3.8 MFC after: 3 days
2011-04-01 00:40:24 +00:00
NULL, "LR", "SR", "ER", "TWINAX", "active TWINAX", "LRM"
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
};
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
KASSERT((pi->flags & FIXED_IFMEDIA) == 0,
("%s: port_type %u", __func__, pi->port_type));
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
cxgbe(4): Always use the first and not the last virtual interface associated with a port in begin_synchronized_op. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 01:28:06 +00:00
vi = &pi->vi[0];
cxgbe(4): Maintain one ifmedia per physical port instead of one per Virtual Interface (VI). All autonomous VIs that share a port share the same media. MFC after: 1 week Sponsored by: Chelsio Communications
2017-08-28 21:44:25 +00:00
if (begin_synchronized_op(sc, vi, HOLD_LOCK, "t4mod") == 0) {
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
PORT_LOCK(pi);
cxgbe(4): Link related changes. - Switch to using 32b port/link capabilities in the driver. The 32b format is used internally by firmwares > 1.16.45.0 and the driver will now interact with the firmware in its native format, whether it's 16b or 32b. Note that the 16b format doesn't have room for 50G, 200G, or 400G speeds. - Add a bit in the pause_settings knobs to allow negotiated PAUSE settings to override manual settings. - Ensure that manual link settings persist across an administrative down/up as well as transceiver unplug/replug. - Remove unused is_*G_port() functions. Approved by: re@ (gjb@) MFC after: 1 month Sponsored by: Chelsio Communications
2018-09-25 05:52:42 +00:00
build_medialist(pi);
if (pi->mod_type != FW_PORT_MOD_TYPE_NONE) {
fixup_link_config(pi);
apply_link_config(pi);
}
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
PORT_UNLOCK(pi);
cxgbe(4): Maintain one ifmedia per physical port instead of one per Virtual Interface (VI). All autonomous VIs that share a port share the same media. MFC after: 1 week Sponsored by: Chelsio Communications
2017-08-28 21:44:25 +00:00
end_synchronized_op(sc, LOCK_HELD);
cxgbe(4): Various link/media related improvements. - Deal with changes to port_type, and not just port_mod when a transceiver is changed. This fixes hot swapping of transceivers of different types (QSFP+ or QSA or QSFP28 in a QSFP28 port, SFP+ or SFP28 in a SFP28 port, etc.). - Always refresh media information for ifconfig if the port is down. The firmware does not generate tranceiver-change interrupts unless at least one VI is enabled on the physical port. Before this change ifconfig diplayed potentially stale information for ports that were administratively down. - Always recalculate and reapply L1 config on a transceiver change. - Display PAUSE settings in ifconfig. The driver sysctls for this continue to work as well. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-17 00:42:13 +00:00
}
Fix some bad interaction between cxgbe(4) and lacp lagg(4) that could leave a port permanently disabled when a copper cable is unplugged and then plugged right back in. lacp_linkstate goes looking for the current ifmedia on a link state change and it could get stale information from cxgbe(4) on a module unplug followed by replug. The fix is to process module events before link-state events within the driver, and to always rebuild the ifmedia list on a module change event (instead of rebuilding it lazily). Thanks to asomers@ for the problem report and detailed analysis to go with it. MFC after: 1 week
2014-11-12 23:29:22 +00:00
cxgbe(4): Always use the first and not the last virtual interface associated with a port in begin_synchronized_op. MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-04 01:28:06 +00:00
ifp = vi->ifp;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if (pi->mod_type == FW_PORT_MOD_TYPE_NONE)
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if_printf(ifp, "transceiver unplugged.\n");
Update header and related code for firmware 1.3.8 MFC after: 3 days
2011-04-01 00:40:24 +00:00
else if (pi->mod_type == FW_PORT_MOD_TYPE_UNKNOWN)
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if_printf(ifp, "unknown transceiver inserted.\n");
Update header and related code for firmware 1.3.8 MFC after: 3 days
2011-04-01 00:40:24 +00:00
else if (pi->mod_type == FW_PORT_MOD_TYPE_NOTSUPPORTED)
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if_printf(ifp, "unsupported transceiver inserted.\n");
Use native FreeBSD facilities everywhere except the shared code in common/ MFC after: 1 week
2012-09-13 09:10:10 +00:00
else if (pi->mod_type > 0 && pi->mod_type < nitems(mod_str)) {
cxgbe(4): Various link/media related improvements. - Deal with changes to port_type, and not just port_mod when a transceiver is changed. This fixes hot swapping of transceivers of different types (QSFP+ or QSA or QSFP28 in a QSFP28 port, SFP+ or SFP28 in a SFP28 port, etc.). - Always refresh media information for ifconfig if the port is down. The firmware does not generate tranceiver-change interrupts unless at least one VI is enabled on the physical port. Before this change ifconfig diplayed potentially stale information for ports that were administratively down. - Always recalculate and reapply L1 config on a transceiver change. - Display PAUSE settings in ifconfig. The driver sysctls for this continue to work as well. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-17 00:42:13 +00:00
if_printf(ifp, "%dGbps %s transceiver inserted.\n",
port_top_speed(pi), mod_str[pi->mod_type]);
Be sure to stay within the bounds of the mod_str array when displaying the transceiver type.
2011-03-05 04:19:38 +00:00
} else {
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if_printf(ifp, "transceiver (type %d) inserted.\n",
Be sure to stay within the bounds of the mod_str array when displaying the transceiver type.
2011-03-05 04:19:38 +00:00
pi->mod_type);
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
void
cxgbe(4): Save the last reported link parameters and compare them with the current state to determine whether to generate a link-state change notification. This fixes a bug introduced in r321063 that caused the driver to sometimes skip these notifications. Reported by: Jason Eggleston @ LLNW MFC after: 3 days Sponsored by: Chelsio Communications
2017-08-12 14:02:19 +00:00
t4_os_link_changed(struct port_info *pi)
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
{
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct vi_info *vi;
struct ifnet *ifp;
cxgbe(4): Various link/media related improvements. - Deal with changes to port_type, and not just port_mod when a transceiver is changed. This fixes hot swapping of transceivers of different types (QSFP+ or QSA or QSFP28 in a QSFP28 port, SFP+ or SFP28 in a SFP28 port, etc.). - Always refresh media information for ifconfig if the port is down. The firmware does not generate tranceiver-change interrupts unless at least one VI is enabled on the physical port. Before this change ifconfig diplayed potentially stale information for ports that were administratively down. - Always recalculate and reapply L1 config on a transceiver change. - Display PAUSE settings in ifconfig. The driver sysctls for this continue to work as well. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-17 00:42:13 +00:00
struct link_config *lc;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
int v;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Consider all supported speeds when building the ifmedia list for a port. Fix other related issues while here: - Require port lock for access to link_config. - Allow 100Mbps operation by tracking the speed in Mbps. Yes, really. - New port flag to indicate that the media list is immutable. It will be used in future refinements. This also fixes a bug where the driver reports incorrect media with recent firmwares. MFC after: 2 days Sponsored by: Chelsio Communications
2018-05-30 22:36:09 +00:00
PORT_LOCK_ASSERT_OWNED(pi);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
for_each_vi(pi, v, vi) {
ifp = vi->ifp;
if (ifp == NULL)
continue;
cxgbe(4): Various link/media related improvements. - Deal with changes to port_type, and not just port_mod when a transceiver is changed. This fixes hot swapping of transceivers of different types (QSFP+ or QSA or QSFP28 in a QSFP28 port, SFP+ or SFP28 in a SFP28 port, etc.). - Always refresh media information for ifconfig if the port is down. The firmware does not generate tranceiver-change interrupts unless at least one VI is enabled on the physical port. Before this change ifconfig diplayed potentially stale information for ports that were administratively down. - Always recalculate and reapply L1 config on a transceiver change. - Display PAUSE settings in ifconfig. The driver sysctls for this continue to work as well. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-17 00:42:13 +00:00
lc = &pi->link_cfg;
if (lc->link_ok) {
ifp->if_baudrate = IF_Mbps(lc->speed);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if_link_state_change(ifp, LINK_STATE_UP);
} else {
if_link_state_change(ifp, LINK_STATE_DOWN);
}
- Show the reason why link is down if this information is available. - Display the temperature and PHY firmware version of the BT PHY. MFC after: 1 day
2013-07-05 01:53:51 +00:00
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
void
t4_iterate(void (*func)(struct adapter *, void *), void *arg)
{
struct adapter *sc;
Change t4_list_lock and t4_uld_list_lock from mutexes to sx'es. - tom_uninit had to be reworked not to hold the adapter lock (a mutex) around t4_deactivate_uld, which acquires the uld_list_lock. - the ifc_match for the interface cloner that creates the tracer ifnet had to be reworked as the kernel calls ifc_match with the global if_cloners_mtx held.
2013-08-28 20:59:22 +00:00
sx_slock(&t4_list_lock);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
SLIST_FOREACH(sc, &t4_list, link) {
/*
* func should not make any assumptions about what state sc is
* in - the only guarantee is that sc->sc_lock is a valid lock.
*/
func(sc, arg);
}
Change t4_list_lock and t4_uld_list_lock from mutexes to sx'es. - tom_uninit had to be reworked not to hold the adapter lock (a mutex) around t4_deactivate_uld, which acquires the uld_list_lock. - the ifc_match for the interface cloner that creates the tracer ifnet had to be reworked as the kernel calls ifc_match with the global if_cloners_mtx held.
2013-08-28 20:59:22 +00:00
sx_sunlock(&t4_list_lock);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
static int
t4_ioctl(struct cdev *dev, unsigned long cmd, caddr_t data, int fflag,
struct thread *td)
{
int rc;
struct adapter *sc = dev->si_drv1;
rc = priv_check(td, PRIV_DRIVER);
if (rc != 0)
return (rc);
switch (cmd) {
Modify read/write ioctls to work with 64 bit registers too. MFC after: 3 days
2011-04-07 07:10:42 +00:00
case CHELSIO_T4_GETREG: {
struct t4_reg *edata = (struct t4_reg *)data;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if ((edata->addr & 0x3) != 0 || edata->addr >= sc->mmio_len)
return (EFAULT);
Modify read/write ioctls to work with 64 bit registers too. MFC after: 3 days
2011-04-07 07:10:42 +00:00
if (edata->size == 4)
edata->val = t4_read_reg(sc, edata->addr);
else if (edata->size == 8)
edata->val = t4_read_reg64(sc, edata->addr);
else
return (EINVAL);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
break;
}
Modify read/write ioctls to work with 64 bit registers too. MFC after: 3 days
2011-04-07 07:10:42 +00:00
case CHELSIO_T4_SETREG: {
struct t4_reg *edata = (struct t4_reg *)data;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
if ((edata->addr & 0x3) != 0 || edata->addr >= sc->mmio_len)
return (EFAULT);
Modify read/write ioctls to work with 64 bit registers too. MFC after: 3 days
2011-04-07 07:10:42 +00:00
if (edata->size == 4) {
if (edata->val & 0xffffffff00000000)
return (EINVAL);
t4_write_reg(sc, edata->addr, (uint32_t) edata->val);
} else if (edata->size == 8)
t4_write_reg64(sc, edata->addr, edata->val);
else
return (EINVAL);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
break;
}
case CHELSIO_T4_REGDUMP: {
struct t4_regdump *regs = (struct t4_regdump *)data;
Add support for register dumps on VF devices. - Add handling of VF register sets to t4_get_regs_len() and t4_get_regs(). - While here, use t4_get_regs_len() in the ioctl handler for regdump instead of inlining it. Reviewed by: np Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7484
2016-08-15 17:42:54 +00:00
int reglen = t4_get_regs_len(sc);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
uint8_t *buf;
if (regs->len < reglen) {
regs->len = reglen; /* hint to the caller */
return (ENOBUFS);
}
regs->len = reglen;
buf = malloc(reglen, M_CXGBE, M_WAITOK | M_ZERO);
cxgbe(4): Updated register dumps. - Get the list of registers to read during a regdump from the shared code instead of the OS specific code. This follows a similar move internally. The shared code includes the list for T6. - Update cxgbetool to be able to decode T5 VF, T6, and T6 VF register dumps (and catch up with some updates to T4 and T5 register decode). Obtained from: Chelsio Communications Sponsored by: Chelsio Communications
2016-03-07 21:11:35 +00:00
get_regs(sc, regs, buf);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
rc = copyout(buf, regs->data, reglen);
free(buf, M_CXGBE);
break;
}
T4 packet filtering/steering. - Enable 5-tuple and every-packet lookup. - Setup the default filter mode to allow filtering/steering based on IP protocol, ingress port, inner VLAN ID, IP frag, FCoE, and MPS match type; all combined together. You can also filter based on MAC index, Ethernet type, IP TOS/IPv6 Traffic Class, and outer VLAN ID but you'll have to modify the default filter mode and exclude some of the match-fields in it. IPv4 and IPv6 SIP/DIP/SPORT/DPORT are always available in all filter rules. - Add driver ioctls to get/set the global filter mode. - Add driver ioctls to program and delete hardware filters. A couple of the "switch" actions that rewrite Ethernet and VLAN information and switch the packet out of another port may not work as the L2 code is not yet in place. Everything else, including all "drop" and "pass" rules with RSS or absolute qid, should work. Obtained from: Chelsio Communications
2011-05-05 02:04:56 +00:00
case CHELSIO_T4_GET_FILTER_MODE:
rc = get_filter_mode(sc, (uint32_t *)data);
break;
case CHELSIO_T4_SET_FILTER_MODE:
rc = set_filter_mode(sc, *(uint32_t *)data);
break;
case CHELSIO_T4_GET_FILTER:
rc = get_filter(sc, (struct t4_filter *)data);
break;
case CHELSIO_T4_SET_FILTER:
rc = set_filter(sc, (struct t4_filter *)data);
break;
case CHELSIO_T4_DEL_FILTER:
rc = del_filter(sc, (struct t4_filter *)data);
break;
- driver ioctl to get SGE context for any given queue. - sysctls to display the context id, cidx, and pidx of all kinds of queues. MFC after: 3 days
2011-06-11 04:50:54 +00:00
case CHELSIO_T4_GET_SGE_CONTEXT:
rc = get_sge_context(sc, (struct t4_sge_context *)data);
break;
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
case CHELSIO_T4_LOAD_FW:
rc = load_fw(sc, (struct t4_data *)data);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
break;
case CHELSIO_T4_GET_MEM:
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
rc = read_card_mem(sc, 2, (struct t4_mem_range *)data);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
break;
Add a driver ioctl to read a byte from any device on a port's i2c bus. This lets userspace read arbitrary information from the SFP+ modules etc. on this bus. Reading multiple bytes in the same transaction isn't possible right now. I'll update the driver once the chip's firmware supports this. MFC after: 3 days
2012-10-10 17:13:46 +00:00
case CHELSIO_T4_GET_I2C:
rc = read_i2c(sc, (struct t4_i2c_data *)data);
break;
cxgbe(4): Clear the freelist statistics in the clearstats ioctl. Move all clearstats code into its own function while here. MFC after: 1 week Sponsored by: Chelsio Communications
2019-07-09 22:24:22 +00:00
case CHELSIO_T4_CLEAR_STATS:
rc = clear_stats(sc, *(uint32_t *)data);
Add a driver ioctl to clear a port's MAC statistics. Submitted by: gnn@ MFC after: 3 days
2012-10-10 19:27:40 +00:00
break;
cxgbe(4): T4_SET_SCHED_CLASS and T4_SET_SCHED_QUEUE ioctls to program scheduling classes in the chip and to bind tx queue(s) to a scheduling class respectively. These can be used for various kinds of tx traffic throttling (to force selected tx queues to drain at a fixed Kbps rate, or a % of the port's total bandwidth, or at a fixed pps rate, etc.). Obtained from: Chelsio
2013-12-03 18:34:52 +00:00
case CHELSIO_T4_SCHED_CLASS:
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
rc = t4_set_sched_class(sc, (struct t4_sched_params *)data);
cxgbe(4): T4_SET_SCHED_CLASS and T4_SET_SCHED_QUEUE ioctls to program scheduling classes in the chip and to bind tx queue(s) to a scheduling class respectively. These can be used for various kinds of tx traffic throttling (to force selected tx queues to drain at a fixed Kbps rate, or a % of the port's total bandwidth, or at a fixed pps rate, etc.). Obtained from: Chelsio
2013-12-03 18:34:52 +00:00
break;
case CHELSIO_T4_SCHED_QUEUE:
Chelsio T4/T5 VF driver. The cxgbev/cxlv driver supports Virtual Function devices for Chelsio T4 and T4 adapters. The VF devices share most of their code with the existing PF4 driver (cxgbe/cxl) and as such the VF device driver currently depends on the PF4 driver. Similar to the cxgbe/cxl drivers, the VF driver includes a t4vf/t5vf PCI device driver that attaches to the VF device. It then creates child cxgbev/cxlv devices representing ports assigned to the VF. By default, the PF driver assigns a single port to each VF. t4vf_hw.c contains VF-specific routines from the shared code used to fetch VF-specific parameters from the firmware. t4_vf.c contains the VF-specific PCI device driver and includes its own attach routine. VF devices are required to use a different firmware request when transmitting packets (which in turn requires a different CPL message to encapsulate messages). This alternate firmware request does not permit chaining multiple packets in a single message, so each packet results in a firmware request. In addition, the different CPL message requires more detailed information when enabling hardware checksums, so parse_pkt() on VF devices must examine L2 and L3 headers for all packets (not just TSO packets) for VF devices. Finally, L2 checksums on non-UDP/non-TCP packets do not work reliably (the firmware trashes the IPv4 fragment field), so IPv4 checksums for such packets are calculated in software. Most of the other changes in the non-VF-specific code are to expose various variables and functions private to the PF driver so that they can be used by the VF driver. Note that a limited subset of cxgbetool functions are supported on VF devices including register dumps, scheduler classes, and clearing of statistics. In addition, TOE is not supported on VF devices, only for the PF interfaces. Reviewed by: np MFC after: 2 months Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D7599
2016-09-07 18:13:57 +00:00
rc = t4_set_sched_queue(sc, (struct t4_sched_queue *)data);
cxgbe(4): T4_SET_SCHED_CLASS and T4_SET_SCHED_QUEUE ioctls to program scheduling classes in the chip and to bind tx queue(s) to a scheduling class respectively. These can be used for various kinds of tx traffic throttling (to force selected tx queues to drain at a fixed Kbps rate, or a % of the port's total bandwidth, or at a fixed pps rate, etc.). Obtained from: Chelsio
2013-12-03 18:34:52 +00:00
break;
Add support for packet-sniffing tracers to cxgbe(4). This works with all T4 and T5 based cards and is useful for analyzing TSO, LRO, TOE, and for general purpose monitoring without tapping any cxgbe or cxl ifnet directly. Tracers on the T4/T5 chips provide access to Ethernet frames exactly as they were received from or transmitted on the wire. On transmit, a tracer will capture a frame after TSO segmentation, hw VLAN tag insertion, hw L3 & L4 checksum insertion, etc. It will also capture frames generated by the TCP offload engine (TOE traffic is normally invisible to the kernel). On receive, a tracer will capture a frame before hw VLAN extraction, runt filtering, other badness filtering, before the steering/drop/L2-rewrite filters or the TOE have had a go at it, and of course before sw LRO in the driver. There are 4 tracers on a chip. A tracer can trace only in one direction (tx or rx). For now cxgbetool will set up tracers to capture the first 128B of every transmitted or received frame on a given port. This is a small subset of what the hardware can do. A pseudo ifnet with the same name as the nexus driver (t4nex0 or t5nex0) will be created for tracing. The data delivered to this ifnet is an additional copy made inside the chip. Normal delivery to cxgbe<n> or cxl<n> will be made as usual. /* watch cxl0, which is the first port hanging off t5nex0. */ # cxgbetool t5nex0 tracer 0 tx0 (watch what cxl0 is transmitting) # cxgbetool t5nex0 tracer 1 rx0 (watch what cxl0 is receiving) # cxgbetool t5nex0 tracer list # tcpdump -i t5nex0 <== all that cxl0 sees and puts on the wire If you were doing TSO, a tcpdump on cxl0 may have shown you ~64K "frames" with no L3/L4 checksum but this will show you the frames that were actually transmitted. /* all done */ # cxgbetool t5nex0 tracer 0 disable # cxgbetool t5nex0 tracer 1 disable # cxgbetool t5nex0 tracer list # ifconfig t5nex0 destroy
2013-07-26 22:04:11 +00:00
case CHELSIO_T4_GET_TRACER:
rc = t4_get_tracer(sc, (struct t4_tracer *)data);
break;
case CHELSIO_T4_SET_TRACER:
rc = t4_set_tracer(sc, (struct t4_tracer *)data);
break;
cxgbe(4): Add an ioctl to copy a firmware config file to the card's flash.
2016-10-07 19:02:39 +00:00
case CHELSIO_T4_LOAD_CFG:
rc = load_cfg(sc, (struct t4_data *)data);
break;
cxgbe(4): New ioctls to flash bootrom and boot config to the card. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-07-18 00:50:58 +00:00
case CHELSIO_T4_LOAD_BOOT:
rc = load_boot(sc, (struct t4_bootrom *)data);
break;
case CHELSIO_T4_LOAD_BOOTCFG:
rc = load_bootcfg(sc, (struct t4_data *)data);
break;
cxgbe(4): Initial import of the "collect" component of Chelsio unified debug (cudbg) code, hooked up to the main driver via an ioctl. The ioctl can be used to collect the chip's internal state in a compressed dump file. These dumps can be decoded with the "view" component of cudbg. Obtained from: Chelsio Communications MFC after: 2 months Sponsored by: Chelsio Communications
2017-08-03 14:43:30 +00:00
case CHELSIO_T4_CUDBG_DUMP:
rc = cudbg_dump(sc, (struct t4_cudbg_dump *)data);
break;
cxgbe(4): Add support for Connection Offload Policy (aka COP). COP allows fine-grained control on whether to offload a TCP connection using t4_tom, and what settings to apply to a connection selected for offload. t4_tom must still be loaded and IFCAP_TOE must still be enabled for full TCP offload to take place on an interface. The difference is that IFCAP_TOE used to be the only knob and would enable TOE for all new connections on the inteface, but now the driver will also consult the COP, if any, before offloading to the hardware TOE. A policy is a plain text file with any number of rules, one per line. Each rule has a "match" part consisting of a socket-type (L = listen, A = active open, P = passive open, D = don't care) and a pcap-filter(7) expression, and a "settings" part that specifies whether to offload the connection or not and the parameters to use if so. The general format of a rule is: [socket-type] expr => settings Example. See cxgbetool(8) for more information. [L] ip && port http => offload [L] port 443 => !offload [L] port ssh => offload [P] src net 192.168/16 && dst port ssh => offload !nagle !timestamp cong newreno [P] dst port ssh => offload !nagle ecn cong tahoe [P] dst port http => offload [A] dst port 443 => offload tls [A] dst net 192.168/16 => offload !timestamp cong highspeed The driver processes the rules for each new listen, active open, or passive open and stops at the first match. There is an implicit rule at the end of every policy that prohibits offload when no rule in the policy matches: [D] all => !offload This is a reworked and expanded version of a patch submitted by Krishnamraju Eraparaju @ Chelsio. Sponsored by: Chelsio Communications
2018-04-14 19:07:56 +00:00
case CHELSIO_T4_SET_OFLD_POLICY:
rc = set_offload_policy(sc, (struct t4_offload_policy *)data);
break;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
default:
Various fixes to the t4/5nex character device. - Remove null open/close methods. - Don't set d_flags to 0 explicitly. - Remove t5_cdevsw as the .d_name member isn't really used and doesn't warrant a separate cdevsw just for the name. - Use ENOTTY as the error value for an unknown ioctl request. - Use make_dev_s() to close race with setting si_drv1. Sponsored by: Chelsio Communications
2016-07-29 22:11:29 +00:00
rc = ENOTTY;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
}
return (rc);
}
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
#ifdef TCP_OFFLOAD
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
static int
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
toe_capability(struct vi_info *vi, int enable)
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
{
int rc;
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
struct port_info *pi = vi->pi;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
struct adapter *sc = pi->adapter;
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
ASSERT_SYNCHRONIZED_OP(sc);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
if (!is_offload(sc))
return (ENODEV);
if (enable) {
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if ((vi->ifp->if_capenable & IFCAP_TOE) != 0) {
/* TOE is already enabled. */
return (0);
}
cxgbe(4): fix a test made while enabling TOE. MFC after: 1 week
2015-02-07 01:50:32 +00:00
/*
* We need the port's queues around so that we're able to send
* and receive CPLs to/from the TOE even if the ifnet for this
* port has never been UP'd administratively.
*/
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (!(vi->flags & VI_INIT_DONE)) {
cxgbe(4): Do not bring up an interface when IFCAP_TOE is enabled on it. The interface's queues are functional after VI_INIT_DONE (which is short of interface-up) and that's all that's needed for t4_tom to communicate with the chip. Approved by: re@ (gjb@) Sponsored by: Chelsio Communications
2016-06-29 06:55:30 +00:00
rc = vi_full_init(vi);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (rc)
return (rc);
}
if (!(pi->vi[0].flags & VI_INIT_DONE)) {
cxgbe(4): Do not bring up an interface when IFCAP_TOE is enabled on it. The interface's queues are functional after VI_INIT_DONE (which is short of interface-up) and that's all that's needed for t4_tom to communicate with the chip. Approved by: re@ (gjb@) Sponsored by: Chelsio Communications
2016-06-29 06:55:30 +00:00
rc = vi_full_init(&pi->vi[0]);
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
if (rc)
return (rc);
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
}
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
if (isset(&sc->offload_map, pi->port_id)) {
/* TOE is enabled on another VI of this port. */
pi->uld_vis++;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
return (0);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): tidy up some of the interaction between the Upper Layer Drivers (ULDs) and the base if_cxgbe driver. Track the per-adapter activation of ULDs in a new "active_ulds" field. This was done pretty arbitrarily before this change -- via TOM_INIT_DONE in adapter->flags for TOM, and the (1 << MAX_NPORTS) bit in adapter->offload_map for iWARP. iWARP and hw-accelerated iSCSI rely on the TOE (supported by the TOM ULD). The rules are: a) If the iWARP and/or iSCSI ULDs are available when TOE is enabled then iWARP and/or iSCSI are enabled too. b) When the iWARP and iSCSI modules are loaded they go looking for adapters with TOE enabled and enable themselves on that adapter. c) You cannot deactivate or unload the TOM module from underneath iWARP or iSCSI. Any such attempt will fail with EBUSY. MFC after: 2 weeks
2015-02-08 09:28:55 +00:00
if (!uld_active(sc, ULD_TOM)) {
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
rc = t4_activate_uld(sc, ULD_TOM);
if (rc == EAGAIN) {
log(LOG_WARNING,
"You must kldload t4_tom.ko before trying "
"to enable TOE on a cxgbe interface.\n");
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
if (rc != 0)
return (rc);
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
KASSERT(sc->tom_softc != NULL,
("%s: TOM activated but softc NULL", __func__));
cxgbe(4): tidy up some of the interaction between the Upper Layer Drivers (ULDs) and the base if_cxgbe driver. Track the per-adapter activation of ULDs in a new "active_ulds" field. This was done pretty arbitrarily before this change -- via TOM_INIT_DONE in adapter->flags for TOM, and the (1 << MAX_NPORTS) bit in adapter->offload_map for iWARP. iWARP and hw-accelerated iSCSI rely on the TOE (supported by the TOM ULD). The rules are: a) If the iWARP and/or iSCSI ULDs are available when TOE is enabled then iWARP and/or iSCSI are enabled too. b) When the iWARP and iSCSI modules are loaded they go looking for adapters with TOE enabled and enable themselves on that adapter. c) You cannot deactivate or unload the TOM module from underneath iWARP or iSCSI. Any such attempt will fail with EBUSY. MFC after: 2 weeks
2015-02-08 09:28:55 +00:00
KASSERT(uld_active(sc, ULD_TOM),
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
("%s: TOM activated but flag not set", __func__));
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
cxgbe(4): tidy up some of the interaction between the Upper Layer Drivers (ULDs) and the base if_cxgbe driver. Track the per-adapter activation of ULDs in a new "active_ulds" field. This was done pretty arbitrarily before this change -- via TOM_INIT_DONE in adapter->flags for TOM, and the (1 << MAX_NPORTS) bit in adapter->offload_map for iWARP. iWARP and hw-accelerated iSCSI rely on the TOE (supported by the TOM ULD). The rules are: a) If the iWARP and/or iSCSI ULDs are available when TOE is enabled then iWARP and/or iSCSI are enabled too. b) When the iWARP and iSCSI modules are loaded they go looking for adapters with TOE enabled and enable themselves on that adapter. c) You cannot deactivate or unload the TOM module from underneath iWARP or iSCSI. Any such attempt will fail with EBUSY. MFC after: 2 weeks
2015-02-08 09:28:55 +00:00
/* Activate iWARP and iSCSI too, if the modules are loaded. */
if (!uld_active(sc, ULD_IWARP))
(void) t4_activate_uld(sc, ULD_IWARP);
if (!uld_active(sc, ULD_ISCSI))
(void) t4_activate_uld(sc, ULD_ISCSI);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
pi->uld_vis++;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
setbit(&sc->offload_map, pi->port_id);
} else {
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
pi->uld_vis--;
if (!isset(&sc->offload_map, pi->port_id) || pi->uld_vis > 0)
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
return (0);
cxgbe(4): tidy up some of the interaction between the Upper Layer Drivers (ULDs) and the base if_cxgbe driver. Track the per-adapter activation of ULDs in a new "active_ulds" field. This was done pretty arbitrarily before this change -- via TOM_INIT_DONE in adapter->flags for TOM, and the (1 << MAX_NPORTS) bit in adapter->offload_map for iWARP. iWARP and hw-accelerated iSCSI rely on the TOE (supported by the TOM ULD). The rules are: a) If the iWARP and/or iSCSI ULDs are available when TOE is enabled then iWARP and/or iSCSI are enabled too. b) When the iWARP and iSCSI modules are loaded they go looking for adapters with TOE enabled and enable themselves on that adapter. c) You cannot deactivate or unload the TOM module from underneath iWARP or iSCSI. Any such attempt will fail with EBUSY. MFC after: 2 weeks
2015-02-08 09:28:55 +00:00
KASSERT(uld_active(sc, ULD_TOM),
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
("%s: TOM never initialized?", __func__));
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
clrbit(&sc->offload_map, pi->port_id);
}
return (0);
}
/*
* Add an upper layer driver to the global list.
*/
int
t4_register_uld(struct uld_info *ui)
{
int rc = 0;
struct uld_info *u;
Change t4_list_lock and t4_uld_list_lock from mutexes to sx'es. - tom_uninit had to be reworked not to hold the adapter lock (a mutex) around t4_deactivate_uld, which acquires the uld_list_lock. - the ifc_match for the interface cloner that creates the tracer ifnet had to be reworked as the kernel calls ifc_match with the global if_cloners_mtx held.
2013-08-28 20:59:22 +00:00
sx_xlock(&t4_uld_list_lock);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
SLIST_FOREACH(u, &t4_uld_list, link) {
if (u->uld_id == ui->uld_id) {
rc = EEXIST;
goto done;
}
}
SLIST_INSERT_HEAD(&t4_uld_list, ui, link);
ui->refcount = 0;
done:
Change t4_list_lock and t4_uld_list_lock from mutexes to sx'es. - tom_uninit had to be reworked not to hold the adapter lock (a mutex) around t4_deactivate_uld, which acquires the uld_list_lock. - the ifc_match for the interface cloner that creates the tracer ifnet had to be reworked as the kernel calls ifc_match with the global if_cloners_mtx held.
2013-08-28 20:59:22 +00:00
sx_xunlock(&t4_uld_list_lock);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
return (rc);
}
int
t4_unregister_uld(struct uld_info *ui)
{
int rc = EINVAL;
struct uld_info *u;
Change t4_list_lock and t4_uld_list_lock from mutexes to sx'es. - tom_uninit had to be reworked not to hold the adapter lock (a mutex) around t4_deactivate_uld, which acquires the uld_list_lock. - the ifc_match for the interface cloner that creates the tracer ifnet had to be reworked as the kernel calls ifc_match with the global if_cloners_mtx held.
2013-08-28 20:59:22 +00:00
sx_xlock(&t4_uld_list_lock);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
SLIST_FOREACH(u, &t4_uld_list, link) {
if (u == ui) {
if (ui->refcount > 0) {
rc = EBUSY;
goto done;
}
SLIST_REMOVE(&t4_uld_list, ui, uld_info, link);
rc = 0;
goto done;
}
}
done:
Change t4_list_lock and t4_uld_list_lock from mutexes to sx'es. - tom_uninit had to be reworked not to hold the adapter lock (a mutex) around t4_deactivate_uld, which acquires the uld_list_lock. - the ifc_match for the interface cloner that creates the tracer ifnet had to be reworked as the kernel calls ifc_match with the global if_cloners_mtx held.
2013-08-28 20:59:22 +00:00
sx_xunlock(&t4_uld_list_lock);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
return (rc);
}
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
int
t4_activate_uld(struct adapter *sc, int id)
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
{
cxgbe(4): tidy up some of the interaction between the Upper Layer Drivers (ULDs) and the base if_cxgbe driver. Track the per-adapter activation of ULDs in a new "active_ulds" field. This was done pretty arbitrarily before this change -- via TOM_INIT_DONE in adapter->flags for TOM, and the (1 << MAX_NPORTS) bit in adapter->offload_map for iWARP. iWARP and hw-accelerated iSCSI rely on the TOE (supported by the TOM ULD). The rules are: a) If the iWARP and/or iSCSI ULDs are available when TOE is enabled then iWARP and/or iSCSI are enabled too. b) When the iWARP and iSCSI modules are loaded they go looking for adapters with TOE enabled and enable themselves on that adapter. c) You cannot deactivate or unload the TOM module from underneath iWARP or iSCSI. Any such attempt will fail with EBUSY. MFC after: 2 weeks
2015-02-08 09:28:55 +00:00
int rc;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
struct uld_info *ui;
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
ASSERT_SYNCHRONIZED_OP(sc);
cxgbe(4): tidy up some of the interaction between the Upper Layer Drivers (ULDs) and the base if_cxgbe driver. Track the per-adapter activation of ULDs in a new "active_ulds" field. This was done pretty arbitrarily before this change -- via TOM_INIT_DONE in adapter->flags for TOM, and the (1 << MAX_NPORTS) bit in adapter->offload_map for iWARP. iWARP and hw-accelerated iSCSI rely on the TOE (supported by the TOM ULD). The rules are: a) If the iWARP and/or iSCSI ULDs are available when TOE is enabled then iWARP and/or iSCSI are enabled too. b) When the iWARP and iSCSI modules are loaded they go looking for adapters with TOE enabled and enable themselves on that adapter. c) You cannot deactivate or unload the TOM module from underneath iWARP or iSCSI. Any such attempt will fail with EBUSY. MFC after: 2 weeks
2015-02-08 09:28:55 +00:00
if (id < 0 || id > ULD_MAX)
return (EINVAL);
rc = EAGAIN; /* kldoad the module with this ULD and try again. */
Change t4_list_lock and t4_uld_list_lock from mutexes to sx'es. - tom_uninit had to be reworked not to hold the adapter lock (a mutex) around t4_deactivate_uld, which acquires the uld_list_lock. - the ifc_match for the interface cloner that creates the tracer ifnet had to be reworked as the kernel calls ifc_match with the global if_cloners_mtx held.
2013-08-28 20:59:22 +00:00
sx_slock(&t4_uld_list_lock);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
SLIST_FOREACH(ui, &t4_uld_list, link) {
if (ui->uld_id == id) {
Make sure the adapter's management queue and the event queue are available before any uppper layer driver (TOE, iWARP, or iSCSI) registers with the base cxgbe(4) driver. Submitted by: Hariprasad at chelsio dot com Reviewed by: np@
2014-09-26 18:53:00 +00:00
if (!(sc->flags & FULL_INIT_DONE)) {
rc = adapter_full_init(sc);
if (rc != 0)
cxgbe(4): tidy up some of the interaction between the Upper Layer Drivers (ULDs) and the base if_cxgbe driver. Track the per-adapter activation of ULDs in a new "active_ulds" field. This was done pretty arbitrarily before this change -- via TOM_INIT_DONE in adapter->flags for TOM, and the (1 << MAX_NPORTS) bit in adapter->offload_map for iWARP. iWARP and hw-accelerated iSCSI rely on the TOE (supported by the TOM ULD). The rules are: a) If the iWARP and/or iSCSI ULDs are available when TOE is enabled then iWARP and/or iSCSI are enabled too. b) When the iWARP and iSCSI modules are loaded they go looking for adapters with TOE enabled and enable themselves on that adapter. c) You cannot deactivate or unload the TOM module from underneath iWARP or iSCSI. Any such attempt will fail with EBUSY. MFC after: 2 weeks
2015-02-08 09:28:55 +00:00
break;
Make sure the adapter's management queue and the event queue are available before any uppper layer driver (TOE, iWARP, or iSCSI) registers with the base cxgbe(4) driver. Submitted by: Hariprasad at chelsio dot com Reviewed by: np@
2014-09-26 18:53:00 +00:00
}
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
rc = ui->activate(sc);
cxgbe(4): tidy up some of the interaction between the Upper Layer Drivers (ULDs) and the base if_cxgbe driver. Track the per-adapter activation of ULDs in a new "active_ulds" field. This was done pretty arbitrarily before this change -- via TOM_INIT_DONE in adapter->flags for TOM, and the (1 << MAX_NPORTS) bit in adapter->offload_map for iWARP. iWARP and hw-accelerated iSCSI rely on the TOE (supported by the TOM ULD). The rules are: a) If the iWARP and/or iSCSI ULDs are available when TOE is enabled then iWARP and/or iSCSI are enabled too. b) When the iWARP and iSCSI modules are loaded they go looking for adapters with TOE enabled and enable themselves on that adapter. c) You cannot deactivate or unload the TOM module from underneath iWARP or iSCSI. Any such attempt will fail with EBUSY. MFC after: 2 weeks
2015-02-08 09:28:55 +00:00
if (rc == 0) {
setbit(&sc->active_ulds, id);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
ui->refcount++;
cxgbe(4): tidy up some of the interaction between the Upper Layer Drivers (ULDs) and the base if_cxgbe driver. Track the per-adapter activation of ULDs in a new "active_ulds" field. This was done pretty arbitrarily before this change -- via TOM_INIT_DONE in adapter->flags for TOM, and the (1 << MAX_NPORTS) bit in adapter->offload_map for iWARP. iWARP and hw-accelerated iSCSI rely on the TOE (supported by the TOM ULD). The rules are: a) If the iWARP and/or iSCSI ULDs are available when TOE is enabled then iWARP and/or iSCSI are enabled too. b) When the iWARP and iSCSI modules are loaded they go looking for adapters with TOE enabled and enable themselves on that adapter. c) You cannot deactivate or unload the TOM module from underneath iWARP or iSCSI. Any such attempt will fail with EBUSY. MFC after: 2 weeks
2015-02-08 09:28:55 +00:00
}
break;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
}
cxgbe(4): tidy up some of the interaction between the Upper Layer Drivers (ULDs) and the base if_cxgbe driver. Track the per-adapter activation of ULDs in a new "active_ulds" field. This was done pretty arbitrarily before this change -- via TOM_INIT_DONE in adapter->flags for TOM, and the (1 << MAX_NPORTS) bit in adapter->offload_map for iWARP. iWARP and hw-accelerated iSCSI rely on the TOE (supported by the TOM ULD). The rules are: a) If the iWARP and/or iSCSI ULDs are available when TOE is enabled then iWARP and/or iSCSI are enabled too. b) When the iWARP and iSCSI modules are loaded they go looking for adapters with TOE enabled and enable themselves on that adapter. c) You cannot deactivate or unload the TOM module from underneath iWARP or iSCSI. Any such attempt will fail with EBUSY. MFC after: 2 weeks
2015-02-08 09:28:55 +00:00
Change t4_list_lock and t4_uld_list_lock from mutexes to sx'es. - tom_uninit had to be reworked not to hold the adapter lock (a mutex) around t4_deactivate_uld, which acquires the uld_list_lock. - the ifc_match for the interface cloner that creates the tracer ifnet had to be reworked as the kernel calls ifc_match with the global if_cloners_mtx held.
2013-08-28 20:59:22 +00:00
sx_sunlock(&t4_uld_list_lock);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
return (rc);
}
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
int
t4_deactivate_uld(struct adapter *sc, int id)
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
{
cxgbe(4): tidy up some of the interaction between the Upper Layer Drivers (ULDs) and the base if_cxgbe driver. Track the per-adapter activation of ULDs in a new "active_ulds" field. This was done pretty arbitrarily before this change -- via TOM_INIT_DONE in adapter->flags for TOM, and the (1 << MAX_NPORTS) bit in adapter->offload_map for iWARP. iWARP and hw-accelerated iSCSI rely on the TOE (supported by the TOM ULD). The rules are: a) If the iWARP and/or iSCSI ULDs are available when TOE is enabled then iWARP and/or iSCSI are enabled too. b) When the iWARP and iSCSI modules are loaded they go looking for adapters with TOE enabled and enable themselves on that adapter. c) You cannot deactivate or unload the TOM module from underneath iWARP or iSCSI. Any such attempt will fail with EBUSY. MFC after: 2 weeks
2015-02-08 09:28:55 +00:00
int rc;
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
struct uld_info *ui;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): Add functions to help synchronize "slow" operations (those not on the fast data path) and use them instead of frobbing the adapter lock and busy flag directly. Other changes made while reworking all slow operations: - Wait for the reply to a filter request (add/delete). This guarantees that the operation is complete by the time the ioctl returns. - Tidy up the tid_info structure. - Do not allow the tx queue size to be set to something that's not a power of 2. MFC after: 1 week
2013-01-10 23:56:50 +00:00
ASSERT_SYNCHRONIZED_OP(sc);
cxgbe(4): tidy up some of the interaction between the Upper Layer Drivers (ULDs) and the base if_cxgbe driver. Track the per-adapter activation of ULDs in a new "active_ulds" field. This was done pretty arbitrarily before this change -- via TOM_INIT_DONE in adapter->flags for TOM, and the (1 << MAX_NPORTS) bit in adapter->offload_map for iWARP. iWARP and hw-accelerated iSCSI rely on the TOE (supported by the TOM ULD). The rules are: a) If the iWARP and/or iSCSI ULDs are available when TOE is enabled then iWARP and/or iSCSI are enabled too. b) When the iWARP and iSCSI modules are loaded they go looking for adapters with TOE enabled and enable themselves on that adapter. c) You cannot deactivate or unload the TOM module from underneath iWARP or iSCSI. Any such attempt will fail with EBUSY. MFC after: 2 weeks
2015-02-08 09:28:55 +00:00
if (id < 0 || id > ULD_MAX)
return (EINVAL);
rc = ENXIO;
Change t4_list_lock and t4_uld_list_lock from mutexes to sx'es. - tom_uninit had to be reworked not to hold the adapter lock (a mutex) around t4_deactivate_uld, which acquires the uld_list_lock. - the ifc_match for the interface cloner that creates the tracer ifnet had to be reworked as the kernel calls ifc_match with the global if_cloners_mtx held.
2013-08-28 20:59:22 +00:00
sx_slock(&t4_uld_list_lock);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
SLIST_FOREACH(ui, &t4_uld_list, link) {
if (ui->uld_id == id) {
rc = ui->deactivate(sc);
cxgbe(4): tidy up some of the interaction between the Upper Layer Drivers (ULDs) and the base if_cxgbe driver. Track the per-adapter activation of ULDs in a new "active_ulds" field. This was done pretty arbitrarily before this change -- via TOM_INIT_DONE in adapter->flags for TOM, and the (1 << MAX_NPORTS) bit in adapter->offload_map for iWARP. iWARP and hw-accelerated iSCSI rely on the TOE (supported by the TOM ULD). The rules are: a) If the iWARP and/or iSCSI ULDs are available when TOE is enabled then iWARP and/or iSCSI are enabled too. b) When the iWARP and iSCSI modules are loaded they go looking for adapters with TOE enabled and enable themselves on that adapter. c) You cannot deactivate or unload the TOM module from underneath iWARP or iSCSI. Any such attempt will fail with EBUSY. MFC after: 2 weeks
2015-02-08 09:28:55 +00:00
if (rc == 0) {
clrbit(&sc->active_ulds, id);
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
ui->refcount--;
cxgbe(4): tidy up some of the interaction between the Upper Layer Drivers (ULDs) and the base if_cxgbe driver. Track the per-adapter activation of ULDs in a new "active_ulds" field. This was done pretty arbitrarily before this change -- via TOM_INIT_DONE in adapter->flags for TOM, and the (1 << MAX_NPORTS) bit in adapter->offload_map for iWARP. iWARP and hw-accelerated iSCSI rely on the TOE (supported by the TOM ULD). The rules are: a) If the iWARP and/or iSCSI ULDs are available when TOE is enabled then iWARP and/or iSCSI are enabled too. b) When the iWARP and iSCSI modules are loaded they go looking for adapters with TOE enabled and enable themselves on that adapter. c) You cannot deactivate or unload the TOM module from underneath iWARP or iSCSI. Any such attempt will fail with EBUSY. MFC after: 2 weeks
2015-02-08 09:28:55 +00:00
}
break;
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
cxgbe(4): tidy up some of the interaction between the Upper Layer Drivers (ULDs) and the base if_cxgbe driver. Track the per-adapter activation of ULDs in a new "active_ulds" field. This was done pretty arbitrarily before this change -- via TOM_INIT_DONE in adapter->flags for TOM, and the (1 << MAX_NPORTS) bit in adapter->offload_map for iWARP. iWARP and hw-accelerated iSCSI rely on the TOE (supported by the TOM ULD). The rules are: a) If the iWARP and/or iSCSI ULDs are available when TOE is enabled then iWARP and/or iSCSI are enabled too. b) When the iWARP and iSCSI modules are loaded they go looking for adapters with TOE enabled and enable themselves on that adapter. c) You cannot deactivate or unload the TOM module from underneath iWARP or iSCSI. Any such attempt will fail with EBUSY. MFC after: 2 weeks
2015-02-08 09:28:55 +00:00
Change t4_list_lock and t4_uld_list_lock from mutexes to sx'es. - tom_uninit had to be reworked not to hold the adapter lock (a mutex) around t4_deactivate_uld, which acquires the uld_list_lock. - the ifc_match for the interface cloner that creates the tracer ifnet had to be reworked as the kernel calls ifc_match with the global if_cloners_mtx held.
2013-08-28 20:59:22 +00:00
sx_sunlock(&t4_uld_list_lock);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
return (rc);
}
cxgbe(4): tidy up some of the interaction between the Upper Layer Drivers (ULDs) and the base if_cxgbe driver. Track the per-adapter activation of ULDs in a new "active_ulds" field. This was done pretty arbitrarily before this change -- via TOM_INIT_DONE in adapter->flags for TOM, and the (1 << MAX_NPORTS) bit in adapter->offload_map for iWARP. iWARP and hw-accelerated iSCSI rely on the TOE (supported by the TOM ULD). The rules are: a) If the iWARP and/or iSCSI ULDs are available when TOE is enabled then iWARP and/or iSCSI are enabled too. b) When the iWARP and iSCSI modules are loaded they go looking for adapters with TOE enabled and enable themselves on that adapter. c) You cannot deactivate or unload the TOM module from underneath iWARP or iSCSI. Any such attempt will fail with EBUSY. MFC after: 2 weeks
2015-02-08 09:28:55 +00:00
int
uld_active(struct adapter *sc, int uld_id)
{
MPASS(uld_id >= 0 && uld_id <= ULD_MAX);
return (isset(&sc->active_ulds, uld_id));
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
#endif
cxgbe(4): Allow tunables that control the number of queues to be set to '-n' to tell the driver to create _up to_ 'n' queues if enough cores are available. For example, setting hw.cxgbe.nrxq10g="-32" will result in 16 queues if the system has 16 cores, 32 if it has 32. There is no change in the default number of queues of any type. MFC after: 1 week Sponsored by: Chelsio Communications
2017-02-06 05:19:29 +00:00
/*
* t = ptr to tunable.
* nc = number of CPUs.
* c = compiled in default for that tunable.
*/
static void
calculate_nqueues(int *t, int nc, const int c)
{
int nq;
if (*t > 0)
return;
nq = *t < 0 ? -*t : c;
*t = min(nc, nq);
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
/*
* Come up with reasonable defaults for some of the tunables, provided they're
* not set by the user (in which case we'll use the values as is).
*/
static void
tweak_tunables(void)
{
int nc = mp_ncpus; /* our snapshot of the number of CPUs */
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
if (t4_ntxq < 1) {
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
#ifdef RSS
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
t4_ntxq = rss_getnumbuckets();
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
#else
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
calculate_nqueues(&t4_ntxq, nc, NTXQ);
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
#endif
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): Allow tunables that control the number of queues to be set to '-n' to tell the driver to create _up to_ 'n' queues if enough cores are available. For example, setting hw.cxgbe.nrxq10g="-32" will result in 16 queues if the system has 16 cores, 32 if it has 32. There is no change in the default number of queues of any type. MFC after: 1 week Sponsored by: Chelsio Communications
2017-02-06 05:19:29 +00:00
calculate_nqueues(&t4_ntxq_vi, nc, NTXQ_VI);
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
if (t4_nrxq < 1) {
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
#ifdef RSS
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
t4_nrxq = rss_getnumbuckets();
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
#else
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
calculate_nqueues(&t4_nrxq, nc, NRXQ);
cxgbe(4): support for the kernel RSS option. You need PCBGROUP and RSS in the kernel config to use this. Relnotes: Yes Sponsored by: Chelsio Communications
2015-10-16 01:19:55 +00:00
#endif
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): Allow tunables that control the number of queues to be set to '-n' to tell the driver to create _up to_ 'n' queues if enough cores are available. For example, setting hw.cxgbe.nrxq10g="-32" will result in 16 queues if the system has 16 cores, 32 if it has 32. There is no change in the default number of queues of any type. MFC after: 1 week Sponsored by: Chelsio Communications
2017-02-06 05:19:29 +00:00
calculate_nqueues(&t4_nrxq_vi, nc, NRXQ_VI);
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
cxgbe(4): Allocate offload Tx queues when a card has resources provisioned for NIC_ETHOFLD and the kernel has option RATELIMIT. It is possible to use the chip's offload queues for normal NIC Tx and not just TOE Tx. The difference is that these queues support out of order processing of work requests and have a per-"flowid" mechanism for tracking credits between the driver and hardware. This allows Tx for any number of flows bound to different rate limits to be submitted to a single Tx queue and the work requests for slow flows won't cause HOL blocking for the rest. Sponsored by: Chelsio Communications
2018-05-17 01:42:18 +00:00
#if defined(TCP_OFFLOAD) || defined(RATELIMIT)
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
calculate_nqueues(&t4_nofldtxq, nc, NOFLDTXQ);
cxgbe(4): Allow tunables that control the number of queues to be set to '-n' to tell the driver to create _up to_ 'n' queues if enough cores are available. For example, setting hw.cxgbe.nrxq10g="-32" will result in 16 queues if the system has 16 cores, 32 if it has 32. There is no change in the default number of queues of any type. MFC after: 1 week Sponsored by: Chelsio Communications
2017-02-06 05:19:29 +00:00
calculate_nqueues(&t4_nofldtxq_vi, nc, NOFLDTXQ_VI);
cxgbe(4): Allocate offload Tx queues when a card has resources provisioned for NIC_ETHOFLD and the kernel has option RATELIMIT. It is possible to use the chip's offload queues for normal NIC Tx and not just TOE Tx. The difference is that these queues support out of order processing of work requests and have a per-"flowid" mechanism for tracking credits between the driver and hardware. This allows Tx for any number of flows bound to different rate limits to be submitted to a single Tx queue and the work requests for slow flows won't cause HOL blocking for the rest. Sponsored by: Chelsio Communications
2018-05-17 01:42:18 +00:00
#endif
#ifdef TCP_OFFLOAD
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
calculate_nqueues(&t4_nofldrxq, nc, NOFLDRXQ);
cxgbe(4): Allow tunables that control the number of queues to be set to '-n' to tell the driver to create _up to_ 'n' queues if enough cores are available. For example, setting hw.cxgbe.nrxq10g="-32" will result in 16 queues if the system has 16 cores, 32 if it has 32. There is no change in the default number of queues of any type. MFC after: 1 week Sponsored by: Chelsio Communications
2017-02-06 05:19:29 +00:00
calculate_nqueues(&t4_nofldrxq_vi, nc, NOFLDRXQ_VI);
cxgbe(4): Merge netmap support from the ncxgbe/ncxl interfaces to the vcxgbe/vcxl interfaces and retire the 'n' interfaces. The main cxgbe/cxl interfaces and tunables related to them are not affected by any of this and will continue to operate as usual. The driver used to create an additional 'n' interface for every cxgbe/cxl interface if "device netmap" was in the kernel. The 'n' interface shared the wire with the main interface but was otherwise autonomous (with its own MAC address, etc.). It did not have normal tx/rx but had a specialized netmap-only data path. r291665 added another set of virtual interfaces (the 'v' interfaces) to the driver. These had normal tx/rx but no netmap support. This revision consolidates the features of both the interfaces into the 'v' interface which now has a normal data path, TOE support, and native netmap support. The 'v' interfaces need to be created explicitly with the hw.cxgbe.num_vis tunable. This means "device netmap" will not result in the automatic creation of any virtual interfaces. The following tunables can be used to override the default number of queues allocated for each 'v' interface. nofld* = 0 will disable TOE on the virtual interface and nnm* = 0 to will disable native netmap support. # number of normal NIC queues hw.cxgbe.ntxq_vi hw.cxgbe.nrxq_vi # number of TOE queues hw.cxgbe.nofldtxq_vi hw.cxgbe.nofldrxq_vi # number of netmap queues hw.cxgbe.nnmtxq_vi hw.cxgbe.nnmrxq_vi hw.cxgbe.nnm{t,r}xq{10,1}g tunables have been removed. --- tl;dr version --- The workflow for netmap on cxgbe starting with FreeBSD 11 is: 1) "device netmap" in the kernel config. 2) "hw.cxgbe.num_vis=2" in loader.conf. num_vis > 2 is ok too, you'll end up with multiple autonomous netmap-capable interfaces for every port. 3) "dmesg | grep vcxl | grep netmap" to verify that the interface has netmap queues. 4) Use any of the 'v' interfaces for netmap. pkt-gen -i vcxl<n>... . One major improvement is that the netmap interface has a normal data path as expected. 5) Just ignore the cxl interfaces if you want to use netmap only. No need to bring them up. The vcxl interfaces are completely independent and everything should just work. --------------------- Approved by: re@ (gjb@) Relnotes: Yes Sponsored by: Chelsio Communications
2016-06-23 02:53:00 +00:00
Instruct the firmware not to provision resources for TCP offload if the kernel is being built without TCP_OFFLOAD. But never override toecaps_allowed if it has been set manually.
2012-07-02 20:42:43 +00:00
if (t4_toecaps_allowed == -1)
t4_toecaps_allowed = FW_CAPS_CONFIG_TOE;
cxgbe(4): Enable additional capabilities in the default configuration files. All features with FreeBSD drivers of some kind are now in the default configuration.
2016-03-16 19:43:44 +00:00
if (t4_rdmacaps_allowed == -1) {
t4_rdmacaps_allowed = FW_CAPS_CONFIG_RDMA_RDDP |
FW_CAPS_CONFIG_RDMA_RDMAC;
}
if (t4_iscsicaps_allowed == -1) {
t4_iscsicaps_allowed = FW_CAPS_CONFIG_ISCSI_INITIATOR_PDU |
FW_CAPS_CONFIG_ISCSI_TARGET_PDU |
FW_CAPS_CONFIG_ISCSI_T10DIF;
}
cxgbe(4): Provide knobs to set the holdoff parameters of TOE rx queues separately from NIC rx queues instead of using the same parameters for both types of queues. MFC after: 2 weeks Sponsored by: Chelsio Communications
2017-10-05 07:18:16 +00:00
if (t4_tmr_idx_ofld < 0 || t4_tmr_idx_ofld >= SGE_NTIMERS)
t4_tmr_idx_ofld = TMR_IDX_OFLD;
if (t4_pktc_idx_ofld < -1 || t4_pktc_idx_ofld >= SGE_NCOUNTERS)
t4_pktc_idx_ofld = PKTC_IDX_OFLD;
Instruct the firmware not to provision resources for TCP offload if the kernel is being built without TCP_OFFLOAD. But never override toecaps_allowed if it has been set manually.
2012-07-02 20:42:43 +00:00
#else
if (t4_toecaps_allowed == -1)
t4_toecaps_allowed = 0;
cxgbe(4): Enable additional capabilities in the default configuration files. All features with FreeBSD drivers of some kind are now in the default configuration.
2016-03-16 19:43:44 +00:00
if (t4_rdmacaps_allowed == -1)
t4_rdmacaps_allowed = 0;
if (t4_iscsicaps_allowed == -1)
t4_iscsicaps_allowed = 0;
cxgbe(4): netmap support for Terminator 5 (T5) based 10G/40G cards. Netmap gets its own hardware-assisted virtual interface and won't take over or disrupt the "normal" interface in any way. You can use both simultaneously. For kernels with DEV_NETMAP, cxgbe(4) carves out an ncxl<N> interface (note the 'n' prefix) in the hardware to accompany each cxl<N> interface. These two ifnet's per port share the same wire but really are separate interfaces in the hardware and software. Each gets its own L2 MAC addresses (unicast and multicast), MTU, checksum caps, etc. You should run netmap on the 'n' interfaces only, that's what they are for. With this, pkt-gen is able to transmit > 45Mpps out of a single 40G port of a T580 card. 2 port tx is at ~56Mpps total (28M + 28M) as of now. Single port receive is at 33Mpps but this is very much a work in progress. I expect it to be closer to 40Mpps once done. In any case the current effort can already saturate multiple 10G ports of a T5 card at the smallest legal packet size. T4 gear is totally untested. trantor:~# ./pkt-gen -i ncxl0 -f tx -D 00:07:43:ab:cd:ef 881.952141 main [1621] interface is ncxl0 881.952250 extract_ip_range [275] range is 10.0.0.1:0 to 10.0.0.1:0 881.952253 extract_ip_range [275] range is 10.1.0.1:0 to 10.1.0.1:0 881.962540 main [1804] mapped 334980KB at 0x801dff000 Sending on netmap:ncxl0: 4 queues, 1 threads and 1 cpus. 10.0.0.1 -> 10.1.0.1 (00:00:00:00:00:00 -> 00:07:43:ab:cd:ef) 881.962562 main [1882] Sending 512 packets every 0.000000000 s 881.962563 main [1884] Wait 2 secs for phy reset 884.088516 main [1886] Ready... 884.088535 nm_open [457] overriding ifname ncxl0 ringid 0x0 flags 0x1 884.088607 sender_body [996] start 884.093246 sender_body [1064] drop copy 885.090435 main_thread [1418] 45206353 pps (45289533 pkts in 1001840 usec) 886.091600 main_thread [1418] 45322792 pps (45375593 pkts in 1001165 usec) 887.092435 main_thread [1418] 45313992 pps (45351784 pkts in 1000834 usec) 888.094434 main_thread [1418] 45315765 pps (45406397 pkts in 1002000 usec) 889.095434 main_thread [1418] 45333218 pps (45378551 pkts in 1001000 usec) 890.097434 main_thread [1418] 45315247 pps (45405877 pkts in 1002000 usec) 891.099434 main_thread [1418] 45326515 pps (45417168 pkts in 1002000 usec) 892.101434 main_thread [1418] 45333039 pps (45423705 pkts in 1002000 usec) 893.103434 main_thread [1418] 45324105 pps (45414708 pkts in 1001999 usec) 894.105434 main_thread [1418] 45318042 pps (45408723 pkts in 1002001 usec) 895.106434 main_thread [1418] 45332430 pps (45377762 pkts in 1001000 usec) 896.107434 main_thread [1418] 45338072 pps (45383410 pkts in 1001000 usec) ... Relnotes: Yes Sponsored by: Chelsio Communications.
2014-05-27 18:18:41 +00:00
#endif
#ifdef DEV_NETMAP
cxgbe(4): Allow tunables that control the number of queues to be set to '-n' to tell the driver to create _up to_ 'n' queues if enough cores are available. For example, setting hw.cxgbe.nrxq10g="-32" will result in 16 queues if the system has 16 cores, 32 if it has 32. There is no change in the default number of queues of any type. MFC after: 1 week Sponsored by: Chelsio Communications
2017-02-06 05:19:29 +00:00
calculate_nqueues(&t4_nnmtxq_vi, nc, NNMTXQ_VI);
calculate_nqueues(&t4_nnmrxq_vi, nc, NNMRXQ_VI);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
#endif
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
if (t4_tmr_idx < 0 || t4_tmr_idx >= SGE_NTIMERS)
t4_tmr_idx = TMR_IDX;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
cxgbe(4): Combine all _10g and _1g tunables and drop the suffix from their names. The finer-grained knobs weren't practically useful. Sponsored by: Chelsio Communications
2017-11-15 23:48:02 +00:00
if (t4_pktc_idx < -1 || t4_pktc_idx >= SGE_NCOUNTERS)
t4_pktc_idx = PKTC_IDX;
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
if (t4_qsize_txq < 128)
t4_qsize_txq = 128;
if (t4_qsize_rxq < 128)
t4_qsize_rxq = 128;
while (t4_qsize_rxq & 7)
t4_qsize_rxq++;
t4_intr_types &= INTR_MSIX | INTR_MSI | INTR_INTX;
cxgbe(4): Sanitize t4_num_vis during MOD_LOAD like all other t4_* tunables. Add num_vis to the intrs_and_queues structure as it affects the number of interrupts requested and queues created. In future cfg_itype_and_nqueues might lower it incrementally instead of going straight to 1 when enough interrupts aren't available. Sponsored by: Chelsio Communications
2017-11-16 01:33:53 +00:00
/*
* Number of VIs to create per-port. The first VI is the "main" regular
* VI for the port. The rest are additional virtual interfaces on the
* same physical port. Note that the main VI does not have native
* netmap support but the extra VIs do.
*
* Limit the number of VIs per port to the number of available
* MAC addresses per port.
*/
if (t4_num_vis < 1)
t4_num_vis = 1;
if (t4_num_vis > nitems(vi_mac_funcs)) {
t4_num_vis = nitems(vi_mac_funcs);
printf("cxgbe: number of VIs limited to %d\n", t4_num_vis);
}
cxgbe(4): Add a knob to enable/disable PCIe relaxed ordering. Disable it by default when running on Intel CPUs. This is a crude fix for the performance issues alluded to in these Linux commits: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=87e09cdec4dae08acdb4aa49beb793c19d73e73e https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=a99b646afa8a02571ea298bedca6592d818229cd MFC after: 1 week Sponsored by: Chelsio Communications
2018-01-03 19:24:57 +00:00
if (pcie_relaxed_ordering < 0 || pcie_relaxed_ordering > 2) {
pcie_relaxed_ordering = 1;
#if defined(__i386__) || defined(__amd64__)
if (cpu_vendor_id == CPU_VENDOR_INTEL)
pcie_relaxed_ordering = 0;
#endif
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
Add a 'show t4 tcb <nexus> <tid>' command to dump a TCB from DDB. This allows the contents of a TCB to be extracted from a T4/T5 card in DDB after a panic.
2016-04-10 05:06:58 +00:00
#ifdef DDB
static void
t4_dump_tcb(struct adapter *sc, int tid)
{
uint32_t base, i, j, off, pf, reg, save, tcb_addr, win_pos;
reg = PCIE_MEM_ACCESS_REG(A_PCIE_MEM_ACCESS_OFFSET, 2);
save = t4_read_reg(sc, reg);
base = sc->memwin[2].mw_base;
/* Dump TCB for the tid */
tcb_addr = t4_read_reg(sc, A_TP_CMM_TCB_BASE);
tcb_addr += tid * TCB_SIZE;
if (is_t4(sc)) {
pf = 0;
win_pos = tcb_addr & ~0xf; /* start must be 16B aligned */
} else {
pf = V_PFNUM(sc->pf);
win_pos = tcb_addr & ~0x7f; /* start must be 128B aligned */
}
t4_write_reg(sc, reg, win_pos | pf);
t4_read_reg(sc, reg);
off = tcb_addr - win_pos;
for (i = 0; i < 4; i++) {
uint32_t buf[8];
for (j = 0; j < 8; j++, off += 4)
buf[j] = htonl(t4_read_reg(sc, base + off));
db_printf("%08x %08x %08x %08x %08x %08x %08x %08x\n",
buf[0], buf[1], buf[2], buf[3], buf[4], buf[5], buf[6],
buf[7]);
}
t4_write_reg(sc, reg, save);
t4_read_reg(sc, reg);
}
Add a 'show t4 devlog <nexus>' DDB command. This command displays the adapter's firmware device log similar to the dev.<nexus>.misc.devlog sysctl. Sponsored by: Chelsio Communications
2016-04-10 06:19:26 +00:00
static void
t4_dump_devlog(struct adapter *sc)
{
struct devlog_params *dparams = &sc->params.devlog;
struct fw_devlog_e e;
int i, first, j, m, nentries, rc;
uint64_t ftstamp = UINT64_MAX;
if (dparams->start == 0) {
db_printf("devlog params not valid\n");
return;
}
nentries = dparams->size / sizeof(struct fw_devlog_e);
m = fwmtype_to_hwmtype(dparams->memtype);
/* Find the first entry. */
first = -1;
for (i = 0; i < nentries && !db_pager_quit; i++) {
rc = -t4_mem_read(sc, m, dparams->start + i * sizeof(e),
sizeof(e), (void *)&e);
if (rc != 0)
break;
if (e.timestamp == 0)
break;
e.timestamp = be64toh(e.timestamp);
if (e.timestamp < ftstamp) {
ftstamp = e.timestamp;
first = i;
}
}
if (first == -1)
return;
i = first;
do {
rc = -t4_mem_read(sc, m, dparams->start + i * sizeof(e),
sizeof(e), (void *)&e);
if (rc != 0)
return;
if (e.timestamp == 0)
return;
e.timestamp = be64toh(e.timestamp);
e.seqno = be32toh(e.seqno);
for (j = 0; j < 8; j++)
e.params[j] = be32toh(e.params[j]);
db_printf("%10d %15ju %8s %8s ",
e.seqno, e.timestamp,
(e.level < nitems(devlog_level_strings) ?
devlog_level_strings[e.level] : "UNKNOWN"),
(e.facility < nitems(devlog_facility_strings) ?
devlog_facility_strings[e.facility] : "UNKNOWN"));
db_printf(e.fmt, e.params[0], e.params[1], e.params[2],
e.params[3], e.params[4], e.params[5], e.params[6],
e.params[7]);
if (++i == nentries)
i = 0;
} while (i != first && !db_pager_quit);
}
Add a 'show t4 tcb <nexus> <tid>' command to dump a TCB from DDB. This allows the contents of a TCB to be extracted from a T4/T5 card in DDB after a panic.
2016-04-10 05:06:58 +00:00
static struct command_table db_t4_table = LIST_HEAD_INITIALIZER(db_t4_table);
_DB_SET(_show, t4, NULL, db_show_table, 0, &db_t4_table);
Add a 'show t4 devlog <nexus>' DDB command. This command displays the adapter's firmware device log similar to the dev.<nexus>.misc.devlog sysctl. Sponsored by: Chelsio Communications
2016-04-10 06:19:26 +00:00
DB_FUNC(devlog, db_show_devlog, db_t4_table, CS_OWN, NULL)
{
device_t dev;
int t;
bool valid;
valid = false;
t = db_read_token();
if (t == tIDENT) {
dev = device_lookup_by_name(db_tok_string);
valid = true;
}
db_skip_to_eol();
if (!valid) {
db_printf("usage: show t4 devlog <nexus>\n");
return;
}
if (dev == NULL) {
db_printf("device not found\n");
return;
}
t4_dump_devlog(device_get_softc(dev));
}
Add a 'show t4 tcb <nexus> <tid>' command to dump a TCB from DDB. This allows the contents of a TCB to be extracted from a T4/T5 card in DDB after a panic.
2016-04-10 05:06:58 +00:00
DB_FUNC(tcb, db_show_t4tcb, db_t4_table, CS_OWN, NULL)
{
device_t dev;
int radix, tid, t;
bool valid;
valid = false;
radix = db_radix;
db_radix = 10;
t = db_read_token();
if (t == tIDENT) {
dev = device_lookup_by_name(db_tok_string);
t = db_read_token();
if (t == tNUMBER) {
tid = db_tok_number;
valid = true;
}
}
db_radix = radix;
db_skip_to_eol();
if (!valid) {
db_printf("usage: show t4 tcb <nexus> <tid>\n");
return;
}
if (dev == NULL) {
db_printf("device not found\n");
return;
}
if (tid < 0) {
db_printf("invalid tid\n");
return;
}
t4_dump_tcb(device_get_softc(dev), tid);
}
#endif
Move ccr_aes_getdeckey() from ccr(4) to the cxgbe(4) driver. This routine will also be used by the TOE module to manage TLS keys. Sponsored by: Chelsio Communications
2018-02-26 22:12:31 +00:00
/*
* Borrowed from cesa_prep_aes_key().
*
* NB: The crypto engine wants the words in the decryption key in reverse
* order.
*/
void
t4_aes_getdeckey(void *dec_key, const void *enc_key, unsigned int kbits)
{
uint32_t ek[4 * (RIJNDAEL_MAXNR + 1)];
uint32_t *dkey;
int i;
rijndaelKeySetupEnc(ek, enc_key, kbits);
dkey = dec_key;
dkey += (kbits / 8) / 4;
switch (kbits) {
case 128:
for (i = 0; i < 4; i++)
*--dkey = htobe32(ek[4 * 10 + i]);
break;
case 192:
for (i = 0; i < 2; i++)
*--dkey = htobe32(ek[4 * 11 + 2 + i]);
for (i = 0; i < 4; i++)
*--dkey = htobe32(ek[4 * 12 + i]);
break;
case 256:
for (i = 0; i < 4; i++)
*--dkey = htobe32(ek[4 * 13 + i]);
for (i = 0; i < 4; i++)
*--dkey = htobe32(ek[4 * 14 + i]);
break;
}
MPASS(dkey == dec_key);
}
cxgbe(4): Keep track of the clusters that have to be freed by the custom free routine (rxb_free) in the driver. Fail MOD_UNLOAD with EBUSY if any such cluster has been handed up to the kernel but hasn't been freed yet. This prevents a panic later when the cluster finally needs to be freed but rxb_free is gone from the kernel. MFC after: 1 week
2014-07-23 22:29:22 +00:00
static struct sx mlu; /* mod load unload */
SX_SYSINIT(cxgbe_mlu, &mlu, "cxgbe mod load/unload");
cxgbe shouldn't directly know of the UMA zones where network buffers come from. MFC after: 1 week
2011-03-08 03:04:07 +00:00
static int
cxgbe(4): Ensure that the MOD_LOAD handler runs before either t4nex or t5nex attach to their devices. MFC after: 3 days
2013-04-11 17:50:50 +00:00
mod_event(module_t mod, int cmd, void *arg)
cxgbe shouldn't directly know of the UMA zones where network buffers come from. MFC after: 1 week
2011-03-08 03:04:07 +00:00
{
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
int rc = 0;
cxgbe(4): Ensure that the MOD_LOAD handler runs before either t4nex or t5nex attach to their devices. MFC after: 3 days
2013-04-11 17:50:50 +00:00
static int loaded = 0;
cxgbe shouldn't directly know of the UMA zones where network buffers come from. MFC after: 1 week
2011-03-08 03:04:07 +00:00
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
switch (cmd) {
case MOD_LOAD:
cxgbe(4): Keep track of the clusters that have to be freed by the custom free routine (rxb_free) in the driver. Fail MOD_UNLOAD with EBUSY if any such cluster has been handed up to the kernel but hasn't been freed yet. This prevents a panic later when the cluster finally needs to be freed but rxb_free is gone from the kernel. MFC after: 1 week
2014-07-23 22:29:22 +00:00
sx_xlock(&mlu);
if (loaded++ == 0) {
t4_sge_modload();
cxgbe(4): Use opaque cookies or tid range-checks to determine the intended recipient of a CPL when it can't be determined solely from the opcode. Retire the per-queue handlers for such CPLs in favor of the new scheme. Sponsored by: Chelsio Communications
2018-04-30 15:18:38 +00:00
t4_register_shared_cpl_handler(CPL_SET_TCB_RPL,
t4_filter_rpl, CPL_COOKIE_FILTER);
t4_register_shared_cpl_handler(CPL_L2T_WRITE_RPL,
do_l2t_write_rpl, CPL_COOKIE_FILTER);
cxgbe(4): Add support for hash filters. These filters reside in the card's memory instead of its TCAM and can be configured via a new "hashfilter" subcommand in cxgbetool. Hash and normal TCAM filters can be used together. The hardware does an exact-match of packet fields for hash filters, unlike the masked match performed for TCAM filters. Any T5/T6 card with memory can support at least half a million hash filters. The sample config file with the driver configures 512K of these, it is possible to double this to 1 million+ in some cases. The chip does an exact-match of fields of incoming datagrams with hash filters and performs the action configured for the filter if it matches. The fields to match are specified in a "filter mask" in the firmware config file. The filter mask always includes the 5-tuple (sip, dip, sport, dport, ipproto). It can, optionally, also include any subset of the filter mode (see filterMode and filterMask in the firmware config file). For example: filterMode = fragmentation, mpshittype, protocol, vlan, port, fcoe filterMask = protocol, port, vlan Exact values of the 5-tuple, the physical port, and VLAN tag would have to be provided while setting up a hash filter with the chip configuration above. Hash filters support all actions supported by TCAM filters. A packet that hits a hash filter can be dropped, let through (with optional steering to a specific queue or RSS region), switched out of another port (with optional L2 rewrite of DMAC, SMAC, VLAN tag), or get NAT'ed. (Support for some of these will show up in the driver in a follow-up commit very shortly). Sponsored by: Chelsio Communications
2018-05-09 04:09:49 +00:00
t4_register_shared_cpl_handler(CPL_ACT_OPEN_RPL,
t4_hashfilter_ao_rpl, CPL_COOKIE_HASHFILTER);
t4_register_shared_cpl_handler(CPL_SET_TCB_RPL,
t4_hashfilter_tcb_rpl, CPL_COOKIE_HASHFILTER);
t4_register_shared_cpl_handler(CPL_ABORT_RPL_RSS,
t4_del_hashfilter_rpl, CPL_COOKIE_HASHFILTER);
cxgbe(4): Changes to the CPL-handler registration mechanism and code related to "shared" CPLs. a) Combine t4_set_tcb_field and t4_set_tcb_field_rpl into a single function. Allow callers to direct the response to any iq. Tidy up set_ulp_mode_iscsi while there to use names from t4_tcb.h instead of magic constants. b) Remove all CPL handler tables from struct adapter. This reduces its size by around 2KB. All handlers are now registered at MOD_LOAD instead of attach or some kind of initialization/activation. The registration functions do not need an adapter parameter any more. c) Add per-iq handlers to deal with CPLs whose destination cannot be determined solely from the opcode. There are 2 such CPLs in use right now: SET_TCB_RPL and L2T_WRITE_RPL. The base driver continues to send filter and L2T_WRITEs over the mgmtq and solicits the reply on fwq. t4_tom (including the DDP code) now uses the port's ctrlq to send L2T_WRITEs and SET_TCB_FIELDs and solicits the reply on an ofld_rxq. fwq and ofld_rxq have different handlers that know what kind of tid to expect in the reply. Update t4_write_l2e and callers to to support any wrq/iq combination. Approved by: re@ (kib@) Sponsored by: Chelsio Communications
2016-07-05 01:29:24 +00:00
t4_register_cpl_handler(CPL_TRACE_PKT, t4_trace_pkt);
t4_register_cpl_handler(CPL_T5_TRACE_PKT, t5_trace_pkt);
cxgbe(4): Add code to deal with the chip's source MAC table (aka SMT). Submitted by: Krishnamraju Eraparaju @ Chelsio Sponsored by: Chelsio Communications
2018-05-31 21:31:08 +00:00
t4_register_cpl_handler(CPL_SMT_WRITE_RPL,
do_smt_write_rpl);
cxgbe(4): Keep track of the clusters that have to be freed by the custom free routine (rxb_free) in the driver. Fail MOD_UNLOAD with EBUSY if any such cluster has been handed up to the kernel but hasn't been freed yet. This prevents a panic later when the cluster finally needs to be freed but rxb_free is gone from the kernel. MFC after: 1 week
2014-07-23 22:29:22 +00:00
sx_init(&t4_list_lock, "T4/T5 adapters");
SLIST_INIT(&t4_list);
cxgbe(4): Delay the panic due to a fatal error by 30s. This lets information logged by the interrupt handler reach the system log before the system goes down.
2019-02-09 01:49:53 +00:00
callout_init(&fatal_callout, 1);
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
#ifdef TCP_OFFLOAD
cxgbe(4): Keep track of the clusters that have to be freed by the custom free routine (rxb_free) in the driver. Fail MOD_UNLOAD with EBUSY if any such cluster has been handed up to the kernel but hasn't been freed yet. This prevents a panic later when the cluster finally needs to be freed but rxb_free is gone from the kernel. MFC after: 1 week
2014-07-23 22:29:22 +00:00
sx_init(&t4_uld_list_lock, "T4/T5 ULDs");
SLIST_INIT(&t4_uld_list);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
#endif
Make most of the CLIP code conditional on #ifdef INET6. This fixes builds of kernels without INET6 such as LINT-NOINET6. Reported by: arybchik Reviewed by: np Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18384
2018-11-29 23:14:54 +00:00
#ifdef INET6
Move CLIP table handling out of TOM and into the base driver. - Store the clip table in 'struct adapter' instead of in the TOM softc. - Init the clip table during attach and teardown during detach. - While here, add a dev.<nexus>.<unit>.misc.clip sysctl to dump the CLIP table. This does mean that we update the clip table even if TOE is not enabled, but non-TOE things need the CLIP table anyway. Reviewed by: np, Krishnamraju Eraparaju @ Chelsio Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18010
2018-11-29 01:15:53 +00:00
t4_clip_modload();
Make most of the CLIP code conditional on #ifdef INET6. This fixes builds of kernels without INET6 such as LINT-NOINET6. Reported by: arybchik Reviewed by: np Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18384
2018-11-29 23:14:54 +00:00
#endif
cxgbe(4): Keep track of the clusters that have to be freed by the custom free routine (rxb_free) in the driver. Fail MOD_UNLOAD with EBUSY if any such cluster has been handed up to the kernel but hasn't been freed yet. This prevents a panic later when the cluster finally needs to be freed but rxb_free is gone from the kernel. MFC after: 1 week
2014-07-23 22:29:22 +00:00
t4_tracer_modload();
tweak_tunables();
}
sx_xunlock(&mlu);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
break;
cxgbe shouldn't directly know of the UMA zones where network buffers come from. MFC after: 1 week
2011-03-08 03:04:07 +00:00
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
case MOD_UNLOAD:
cxgbe(4): Keep track of the clusters that have to be freed by the custom free routine (rxb_free) in the driver. Fail MOD_UNLOAD with EBUSY if any such cluster has been handed up to the kernel but hasn't been freed yet. This prevents a panic later when the cluster finally needs to be freed but rxb_free is gone from the kernel. MFC after: 1 week
2014-07-23 22:29:22 +00:00
sx_xlock(&mlu);
if (--loaded == 0) {
int tries;
sx_slock(&t4_list_lock);
if (!SLIST_EMPTY(&t4_list)) {
rc = EBUSY;
sx_sunlock(&t4_list_lock);
goto done_unload;
}
#ifdef TCP_OFFLOAD
sx_slock(&t4_uld_list_lock);
if (!SLIST_EMPTY(&t4_uld_list)) {
rc = EBUSY;
sx_sunlock(&t4_uld_list_lock);
sx_sunlock(&t4_list_lock);
goto done_unload;
}
#endif
tries = 0;
while (tries++ < 5 && t4_sge_extfree_refs() != 0) {
uprintf("%ju clusters with custom free routine "
"still is use.\n", t4_sge_extfree_refs());
pause("t4unload", 2 * hz);
}
- Updated TOE support in the kernel. - Stateful TCP offload drivers for Terminator 3 and 4 (T3 and T4) ASICs. These are available as t3_tom and t4_tom modules that augment cxgb(4) and cxgbe(4) respectively. The cxgb/cxgbe drivers continue to work as usual with or without these extra features. - iWARP driver for Terminator 3 ASIC (kernel verbs). T4 iWARP in the works and will follow soon. Build-tested with make universe. 30s overview ============ What interfaces support TCP offload? Look for TOE4 and/or TOE6 in the capabilities of an interface: # ifconfig -m | grep TOE Enable/disable TCP offload on an interface (just like any other ifnet capability): # ifconfig cxgbe0 toe # ifconfig cxgbe0 -toe Which connections are offloaded? Look for toe4 and/or toe6 in the output of netstat and sockstat: # netstat -np tcp | grep toe # sockstat -46c | grep toe Reviewed by: bz, gnn Sponsored by: Chelsio communications. MFC after: ~3 months (after 9.1, and after ensuring MFC is feasible)
2012-06-19 07:34:13 +00:00
#ifdef TCP_OFFLOAD
Change t4_list_lock and t4_uld_list_lock from mutexes to sx'es. - tom_uninit had to be reworked not to hold the adapter lock (a mutex) around t4_deactivate_uld, which acquires the uld_list_lock. - the ifc_match for the interface cloner that creates the tracer ifnet had to be reworked as the kernel calls ifc_match with the global if_cloners_mtx held.
2013-08-28 20:59:22 +00:00
sx_sunlock(&t4_uld_list_lock);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
#endif
Change t4_list_lock and t4_uld_list_lock from mutexes to sx'es. - tom_uninit had to be reworked not to hold the adapter lock (a mutex) around t4_deactivate_uld, which acquires the uld_list_lock. - the ifc_match for the interface cloner that creates the tracer ifnet had to be reworked as the kernel calls ifc_match with the global if_cloners_mtx held.
2013-08-28 20:59:22 +00:00
sx_sunlock(&t4_list_lock);
cxgbe(4): Keep track of the clusters that have to be freed by the custom free routine (rxb_free) in the driver. Fail MOD_UNLOAD with EBUSY if any such cluster has been handed up to the kernel but hasn't been freed yet. This prevents a panic later when the cluster finally needs to be freed but rxb_free is gone from the kernel. MFC after: 1 week
2014-07-23 22:29:22 +00:00
if (t4_sge_extfree_refs() == 0) {
t4_tracer_modunload();
Make most of the CLIP code conditional on #ifdef INET6. This fixes builds of kernels without INET6 such as LINT-NOINET6. Reported by: arybchik Reviewed by: np Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18384
2018-11-29 23:14:54 +00:00
#ifdef INET6
Move CLIP table handling out of TOM and into the base driver. - Store the clip table in 'struct adapter' instead of in the TOM softc. - Init the clip table during attach and teardown during detach. - While here, add a dev.<nexus>.<unit>.misc.clip sysctl to dump the CLIP table. This does mean that we update the clip table even if TOE is not enabled, but non-TOE things need the CLIP table anyway. Reviewed by: np, Krishnamraju Eraparaju @ Chelsio Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18010
2018-11-29 01:15:53 +00:00
t4_clip_modunload();
Make most of the CLIP code conditional on #ifdef INET6. This fixes builds of kernels without INET6 such as LINT-NOINET6. Reported by: arybchik Reviewed by: np Sponsored by: Chelsio Communications Differential Revision: https://reviews.freebsd.org/D18384
2018-11-29 23:14:54 +00:00
#endif
cxgbe(4): Keep track of the clusters that have to be freed by the custom free routine (rxb_free) in the driver. Fail MOD_UNLOAD with EBUSY if any such cluster has been handed up to the kernel but hasn't been freed yet. This prevents a panic later when the cluster finally needs to be freed but rxb_free is gone from the kernel. MFC after: 1 week
2014-07-23 22:29:22 +00:00
#ifdef TCP_OFFLOAD
sx_destroy(&t4_uld_list_lock);
#endif
sx_destroy(&t4_list_lock);
t4_sge_modunload();
loaded = 0;
} else {
rc = EBUSY;
loaded++; /* undo earlier decrement */
}
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
}
cxgbe(4): Keep track of the clusters that have to be freed by the custom free routine (rxb_free) in the driver. Fail MOD_UNLOAD with EBUSY if any such cluster has been handed up to the kernel but hasn't been freed yet. This prevents a panic later when the cluster finally needs to be freed but rxb_free is gone from the kernel. MFC after: 1 week
2014-07-23 22:29:22 +00:00
done_unload:
sx_xunlock(&mlu);
Many updates to cxgbe(4) - Device configuration via plain text config file. Also able to operate when not attached to the chip as the master driver. - Generic "work request" queue that serves as the base for both ctrl and ofld tx queues. - Generic interrupt handler routine that can process any event on any kind of ingress queue (via a dispatch table). - A couple of new driver ioctls. cxgbetool can now install a firmware to the card ("loadfw" command) and can read the card's memory ("memdump" and "tcb" commands). - Lots of assorted information within dev.t4nex.X.misc.* This is primarily for debugging and won't show up in sysctl -a. - Code to manage the L2 tables on the chip. - Updates to cxgbe(4) man page to go with the tunables that have changed. - Updates to the shared code in common/ - Updates to the driver-firmware interface (now at fw 1.4.16.0) MFC after: 1 month
2011-12-16 02:09:51 +00:00
break;
}
return (rc);
cxgbe shouldn't directly know of the UMA zones where network buffers come from. MFC after: 1 week
2011-03-08 03:04:07 +00:00
}
cxgbe(4): Attach to cards with the Terminator 6 ASIC. T6 cards will come up as 't6nex' nexus devices with 'cc' ports hanging off them. The T6 firmware and configuration files will be added as soon as they are released. For now the driver will try to work with whatever firmware and configuration is on the card's flash. Sponsored by: Chelsio Communications
2016-09-16 00:08:37 +00:00
static devclass_t t4_devclass, t5_devclass, t6_devclass;
static devclass_t cxgbe_devclass, cxl_devclass, cc_devclass;
static devclass_t vcxgbe_devclass, vcxl_devclass, vcc_devclass;
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
cxgbe(4): Ensure that the MOD_LOAD handler runs before either t4nex or t5nex attach to their devices. MFC after: 3 days
2013-04-11 17:50:50 +00:00
DRIVER_MODULE(t4nex, pci, t4_driver, t4_devclass, mod_event, 0);
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
MODULE_VERSION(t4nex, 1);
Add dependencies on the firmware, which allows the loading of the cxgb and cxgbe modules. Reviewed and approved by: np MFC after: 1 week
2013-05-16 13:07:02 +00:00
MODULE_DEPEND(t4nex, firmware, 1, 1, 1);
Sync netmap sources with the version in our private tree. This commit contains large contributions from Giuseppe Lettieri and Stefano Garzarella, is partly supported by grants from Verisign and Cisco, and brings in the following: - fix zerocopy monitor ports and introduce copying monitor ports (the latter are lower performance but give access to all traffic in parallel with the application) - exclusive open mode, useful to implement solutions that recover from crashes of the main netmap client (suggested by Patrick Kelsey) - revised memory allocator in preparation for the 'passthrough mode' (ptnetmap) recently presented at bsdcan. ptnetmap is described in S. Garzarella, G. Lettieri, L. Rizzo; Virtual device passthrough for high speed VM networking, ACM/IEEE ANCS 2015, Oakland (CA) May 2015 http://info.iet.unipi.it/~luigi/research.html - fix rx CRC handing on ixl - add module dependencies for netmap when building drivers as modules - minor simplifications to device-specific routines (*txsync, *rxsync) - general code cleanup (remove unused variables, introduce macros to access rings and remove duplicate code, Applications do not need to be recompiled, unless of course they want to use the new features (monitors and exclusive open). Those willing to try this code on stable/10 can just update the sys/dev/netmap/*, sys/net/netmap* with the version in HEAD and apply the small patches to individual device drivers. MFC after: 1 month Sponsored by: (partly) Verisign, Cisco
2015-07-10 05:51:36 +00:00
#ifdef DEV_NETMAP
MODULE_DEPEND(t4nex, netmap, 1, 1, 1);
#endif /* DEV_NETMAP */
cxgbe(4): Ensure that the MOD_LOAD handler runs before either t4nex or t5nex attach to their devices. MFC after: 3 days
2013-04-11 17:50:50 +00:00
DRIVER_MODULE(t5nex, pci, t5_driver, t5_devclass, mod_event, 0);
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
MODULE_VERSION(t5nex, 1);
Add dependencies on the firmware, which allows the loading of the cxgb and cxgbe modules. Reviewed and approved by: np MFC after: 1 week
2013-05-16 13:07:02 +00:00
MODULE_DEPEND(t5nex, firmware, 1, 1, 1);
Sync netmap sources with the version in our private tree. This commit contains large contributions from Giuseppe Lettieri and Stefano Garzarella, is partly supported by grants from Verisign and Cisco, and brings in the following: - fix zerocopy monitor ports and introduce copying monitor ports (the latter are lower performance but give access to all traffic in parallel with the application) - exclusive open mode, useful to implement solutions that recover from crashes of the main netmap client (suggested by Patrick Kelsey) - revised memory allocator in preparation for the 'passthrough mode' (ptnetmap) recently presented at bsdcan. ptnetmap is described in S. Garzarella, G. Lettieri, L. Rizzo; Virtual device passthrough for high speed VM networking, ACM/IEEE ANCS 2015, Oakland (CA) May 2015 http://info.iet.unipi.it/~luigi/research.html - fix rx CRC handing on ixl - add module dependencies for netmap when building drivers as modules - minor simplifications to device-specific routines (*txsync, *rxsync) - general code cleanup (remove unused variables, introduce macros to access rings and remove duplicate code, Applications do not need to be recompiled, unless of course they want to use the new features (monitors and exclusive open). Those willing to try this code on stable/10 can just update the sys/dev/netmap/*, sys/net/netmap* with the version in HEAD and apply the small patches to individual device drivers. MFC after: 1 month Sponsored by: (partly) Verisign, Cisco
2015-07-10 05:51:36 +00:00
#ifdef DEV_NETMAP
MODULE_DEPEND(t5nex, netmap, 1, 1, 1);
#endif /* DEV_NETMAP */
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
cxgbe(4): Attach to cards with the Terminator 6 ASIC. T6 cards will come up as 't6nex' nexus devices with 'cc' ports hanging off them. The T6 firmware and configuration files will be added as soon as they are released. For now the driver will try to work with whatever firmware and configuration is on the card's flash. Sponsored by: Chelsio Communications
2016-09-16 00:08:37 +00:00
DRIVER_MODULE(t6nex, pci, t6_driver, t6_devclass, mod_event, 0);
MODULE_VERSION(t6nex, 1);
MODULE_DEPEND(t6nex, firmware, 1, 1, 1);
#ifdef DEV_NETMAP
MODULE_DEPEND(t6nex, netmap, 1, 1, 1);
#endif /* DEV_NETMAP */
cxgbe(4) - NIC driver for Chelsio T4 (Terminator 4) based 10Gb/1Gb adapters. MFC after: 3 weeks
2011-02-18 08:00:26 +00:00
DRIVER_MODULE(cxgbe, t4nex, cxgbe_driver, cxgbe_devclass, 0, 0);
MODULE_VERSION(cxgbe, 1);
cxgbe(4): Add support for Chelsio's Terminator 5 (aka T5) ASIC. This includes support for the NIC and TOE features of the 40G, 10G, and 1G/100M cards based on the T5. The ASIC is mostly backward compatible with the Terminator 4 so cxgbe(4) has been updated instead of writing a brand new driver. T5 cards will show up as cxl (short for cxlgb) ports attached to the t5nex bus driver. Sponsored by: Chelsio
2013-03-30 02:26:20 +00:00
DRIVER_MODULE(cxl, t5nex, cxl_driver, cxl_devclass, 0, 0);
MODULE_VERSION(cxl, 1);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
cxgbe(4): Attach to cards with the Terminator 6 ASIC. T6 cards will come up as 't6nex' nexus devices with 'cc' ports hanging off them. The T6 firmware and configuration files will be added as soon as they are released. For now the driver will try to work with whatever firmware and configuration is on the card's flash. Sponsored by: Chelsio Communications
2016-09-16 00:08:37 +00:00
DRIVER_MODULE(cc, t6nex, cc_driver, cc_devclass, 0, 0);
MODULE_VERSION(cc, 1);
Add support for configuring additional virtual interfaces (VIs) on a port. Each virtual interface has its own MAC address, queues, and statistics. The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented as additional VIs on each port. This change allows additional non-netmap interfaces to be configured on each port. Additional virtual interfaces use the naming scheme vcxgbeX or vcxlX. Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a value greater than 1 before loading the cxgbe(4) or cxl(4) driver. NB: The first VI on each port is the "main" interface (cxgbeX or cxlX). T4/T5 NICs provide a limited number of MAC addresses for each physical port. As a result, a maximum of six VIs can be configured on each port (including the "main" interface and the netmap interface when netmap is enabled). One user-visible result is that when netmap is enabled, packets received or transmitted via the netmap interface are no longer counted in the stats for the "main" interface, but are not accounted to the netmap interface. The netmap interfaces now also have a new-bus device and export various information sysctl nodes via dev.n(cxgbe|cxl).X. The cxgbetool 'clearstats' command clears the stats for all VIs on the specified port along with the port's stats. There is currently no way to clear the stats of an individual VI. Reviewed by: np MFC after: 1 month Sponsored by: Chelsio
2015-12-03 00:02:01 +00:00
DRIVER_MODULE(vcxgbe, cxgbe, vcxgbe_driver, vcxgbe_devclass, 0, 0);
MODULE_VERSION(vcxgbe, 1);
DRIVER_MODULE(vcxl, cxl, vcxl_driver, vcxl_devclass, 0, 0);
MODULE_VERSION(vcxl, 1);
cxgbe(4): Attach to cards with the Terminator 6 ASIC. T6 cards will come up as 't6nex' nexus devices with 'cc' ports hanging off them. The T6 firmware and configuration files will be added as soon as they are released. For now the driver will try to work with whatever firmware and configuration is on the card's flash. Sponsored by: Chelsio Communications
2016-09-16 00:08:37 +00:00
DRIVER_MODULE(vcc, cc, vcc_driver, vcc_devclass, 0, 0);
MODULE_VERSION(vcc, 1);
Reference in New Issue Copy Permalink