d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
3012653750
freebsd-skq/sys/dev/ixgbe/if_ix.c

4797 lines
134 KiB
C
Raw Normal View History

This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
/******************************************************************************
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
Copyright (c) 2001-2015, Intel Corporation
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
All rights reserved.
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.
3. Neither the name of the Intel Corporation nor the names of its
contributors may be used to endorse or promote products derived from
this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS 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 COPYRIGHT OWNER 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.
******************************************************************************/
/*$FreeBSD$*/
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
#ifndef IXGBE_STANDALONE_BUILD
- Add the RX refresh changes from igb to ixgbe - Also a couple minor tweaks to the TX code from the same source. - Add the INET ioctl code which has been missing from this driver, and which caused IP aliases to reset the interface. - Last, some minor logic changes that just reflect upcoming hardware support, but have no other functional effect now. MFC after a week
2011-04-25 23:34:21 +00:00
#include "opt_inet.h"
First off: update the driver README, the old one was horribly crusty, and this still isn't perfect, but its at least a bit more recent. Secondly, a few improvements to the driver from Andrew Boyer, support hint to allow devices to not attach, add VLAN_HWTSO capability so vlans can use TSO, fix in the interrupt handler to make sure the stack TX queue is processed. Oh, and also make sure IPv6 does not cause a re-init in the ioctl routine. Thanks for your efforts Andrew! Thanks to Claudio Jeker for noticing the ixgbe_xmit() routine was not correctly swapping the dma map from the first to the last descriptor in a multi-descriptor transmission, corrected this.
2011-06-02 00:34:57 +00:00
#include "opt_inet6.h"
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
#include "opt_rss.h"
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
#endif
A couple late breaking bugs that testing have turned up. - change include style so build in kernel try OR standalone work. - Limit HWCSUM - I was led to believe that it would work with RSS, but our testing had odd issues which suggests this is false. - A fatfinger error in the ioctl code made ifconfig up not work. Approved by: re
2007-07-12 19:04:11 +00:00
#include "ixgbe.h"
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
#ifdef RSS
Fix ixgbe(4) to compile - with RSS; with ix+ixv in the kernel. * Fix the multiple same-named devclasses; the duplicate name trips up the linker. * Re-do the taskqueue stuff to use the new cpuset API, not the old pinned API. * Add includes for the new location of the RSS configuration routines. This allows ixgbe to compile as a module /and/ linked into the kernel, along with RSS working. Sponsored by: Norse Corp, Inc.
2015-03-18 05:05:30 +00:00
#include <net/rss_config.h>
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
#include <netinet/in_rss.h>
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
#endif
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/*********************************************************************
* Set this to one to display debug statistics
*********************************************************************/
int ixgbe_display_debug_stats = 0;
/*********************************************************************
* Driver version
*********************************************************************/
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
char ixgbe_driver_version[] = "2.8.3";
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/*********************************************************************
* PCI Device ID Table
*
* Used by probe to select devices to load on
* Last field stores an index into ixgbe_strings
* Last entry must be all 0s
*
* { Vendor ID, Device ID, SubVendor ID, SubDevice ID, String Index }
*********************************************************************/
static ixgbe_vendor_info_t ixgbe_vendor_info_array[] =
{
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_DUAL_PORT, 0, 0, 0},
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_SINGLE_PORT, 0, 0, 0},
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_CX4, 0, 0, 0},
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT, 0, 0, 0},
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT2, 0, 0, 0},
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598, 0, 0, 0},
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_DA_DUAL_PORT, 0, 0, 0},
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_CX4_DUAL_PORT, 0, 0, 0},
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_XF_LR, 0, 0, 0},
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM, 0, 0, 0},
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_SFP_LOM, 0, 0, 0},
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4, 0, 0, 0},
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4_MEZZ, 0, 0, 0},
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP, 0, 0, 0},
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_XAUI_LOM, 0, 0, 0},
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_CX4, 0, 0, 0},
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_T3_LOM, 0, 0, 0},
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_COMBO_BACKPLANE, 0, 0, 0},
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BACKPLANE_FCOE, 0, 0, 0},
Update to the ixgbe driver: - Add a couple of new devices - Flow control changes in shared and core code - Bug fix to Flow Director for 82598 - Shared code sync to internal with required core change Thanks to those helping in the testing and improvements to this driver! MFC after:5 days
2012-07-05 20:51:44 +00:00
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF2, 0, 0, 0},
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_FCOE, 0, 0, 0},
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599EN_SFP, 0, 0, 0},
First, sync to internal shared code, and then Fixes: - flow control - don't override user value on re-init - fix to make 1G optics work correctly - change to interrupt enabling - some bits were incorrect for certain hardware. - certain stats fixes, remove a duplicate increment of ierror, thanks to Scott Long for pointing these out. - shared code link interface changed, requiring some core code changes to accomodate this. - add an m_adj() to ETHER_ALIGN on the recieve side, this was requested by Mike Karels, thanks Mike. - Multicast code corrections also thanks to Mike Karels.
2013-03-04 23:07:40 +00:00
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF_QP, 0, 0, 0},
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_QSFP_SF_QP, 0, 0, 0},
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T, 0, 0, 0},
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T1, 0, 0, 0},
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T, 0, 0, 0},
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KR, 0, 0, 0},
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KX4, 0, 0, 0},
{IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_10G_T, 0, 0, 0},
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/* required last entry */
{0, 0, 0, 0, 0}
};
/*********************************************************************
* Table of branding strings
*********************************************************************/
static char *ixgbe_strings[] = {
"Intel(R) PRO/10GbE PCI-Express Network Driver"
};
/*********************************************************************
* Function prototypes
*********************************************************************/
static int ixgbe_probe(device_t);
static int ixgbe_attach(device_t);
static int ixgbe_detach(device_t);
static int ixgbe_shutdown(device_t);
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
static int ixgbe_suspend(device_t);
static int ixgbe_resume(device_t);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
static int ixgbe_ioctl(struct ifnet *, u_long, caddr_t);
Thanks to Michael Tuexen for adding SCTP support for 82599, also for finding a one character bug that kept TSO from working. Sometimes with direct attach cables a failure can occur in init, the old method of calling detach was broken, there is no way to return an error to the system from init, so I have changed it to return failure thru the ioctl. And, have fixed the ALTQ code changes of Max Laier, sorry Max :)
2010-03-30 19:09:18 +00:00
static void ixgbe_init(void *);
Fixes for panic experienced in test at Intel, when doing bidirectional stress traffic on 82598. Also a couple bug fixes from Michael Tuexen, thank you!! Add a workaround into the header so that 8 REL can use the driver (adds local copy of ALTQ fix). MFC: in a few days
2010-06-03 00:00:45 +00:00
static void ixgbe_init_locked(struct adapter *);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
static void ixgbe_stop(void *);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
#if __FreeBSD_version >= 1100036
Convert to if_get_counter().
2014-09-28 07:29:45 +00:00
static uint64_t ixgbe_get_counter(struct ifnet *, ift_counter);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
#endif
static void ixgbe_add_media_types(struct adapter *);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
static void ixgbe_media_status(struct ifnet *, struct ifmediareq *);
static int ixgbe_media_change(struct ifnet *);
static void ixgbe_identify_hardware(struct adapter *);
static int ixgbe_allocate_pci_resources(struct adapter *);
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
static void ixgbe_get_slot_info(struct ixgbe_hw *);
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
static int ixgbe_allocate_msix(struct adapter *);
static int ixgbe_allocate_legacy(struct adapter *);
static int ixgbe_setup_msix(struct adapter *);
static void ixgbe_free_pci_resources(struct adapter *);
A couple fixes got clobbered, putting them back.
2010-11-26 23:57:13 +00:00
static void ixgbe_local_timer(void *);
static int ixgbe_setup_interface(device_t, struct adapter *);
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
static void ixgbe_config_dmac(struct adapter *);
static void ixgbe_config_delay_values(struct adapter *);
A couple fixes got clobbered, putting them back.
2010-11-26 23:57:13 +00:00
static void ixgbe_config_link(struct adapter *);
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
static void ixgbe_check_eee_support(struct adapter *);
static void ixgbe_check_wol_support(struct adapter *);
static int ixgbe_setup_low_power_mode(struct adapter *);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
static void ixgbe_rearm_queues(struct adapter *, u64);
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
static void ixgbe_initialize_transmit_units(struct adapter *);
static void ixgbe_initialize_receive_units(struct adapter *);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
static void ixgbe_enable_rx_drop(struct adapter *);
static void ixgbe_disable_rx_drop(struct adapter *);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
static void ixgbe_enable_intr(struct adapter *);
static void ixgbe_disable_intr(struct adapter *);
static void ixgbe_update_stats_counters(struct adapter *);
static void ixgbe_set_promisc(struct adapter *);
static void ixgbe_set_multi(struct adapter *);
static void ixgbe_update_link_status(struct adapter *);
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
static void ixgbe_set_ivar(struct adapter *, u8, u8, s8);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
static void ixgbe_configure_ivars(struct adapter *);
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
static u8 * ixgbe_mc_array_itr(struct ixgbe_hw *, u8 **, u32 *);
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
static void ixgbe_setup_vlan_hw_support(struct adapter *);
This updates the ixgbe driver to Intel internal version 1.4.7 Shared code changes, core driver fixes, vlan event/filter support Also Kip Macy's fix to allow any number of queues, thanks Kip!
2008-07-30 18:15:18 +00:00
static void ixgbe_register_vlan(void *, struct ifnet *, u16);
static void ixgbe_unregister_vlan(void *, struct ifnet *, u16);
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
static void ixgbe_add_device_sysctls(struct adapter *);
static void ixgbe_add_hw_stats(struct adapter *);
/* Sysctl handlers */
static int ixgbe_set_flowcntl(SYSCTL_HANDLER_ARGS);
static int ixgbe_set_advertise(SYSCTL_HANDLER_ARGS);
static int ixgbe_sysctl_thermal_test(SYSCTL_HANDLER_ARGS);
static int ixgbe_sysctl_dmac(SYSCTL_HANDLER_ARGS);
static int ixgbe_sysctl_phy_temp(SYSCTL_HANDLER_ARGS);
static int ixgbe_sysctl_phy_overtemp_occurred(SYSCTL_HANDLER_ARGS);
static int ixgbe_sysctl_wol_enable(SYSCTL_HANDLER_ARGS);
static int ixgbe_sysctl_wufc(SYSCTL_HANDLER_ARGS);
static int ixgbe_sysctl_eee_enable(SYSCTL_HANDLER_ARGS);
static int ixgbe_sysctl_eee_negotiated(SYSCTL_HANDLER_ARGS);
static int ixgbe_sysctl_eee_rx_lpi_status(SYSCTL_HANDLER_ARGS);
static int ixgbe_sysctl_eee_tx_lpi_status(SYSCTL_HANDLER_ARGS);
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
/* Support for pluggable optic modules */
static bool ixgbe_sfp_probe(struct adapter *);
Update driver to version 2.3.8: CRITICAL FIX - with stats changes the older 82598 will panic and trash the stack on driver load, FCOE registers ONLY exist in 82599 and must not be read otherwise. kern/153951 - to correct incorrect media type on adapters with pluggable modules I have eliminated the old static table in favor of a new dynamic shared code routine. This also has the benefit of detecting changes when a different module is inserted. Performance/enhancement to the Flow Director code from my linux coworker (the developer of the code). Fixes from Michael Tuexen - a data corruption problem on the 82599 (CRITICAL), fix so the buf size correctly adjusts as the cluster changes, and max descriptors are set properly. Also added 16K clusters for those REALLY big jumbos :) In the RX path, the RX LOCK was not being released, and this causes LOR problems. Add the code that igb already has. Sync with in house shared code, this was necessary for the Flow Director fix. MFC in 2 days
2011-01-19 19:36:27 +00:00
static void ixgbe_setup_optics(struct adapter *);
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
/* Legacy (single vector interrupt handler */
static void ixgbe_legacy_irq(void *);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/* The MSI/X Interrupt handlers */
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
static void ixgbe_msix_que(void *);
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
static void ixgbe_msix_link(void *);
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
/* Deferred interrupt tasklets */
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
static void ixgbe_handle_que(void *, int);
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
static void ixgbe_handle_link(void *, int);
static void ixgbe_handle_msf(void *, int);
static void ixgbe_handle_mod(void *, int);
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
static void ixgbe_handle_phy(void *, int);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
#ifdef IXGBE_FDIR
static void ixgbe_reinit_fdir(void *, int);
#endif
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/*********************************************************************
* FreeBSD Device Interface Entry Points
*********************************************************************/
Fix ixgbe(4) to compile - with RSS; with ix+ixv in the kernel. * Fix the multiple same-named devclasses; the duplicate name trips up the linker. * Re-do the taskqueue stuff to use the new cpuset API, not the old pinned API. * Add includes for the new location of the RSS configuration routines. This allows ixgbe to compile as a module /and/ linked into the kernel, along with RSS working. Sponsored by: Norse Corp, Inc.
2015-03-18 05:05:30 +00:00
static device_method_t ix_methods[] = {
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/* Device interface */
DEVMETHOD(device_probe, ixgbe_probe),
DEVMETHOD(device_attach, ixgbe_attach),
DEVMETHOD(device_detach, ixgbe_detach),
DEVMETHOD(device_shutdown, ixgbe_shutdown),
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
DEVMETHOD(device_suspend, ixgbe_suspend),
DEVMETHOD(device_resume, ixgbe_resume),
Use DEVMETHOD_END macro defined in sys/bus.h instead of {0, 0} sentinel on device_method_t arrays Reviewed by: cognet Approved by: cognet
2013-01-30 18:01:20 +00:00
DEVMETHOD_END
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
};
Fix ixgbe(4) to compile - with RSS; with ix+ixv in the kernel. * Fix the multiple same-named devclasses; the duplicate name trips up the linker. * Re-do the taskqueue stuff to use the new cpuset API, not the old pinned API. * Add includes for the new location of the RSS configuration routines. This allows ixgbe to compile as a module /and/ linked into the kernel, along with RSS working. Sponsored by: Norse Corp, Inc.
2015-03-18 05:05:30 +00:00
static driver_t ix_driver = {
"ix", ix_methods, sizeof(struct adapter),
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
};
Fix ixgbe(4) to compile - with RSS; with ix+ixv in the kernel. * Fix the multiple same-named devclasses; the duplicate name trips up the linker. * Re-do the taskqueue stuff to use the new cpuset API, not the old pinned API. * Add includes for the new location of the RSS configuration routines. This allows ixgbe to compile as a module /and/ linked into the kernel, along with RSS working. Sponsored by: Norse Corp, Inc.
2015-03-18 05:05:30 +00:00
devclass_t ix_devclass;
DRIVER_MODULE(ix, pci, ix_driver, ix_devclass, 0, 0);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Fix ixgbe(4) to compile - with RSS; with ix+ixv in the kernel. * Fix the multiple same-named devclasses; the duplicate name trips up the linker. * Re-do the taskqueue stuff to use the new cpuset API, not the old pinned API. * Add includes for the new location of the RSS configuration routines. This allows ixgbe to compile as a module /and/ linked into the kernel, along with RSS working. Sponsored by: Norse Corp, Inc.
2015-03-18 05:05:30 +00:00
MODULE_DEPEND(ix, pci, 1, 1, 1);
MODULE_DEPEND(ix, ether, 1, 1, 1);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
/*
** TUNEABLE PARAMETERS:
*/
Expose system level ixgbe sysctls. Device level sysctls are already exposed as dev.ix.<device> Fixing the case where number of queues for igb is auto-tuned and hw.igb.num_queues does not return current/updated value. Reviewed by: jfv Approved by: re (delphij) MFC after: 2 weeks
2013-10-05 19:17:56 +00:00
static SYSCTL_NODE(_hw, OID_AUTO, ix, CTLFLAG_RD, 0,
"IXGBE driver parameters");
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
/*
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
** AIM: Adaptive Interrupt Moderation
** which means that the interrupt rate
** is varied over time based on the
** traffic for that interrupt vector
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
*/
static int ixgbe_enable_aim = TRUE;
Pull in r267961 and r267973 again. Fix for issues reported will follow.
2014-06-28 03:56:17 +00:00
SYSCTL_INT(_hw_ix, OID_AUTO, enable_aim, CTLFLAG_RWTUN, &ixgbe_enable_aim, 0,
Expose system level ixgbe sysctls. Device level sysctls are already exposed as dev.ix.<device> Fixing the case where number of queues for igb is auto-tuned and hw.igb.num_queues does not return current/updated value. Reviewed by: jfv Approved by: re (delphij) MFC after: 2 weeks
2013-10-05 19:17:56 +00:00
"Enable adaptive interrupt moderation");
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
ixgbe changes: - remove experimental code for disabling CRC - use the correct constant for conversion between interrupt rate and EITR values (the previous values were off by a factor of 2) - make dev.ix.N.queueM.interrupt_rate a RW sysctl variable. Changing individual values affects the queue immediately, and propagates to all interfaces at the next reinit. - add dev.ix.N.queueM.irqs rdonly sysctl, to export the actual interrupt counts Netmap-related changes for ixgbe: - use the "new" format for TX descriptors in netmap mode. - pass interrupt mitigation delays to the user process doing poll() on a netmap file descriptor. On the RX side this means we will not check the ring more than once per interrupt. This gives the process a chance to sleep and process packets in larger batches, thus reducing CPU usage. On the TX side we take this even further: completed transmissions are reclaimed every half ring even if the NIC interrupts more often. This saves even more CPU without any additional tx delays. Generic Netmap-related changes: - align the netmap_kring to cache lines so that there is no false sharing (possibly useful for multiqueue NICs and MSIX interrupts, which are handled by different cores). It's a minor improvement but it does not cost anything. Reviewed by: Jack Vogel Approved by: Jack Vogel
2012-01-26 09:55:16 +00:00
static int ixgbe_max_interrupt_rate = (4000000 / IXGBE_LOW_LATENCY);
Expose system level ixgbe sysctls. Device level sysctls are already exposed as dev.ix.<device> Fixing the case where number of queues for igb is auto-tuned and hw.igb.num_queues does not return current/updated value. Reviewed by: jfv Approved by: re (delphij) MFC after: 2 weeks
2013-10-05 19:17:56 +00:00
SYSCTL_INT(_hw_ix, OID_AUTO, max_interrupt_rate, CTLFLAG_RDTUN,
&ixgbe_max_interrupt_rate, 0, "Maximum interrupts per second");
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/* How many packets rxeof tries to clean at a time */
Patch #7 This is primarily about processing limit control. - add a limit for both RX and TX, change the default to 256 - change the sysctl usage to be common, and now to be called during init for each ring. - the TX limit is not yet used, but the changes in the last patch in this series uses the value. - the motivation behind these changes is to improve data locality in the final code. - rxeof interface changes since it now gets limit from the ring struct
2012-11-30 23:28:01 +00:00
static int ixgbe_rx_process_limit = 256;
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
TUNABLE_INT("hw.ixgbe.rx_process_limit", &ixgbe_rx_process_limit);
Expose system level ixgbe sysctls. Device level sysctls are already exposed as dev.ix.<device> Fixing the case where number of queues for igb is auto-tuned and hw.igb.num_queues does not return current/updated value. Reviewed by: jfv Approved by: re (delphij) MFC after: 2 weeks
2013-10-05 19:17:56 +00:00
SYSCTL_INT(_hw_ix, OID_AUTO, rx_process_limit, CTLFLAG_RDTUN,
&ixgbe_rx_process_limit, 0,
"Maximum number of received packets to process at a time,"
"-1 means unlimited");
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Patch #7 This is primarily about processing limit control. - add a limit for both RX and TX, change the default to 256 - change the sysctl usage to be common, and now to be called during init for each ring. - the TX limit is not yet used, but the changes in the last patch in this series uses the value. - the motivation behind these changes is to improve data locality in the final code. - rxeof interface changes since it now gets limit from the ring struct
2012-11-30 23:28:01 +00:00
/* How many packets txeof tries to clean at a time */
static int ixgbe_tx_process_limit = 256;
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
TUNABLE_INT("hw.ixgbe.tx_process_limit", &ixgbe_tx_process_limit);
Expose system level ixgbe sysctls. Device level sysctls are already exposed as dev.ix.<device> Fixing the case where number of queues for igb is auto-tuned and hw.igb.num_queues does not return current/updated value. Reviewed by: jfv Approved by: re (delphij) MFC after: 2 weeks
2013-10-05 19:17:56 +00:00
SYSCTL_INT(_hw_ix, OID_AUTO, tx_process_limit, CTLFLAG_RDTUN,
&ixgbe_tx_process_limit, 0,
"Maximum number of sent packets to process at a time,"
"-1 means unlimited");
Patch #7 This is primarily about processing limit control. - add a limit for both RX and TX, change the default to 256 - change the sysctl usage to be common, and now to be called during init for each ring. - the TX limit is not yet used, but the changes in the last patch in this series uses the value. - the motivation behind these changes is to improve data locality in the final code. - rxeof interface changes since it now gets limit from the ring struct
2012-11-30 23:28:01 +00:00
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
/*
** Smart speed setting, default to on
** this only works as a compile option
** right now as its during attach, set
** this to 'ixgbe_smart_speed_off' to
** disable.
*/
static int ixgbe_smart_speed = ixgbe_smart_speed_on;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
/*
* MSIX should be the default for best performance,
* but this allows it to be forced off for testing.
*/
static int ixgbe_enable_msix = 1;
Expose system level ixgbe sysctls. Device level sysctls are already exposed as dev.ix.<device> Fixing the case where number of queues for igb is auto-tuned and hw.igb.num_queues does not return current/updated value. Reviewed by: jfv Approved by: re (delphij) MFC after: 2 weeks
2013-10-05 19:17:56 +00:00
SYSCTL_INT(_hw_ix, OID_AUTO, enable_msix, CTLFLAG_RDTUN, &ixgbe_enable_msix, 0,
"Enable MSI-X interrupts");
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
/*
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
* Number of Queues, can be set to 0,
* it then autoconfigures based on the
- Add the RX refresh changes from igb to ixgbe - Also a couple minor tweaks to the TX code from the same source. - Add the INET ioctl code which has been missing from this driver, and which caused IP aliases to reset the interface. - Last, some minor logic changes that just reflect upcoming hardware support, but have no other functional effect now. MFC after a week
2011-04-25 23:34:21 +00:00
* number of cpus with a max of 8. This
* can be overriden manually here.
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
*/
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
static int ixgbe_num_queues = 0;
Expose system level ixgbe sysctls. Device level sysctls are already exposed as dev.ix.<device> Fixing the case where number of queues for igb is auto-tuned and hw.igb.num_queues does not return current/updated value. Reviewed by: jfv Approved by: re (delphij) MFC after: 2 weeks
2013-10-05 19:17:56 +00:00
SYSCTL_INT(_hw_ix, OID_AUTO, num_queues, CTLFLAG_RDTUN, &ixgbe_num_queues, 0,
"Number of queues to configure, 0 indicates autoconfigure");
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
/*
** Number of TX descriptors per ring,
** setting higher than RX as this seems
** the better performing choice.
*/
static int ixgbe_txd = PERFORM_TXD;
Expose system level ixgbe sysctls. Device level sysctls are already exposed as dev.ix.<device> Fixing the case where number of queues for igb is auto-tuned and hw.igb.num_queues does not return current/updated value. Reviewed by: jfv Approved by: re (delphij) MFC after: 2 weeks
2013-10-05 19:17:56 +00:00
SYSCTL_INT(_hw_ix, OID_AUTO, txd, CTLFLAG_RDTUN, &ixgbe_txd, 0,
Fix a typo. PR: 191898 Submitted by: vsjcfm@gmail.com
2014-07-17 06:21:58 +00:00
"Number of transmit descriptors per queue");
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
/* Number of RX descriptors per ring */
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
static int ixgbe_rxd = PERFORM_RXD;
Expose system level ixgbe sysctls. Device level sysctls are already exposed as dev.ix.<device> Fixing the case where number of queues for igb is auto-tuned and hw.igb.num_queues does not return current/updated value. Reviewed by: jfv Approved by: re (delphij) MFC after: 2 weeks
2013-10-05 19:17:56 +00:00
SYSCTL_INT(_hw_ix, OID_AUTO, rxd, CTLFLAG_RDTUN, &ixgbe_rxd, 0,
"Number of receive descriptors per queue");
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
/*
** Defining this on will allow the use
** of unsupported SFP+ modules, note that
** doing so you are on your own :)
*/
static int allow_unsupported_sfp = FALSE;
Renamed hw.ixgbe.unsupported_sfp -> hw.ix.unsupported_sfp This now matches all other ixgbe sysctl / tunables. Sponsored by: Multiplay
2014-08-14 13:25:05 +00:00
TUNABLE_INT("hw.ix.unsupported_sfp", &allow_unsupported_sfp);
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
/* Keep running tab on them for sanity check */
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
static int ixgbe_total_ports;
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
#ifdef IXGBE_FDIR
/*
** Flow Director actually 'steals'
** part of the packet buffer as its
** filter pool, this variable controls
** how much it uses:
** 0 = 64K, 1 = 128K, 2 = 256K
*/
static int fdir_pballoc = 1;
#endif
1. Fix the handling of link reset while in netmap more. A link reset now is completely transparent for the netmap client: even if the NIC resets its own ring (e.g. restarting from 0), the client will not see any change in the current rx/tx positions, because the driver will keep track of the offset between the two. 2. make the device-specific code more uniform across different drivers There were some inconsistencies in the implementation of the netmap support routines, now drivers have been aligned to a common code structure. 3. import netmap support for ixgbe . This is implemented as a very small patch for ixgbe.c (233 lines, 11 chunks, mostly comments: in total the patch has only 54 lines of new code) , as most of the code is in an external file sys/dev/netmap/ixgbe_netmap.h , following some initial comments from Jack Vogel about making changes less intrusive. (Note, i have emailed Jack multiple times asking if he had comments on this structure of the code; i got no reply so i assume he is fine with it). Support for other drivers (em, lem, re, igb) will come later. "ixgbe" is now the reference driver for netmap support. Both the external file (sys/dev/netmap/ixgbe_netmap.h) and the device-specific patches (in sys/dev/ixgbe/ixgbe.c) are heavily commented and should serve as a reference for other device drivers. Tested on i386 and amd64 with the pkt-gen program in tools/tools/netmap, the sender does 14.88 Mpps at 1050 Mhz and 14.2 Mpps at 900 MHz on an i7-860 with 4 cores and 82599 card. Haven't tried yet more aggressive optimizations such as adding 'prefetch' instructions in the time-critical parts of the code.
2011-12-05 12:06:53 +00:00
#ifdef DEV_NETMAP
/*
* The #ifdef DEV_NETMAP / #endif blocks in this file are meant to
* be a reference on how to implement netmap support in a driver.
* Additional comments are in ixgbe_netmap.h .
*
fix a typo in a comment
2012-05-17 14:36:19 +00:00
* <dev/netmap/ixgbe_netmap.h> contains functions for netmap support
1. Fix the handling of link reset while in netmap more. A link reset now is completely transparent for the netmap client: even if the NIC resets its own ring (e.g. restarting from 0), the client will not see any change in the current rx/tx positions, because the driver will keep track of the offset between the two. 2. make the device-specific code more uniform across different drivers There were some inconsistencies in the implementation of the netmap support routines, now drivers have been aligned to a common code structure. 3. import netmap support for ixgbe . This is implemented as a very small patch for ixgbe.c (233 lines, 11 chunks, mostly comments: in total the patch has only 54 lines of new code) , as most of the code is in an external file sys/dev/netmap/ixgbe_netmap.h , following some initial comments from Jack Vogel about making changes less intrusive. (Note, i have emailed Jack multiple times asking if he had comments on this structure of the code; i got no reply so i assume he is fine with it). Support for other drivers (em, lem, re, igb) will come later. "ixgbe" is now the reference driver for netmap support. Both the external file (sys/dev/netmap/ixgbe_netmap.h) and the device-specific patches (in sys/dev/ixgbe/ixgbe.c) are heavily commented and should serve as a reference for other device drivers. Tested on i386 and amd64 with the pkt-gen program in tools/tools/netmap, the sender does 14.88 Mpps at 1050 Mhz and 14.2 Mpps at 900 MHz on an i7-860 with 4 cores and 82599 card. Haven't tried yet more aggressive optimizations such as adding 'prefetch' instructions in the time-critical parts of the code.
2011-12-05 12:06:53 +00:00
* that extend the standard driver.
*/
#include <dev/netmap/ixgbe_netmap.h>
#endif /* DEV_NETMAP */
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/*********************************************************************
* Device identification routine
*
* ixgbe_probe determines if the driver should be loaded on
* adapter based on PCI vendor/device id of the adapter.
*
First off: update the driver README, the old one was horribly crusty, and this still isn't perfect, but its at least a bit more recent. Secondly, a few improvements to the driver from Andrew Boyer, support hint to allow devices to not attach, add VLAN_HWTSO capability so vlans can use TSO, fix in the interrupt handler to make sure the stack TX queue is processed. Oh, and also make sure IPv6 does not cause a re-init in the ioctl routine. Thanks for your efforts Andrew! Thanks to Claudio Jeker for noticing the ixgbe_xmit() routine was not correctly swapping the dma map from the first to the last descriptor in a multi-descriptor transmission, corrected this.
2011-06-02 00:34:57 +00:00
* return BUS_PROBE_DEFAULT on success, positive on failure
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
*********************************************************************/
static int
ixgbe_probe(device_t dev)
{
ixgbe_vendor_info_t *ent;
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
u16 pci_vendor_id = 0;
u16 pci_device_id = 0;
u16 pci_subvendor_id = 0;
u16 pci_subdevice_id = 0;
char adapter_name[256];
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
INIT_DEBUGOUT("ixgbe_probe: begin");
pci_vendor_id = pci_get_vendor(dev);
if (pci_vendor_id != IXGBE_INTEL_VENDOR_ID)
return (ENXIO);
pci_device_id = pci_get_device(dev);
pci_subvendor_id = pci_get_subvendor(dev);
pci_subdevice_id = pci_get_subdevice(dev);
ent = ixgbe_vendor_info_array;
while (ent->vendor_id != 0) {
if ((pci_vendor_id == ent->vendor_id) &&
(pci_device_id == ent->device_id) &&
((pci_subvendor_id == ent->subvendor_id) ||
(ent->subvendor_id == 0)) &&
((pci_subdevice_id == ent->subdevice_id) ||
(ent->subdevice_id == 0))) {
sprintf(adapter_name, "%s, Version - %s",
ixgbe_strings[ent->index],
ixgbe_driver_version);
device_set_desc_copy(dev, adapter_name);
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
++ixgbe_total_ports;
First off: update the driver README, the old one was horribly crusty, and this still isn't perfect, but its at least a bit more recent. Secondly, a few improvements to the driver from Andrew Boyer, support hint to allow devices to not attach, add VLAN_HWTSO capability so vlans can use TSO, fix in the interrupt handler to make sure the stack TX queue is processed. Oh, and also make sure IPv6 does not cause a re-init in the ioctl routine. Thanks for your efforts Andrew! Thanks to Claudio Jeker for noticing the ixgbe_xmit() routine was not correctly swapping the dma map from the first to the last descriptor in a multi-descriptor transmission, corrected this.
2011-06-02 00:34:57 +00:00
return (BUS_PROBE_DEFAULT);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
ent++;
}
return (ENXIO);
}
/*********************************************************************
* Device initialization routine
*
* The attach entry point is called when the driver is being loaded.
* This routine identifies the type of hardware, allocates all resources
* and initializes the hardware.
*
* return 0 on success, positive on failure
*********************************************************************/
static int
ixgbe_attach(device_t dev)
{
struct adapter *adapter;
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
struct ixgbe_hw *hw;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
int error = 0;
Update driver to version 2.3.8: CRITICAL FIX - with stats changes the older 82598 will panic and trash the stack on driver load, FCOE registers ONLY exist in 82599 and must not be read otherwise. kern/153951 - to correct incorrect media type on adapters with pluggable modules I have eliminated the old static table in favor of a new dynamic shared code routine. This also has the benefit of detecting changes when a different module is inserted. Performance/enhancement to the Flow Director code from my linux coworker (the developer of the code). Fixes from Michael Tuexen - a data corruption problem on the 82599 (CRITICAL), fix so the buf size correctly adjusts as the cluster changes, and max descriptors are set properly. Also added 16K clusters for those REALLY big jumbos :) In the RX path, the RX LOCK was not being released, and this causes LOR problems. Add the code that igb already has. Sync with in house shared code, this was necessary for the Flow Director fix. MFC in 2 days
2011-01-19 19:36:27 +00:00
u16 csum;
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
u32 ctrl_ext;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
INIT_DEBUGOUT("ixgbe_attach: begin");
/* Allocate, clear, and link in our adapter structure */
adapter = device_get_softc(dev);
adapter->dev = adapter->osdep.dev = dev;
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
hw = &adapter->hw;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
/* Core Lock Init*/
IXGBE_CORE_LOCK_INIT(adapter, device_get_nameunit(dev));
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/* Set up the timer callout */
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
callout_init_mtx(&adapter->timer, &adapter->core_mtx, 0);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/* Determine hardware revision */
ixgbe_identify_hardware(adapter);
/* Do base PCI setup - map BAR0 */
if (ixgbe_allocate_pci_resources(adapter)) {
device_printf(dev, "Allocation of PCI resources failed\n");
error = ENXIO;
goto err_out;
}
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
/* Do descriptor calc and sanity checks */
if (((ixgbe_txd * sizeof(union ixgbe_adv_tx_desc)) % DBA_ALIGN) != 0 ||
ixgbe_txd < MIN_TXD || ixgbe_txd > MAX_TXD) {
device_printf(dev, "TXD config issue, using default!\n");
adapter->num_tx_desc = DEFAULT_TXD;
} else
adapter->num_tx_desc = ixgbe_txd;
/*
** With many RX rings it is easy to exceed the
** system mbuf allocation. Tuning nmbclusters
** can alleviate this.
*/
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
if (nmbclusters > 0) {
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
int s;
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
s = (ixgbe_rxd * adapter->num_queues) * ixgbe_total_ports;
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
if (s > nmbclusters) {
device_printf(dev, "RX Descriptors exceed "
"system mbuf max, using default instead!\n");
ixgbe_rxd = DEFAULT_RXD;
}
}
if (((ixgbe_rxd * sizeof(union ixgbe_adv_rx_desc)) % DBA_ALIGN) != 0 ||
Use the right constants in comparisons. This is currently a nop, as MIN_RXD == MIN_TXD and MAX_RXD == MAX_TXD. Reviewed by: Eric Joyner @ Intel MFC after: 1 week
2014-09-08 19:24:25 +00:00
ixgbe_rxd < MIN_RXD || ixgbe_rxd > MAX_RXD) {
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
device_printf(dev, "RXD config issue, using default!\n");
adapter->num_rx_desc = DEFAULT_RXD;
} else
adapter->num_rx_desc = ixgbe_rxd;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/* Allocate our TX/RX Queues */
if (ixgbe_allocate_queues(adapter)) {
error = ENOMEM;
goto err_out;
}
A couple fixes got clobbered, putting them back.
2010-11-26 23:57:13 +00:00
/* Allocate multicast array memory. */
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
adapter->mta = malloc(sizeof(u8) * IXGBE_ETH_LENGTH_OF_ADDRESS *
A couple fixes got clobbered, putting them back.
2010-11-26 23:57:13 +00:00
MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT);
if (adapter->mta == NULL) {
device_printf(dev, "Can not allocate multicast setup array\n");
error = ENOMEM;
goto err_late;
}
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/* Initialize the shared code */
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
hw->allow_unsupported_sfp = allow_unsupported_sfp;
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
error = ixgbe_init_shared_code(hw);
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
if (error == IXGBE_ERR_SFP_NOT_PRESENT) {
/*
** No optics in this port, set up
** so the timer routine will probe
** for later insertion.
*/
adapter->sfp_probe = TRUE;
error = 0;
} else if (error == IXGBE_ERR_SFP_NOT_SUPPORTED) {
device_printf(dev,"Unsupported SFP+ module detected!\n");
error = EIO;
goto err_late;
} else if (error) {
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
device_printf(dev,"Unable to initialize the shared code\n");
error = EIO;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
goto err_late;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
/* Make sure we have a good EEPROM before we read from it */
if (ixgbe_validate_eeprom_checksum(&adapter->hw, &csum) < 0) {
device_printf(dev,"The EEPROM Checksum Is Not Valid\n");
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
error = EIO;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
goto err_late;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
error = ixgbe_init_hw(hw);
Update to the ixgbe driver: - Add a couple of new devices - Flow control changes in shared and core code - Bug fix to Flow Director for 82598 - Shared code sync to internal with required core change Thanks to those helping in the testing and improvements to this driver! MFC after:5 days
2012-07-05 20:51:44 +00:00
switch (error) {
case IXGBE_ERR_EEPROM_VERSION:
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
device_printf(dev, "This device is a pre-production adapter/"
"LOM. Please be aware there may be issues associated "
"with your hardware.\n If you are experiencing problems "
"please contact your Intel or hardware representative "
"who provided you with this hardware.\n");
Update to the ixgbe driver: - Add a couple of new devices - Flow control changes in shared and core code - Bug fix to Flow Director for 82598 - Shared code sync to internal with required core change Thanks to those helping in the testing and improvements to this driver! MFC after:5 days
2012-07-05 20:51:44 +00:00
break;
case IXGBE_ERR_SFP_NOT_SUPPORTED:
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
device_printf(dev,"Unsupported SFP+ Module\n");
error = EIO;
goto err_late;
Update to the ixgbe driver: - Add a couple of new devices - Flow control changes in shared and core code - Bug fix to Flow Director for 82598 - Shared code sync to internal with required core change Thanks to those helping in the testing and improvements to this driver! MFC after:5 days
2012-07-05 20:51:44 +00:00
case IXGBE_ERR_SFP_NOT_PRESENT:
device_printf(dev,"No SFP+ Module found\n");
/* falls thru */
default:
break;
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
}
Update driver to version 2.3.8: CRITICAL FIX - with stats changes the older 82598 will panic and trash the stack on driver load, FCOE registers ONLY exist in 82599 and must not be read otherwise. kern/153951 - to correct incorrect media type on adapters with pluggable modules I have eliminated the old static table in favor of a new dynamic shared code routine. This also has the benefit of detecting changes when a different module is inserted. Performance/enhancement to the Flow Director code from my linux coworker (the developer of the code). Fixes from Michael Tuexen - a data corruption problem on the 82599 (CRITICAL), fix so the buf size correctly adjusts as the cluster changes, and max descriptors are set properly. Also added 16K clusters for those REALLY big jumbos :) In the RX path, the RX LOCK was not being released, and this causes LOR problems. Add the code that igb already has. Sync with in house shared code, this was necessary for the Flow Director fix. MFC in 2 days
2011-01-19 19:36:27 +00:00
/* Detect and set physical type */
ixgbe_setup_optics(adapter);
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
if ((adapter->msix > 1) && (ixgbe_enable_msix))
error = ixgbe_allocate_msix(adapter);
else
error = ixgbe_allocate_legacy(adapter);
if (error)
goto err_late;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/* Setup OS specific network interface */
A couple fixes got clobbered, putting them back.
2010-11-26 23:57:13 +00:00
if (ixgbe_setup_interface(dev, adapter) != 0)
goto err_late;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/* Initialize statistics */
ixgbe_update_stats_counters(adapter);
This updates the ixgbe driver to Intel internal version 1.4.7 Shared code changes, core driver fixes, vlan event/filter support Also Kip Macy's fix to allow any number of queues, thanks Kip!
2008-07-30 18:15:18 +00:00
/* Register for VLAN events */
adapter->vlan_attach = EVENTHANDLER_REGISTER(vlan_config,
Improvement on the last change, this gives a precise way to tell the one and only interface that a vlan event is for. Thanks to John Baldwin for the patch. Approved by: re
2009-07-24 21:35:52 +00:00
ixgbe_register_vlan, adapter, EVENTHANDLER_PRI_FIRST);
This updates the ixgbe driver to Intel internal version 1.4.7 Shared code changes, core driver fixes, vlan event/filter support Also Kip Macy's fix to allow any number of queues, thanks Kip!
2008-07-30 18:15:18 +00:00
adapter->vlan_detach = EVENTHANDLER_REGISTER(vlan_unconfig,
Improvement on the last change, this gives a precise way to tell the one and only interface that a vlan event is for. Thanks to John Baldwin for the patch. Approved by: re
2009-07-24 21:35:52 +00:00
ixgbe_unregister_vlan, adapter, EVENTHANDLER_PRI_FIRST);
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
/* Check PCIE slot type/speed/width */
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
ixgbe_get_slot_info(hw);
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
First, sync to internal shared code, and then Fixes: - flow control - don't override user value on re-init - fix to make 1G optics work correctly - change to interrupt enabling - some bits were incorrect for certain hardware. - certain stats fixes, remove a duplicate increment of ierror, thanks to Scott Long for pointing these out. - shared code link interface changed, requiring some core code changes to accomodate this. - add an m_adj() to ETHER_ALIGN on the recieve side, this was requested by Mike Karels, thanks Mike. - Multicast code corrections also thanks to Mike Karels.
2013-03-04 23:07:40 +00:00
/* Set an initial default flow control value */
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
adapter->fc = ixgbe_fc_full;
First, sync to internal shared code, and then Fixes: - flow control - don't override user value on re-init - fix to make 1G optics work correctly - change to interrupt enabling - some bits were incorrect for certain hardware. - certain stats fixes, remove a duplicate increment of ierror, thanks to Scott Long for pointing these out. - shared code link interface changed, requiring some core code changes to accomodate this. - add an m_adj() to ETHER_ALIGN on the recieve side, this was requested by Mike Karels, thanks Mike. - Multicast code corrections also thanks to Mike Karels.
2013-03-04 23:07:40 +00:00
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
/* Check for certain supported features */
ixgbe_check_wol_support(adapter);
ixgbe_check_eee_support(adapter);
/* Add sysctls */
ixgbe_add_device_sysctls(adapter);
ixgbe_add_hw_stats(adapter);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/* let hardware know driver is loaded */
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
ctrl_ext |= IXGBE_CTRL_EXT_DRV_LOAD;
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
1. Fix the handling of link reset while in netmap more. A link reset now is completely transparent for the netmap client: even if the NIC resets its own ring (e.g. restarting from 0), the client will not see any change in the current rx/tx positions, because the driver will keep track of the offset between the two. 2. make the device-specific code more uniform across different drivers There were some inconsistencies in the implementation of the netmap support routines, now drivers have been aligned to a common code structure. 3. import netmap support for ixgbe . This is implemented as a very small patch for ixgbe.c (233 lines, 11 chunks, mostly comments: in total the patch has only 54 lines of new code) , as most of the code is in an external file sys/dev/netmap/ixgbe_netmap.h , following some initial comments from Jack Vogel about making changes less intrusive. (Note, i have emailed Jack multiple times asking if he had comments on this structure of the code; i got no reply so i assume he is fine with it). Support for other drivers (em, lem, re, igb) will come later. "ixgbe" is now the reference driver for netmap support. Both the external file (sys/dev/netmap/ixgbe_netmap.h) and the device-specific patches (in sys/dev/ixgbe/ixgbe.c) are heavily commented and should serve as a reference for other device drivers. Tested on i386 and amd64 with the pkt-gen program in tools/tools/netmap, the sender does 14.88 Mpps at 1050 Mhz and 14.2 Mpps at 900 MHz on an i7-860 with 4 cores and 82599 card. Haven't tried yet more aggressive optimizations such as adding 'prefetch' instructions in the time-critical parts of the code.
2011-12-05 12:06:53 +00:00
#ifdef DEV_NETMAP
ixgbe_netmap_attach(adapter);
#endif /* DEV_NETMAP */
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
INIT_DEBUGOUT("ixgbe_attach: end");
return (0);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
err_late:
ixgbe_free_transmit_structures(adapter);
ixgbe_free_receive_structures(adapter);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
err_out:
A couple fixes got clobbered, putting them back.
2010-11-26 23:57:13 +00:00
if (adapter->ifp != NULL)
if_free(adapter->ifp);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
ixgbe_free_pci_resources(adapter);
A couple fixes got clobbered, putting them back.
2010-11-26 23:57:13 +00:00
free(adapter->mta, M_DEVBUF);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
return (error);
}
/*********************************************************************
* Device removal routine
*
* The detach entry point is called when the driver is being removed.
* This routine stops the adapter and deallocates all the resources
* that were allocated for driver operation.
*
* return 0 on success, positive on failure
*********************************************************************/
static int
ixgbe_detach(device_t dev)
{
struct adapter *adapter = device_get_softc(dev);
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
struct ix_queue *que = adapter->queues;
Merge similar fixes from 223198 from igb to ixgbe: - Use a dedicated task to handle deferred transmits from the if_transmit method instead of reusing the existing per-queue interrupt task. Reusing the per-queue interrupt task could result in both an interrupt thread and the taskqueue thread trying to handle received packets on a single queue resulting in out-of-order packet processing and lock contention. - Don't define ixgbe_start() at all where if_transmit is used. Tested by: Vijay Singh Reviewed by: jfv MFC after: 2 weeks
2012-09-26 18:11:43 +00:00
struct tx_ring *txr = adapter->tx_rings;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
u32 ctrl_ext;
INIT_DEBUGOUT("ixgbe_detach: begin");
/* Make sure VLANS are not using driver */
if (adapter->ifp->if_vlantrunk != NULL) {
device_printf(dev,"Vlan in use, detach first\n");
return (EBUSY);
}
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
/* Stop the adapter */
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
IXGBE_CORE_LOCK(adapter);
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
ixgbe_setup_low_power_mode(adapter);
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
IXGBE_CORE_UNLOCK(adapter);
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
Merge similar fixes from 223198 from igb to ixgbe: - Use a dedicated task to handle deferred transmits from the if_transmit method instead of reusing the existing per-queue interrupt task. Reusing the per-queue interrupt task could result in both an interrupt thread and the taskqueue thread trying to handle received packets on a single queue resulting in out-of-order packet processing and lock contention. - Don't define ixgbe_start() at all where if_transmit is used. Tested by: Vijay Singh Reviewed by: jfv MFC after: 2 weeks
2012-09-26 18:11:43 +00:00
for (int i = 0; i < adapter->num_queues; i++, que++, txr++) {
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
if (que->tq) {
Fix a small, but important bug, a task drain was mistakenly being compiled only when setting LEGACY_TX, this means you would not get the drain when needed on detach!! Thanks to Bryan Venteicher (bryanv@freebsd.org) for catching this little gremlin!! :)
2013-03-04 23:15:07 +00:00
#ifndef IXGBE_LEGACY_TX
Merge similar fixes from 223198 from igb to ixgbe: - Use a dedicated task to handle deferred transmits from the if_transmit method instead of reusing the existing per-queue interrupt task. Reusing the per-queue interrupt task could result in both an interrupt thread and the taskqueue thread trying to handle received packets on a single queue resulting in out-of-order packet processing and lock contention. - Don't define ixgbe_start() at all where if_transmit is used. Tested by: Vijay Singh Reviewed by: jfv MFC after: 2 weeks
2012-09-26 18:11:43 +00:00
taskqueue_drain(que->tq, &txr->txq_task);
#endif
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
taskqueue_drain(que->tq, &que->que_task);
taskqueue_free(que->tq);
This updates the ixgbe driver to Intel internal version 1.4.7 Shared code changes, core driver fixes, vlan event/filter support Also Kip Macy's fix to allow any number of queues, thanks Kip!
2008-07-30 18:15:18 +00:00
}
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
}
This updates the ixgbe driver to Intel internal version 1.4.7 Shared code changes, core driver fixes, vlan event/filter support Also Kip Macy's fix to allow any number of queues, thanks Kip!
2008-07-30 18:15:18 +00:00
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
/* Drain the Link queue */
if (adapter->tq) {
taskqueue_drain(adapter->tq, &adapter->link_task);
taskqueue_drain(adapter->tq, &adapter->mod_task);
taskqueue_drain(adapter->tq, &adapter->msf_task);
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
taskqueue_drain(adapter->tq, &adapter->phy_task);
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
#ifdef IXGBE_FDIR
taskqueue_drain(adapter->tq, &adapter->fdir_task);
#endif
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
taskqueue_free(adapter->tq);
}
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
/* let hardware know driver is unloading */
ctrl_ext = IXGBE_READ_REG(&adapter->hw, IXGBE_CTRL_EXT);
ctrl_ext &= ~IXGBE_CTRL_EXT_DRV_LOAD;
IXGBE_WRITE_REG(&adapter->hw, IXGBE_CTRL_EXT, ctrl_ext);
This updates the ixgbe driver to Intel internal version 1.4.7 Shared code changes, core driver fixes, vlan event/filter support Also Kip Macy's fix to allow any number of queues, thanks Kip!
2008-07-30 18:15:18 +00:00
/* Unregister VLAN events */
if (adapter->vlan_attach != NULL)
EVENTHANDLER_DEREGISTER(vlan_config, adapter->vlan_attach);
if (adapter->vlan_detach != NULL)
EVENTHANDLER_DEREGISTER(vlan_unconfig, adapter->vlan_detach);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
ether_ifdetach(adapter->ifp);
callout_drain(&adapter->timer);
1. Fix the handling of link reset while in netmap more. A link reset now is completely transparent for the netmap client: even if the NIC resets its own ring (e.g. restarting from 0), the client will not see any change in the current rx/tx positions, because the driver will keep track of the offset between the two. 2. make the device-specific code more uniform across different drivers There were some inconsistencies in the implementation of the netmap support routines, now drivers have been aligned to a common code structure. 3. import netmap support for ixgbe . This is implemented as a very small patch for ixgbe.c (233 lines, 11 chunks, mostly comments: in total the patch has only 54 lines of new code) , as most of the code is in an external file sys/dev/netmap/ixgbe_netmap.h , following some initial comments from Jack Vogel about making changes less intrusive. (Note, i have emailed Jack multiple times asking if he had comments on this structure of the code; i got no reply so i assume he is fine with it). Support for other drivers (em, lem, re, igb) will come later. "ixgbe" is now the reference driver for netmap support. Both the external file (sys/dev/netmap/ixgbe_netmap.h) and the device-specific patches (in sys/dev/ixgbe/ixgbe.c) are heavily commented and should serve as a reference for other device drivers. Tested on i386 and amd64 with the pkt-gen program in tools/tools/netmap, the sender does 14.88 Mpps at 1050 Mhz and 14.2 Mpps at 900 MHz on an i7-860 with 4 cores and 82599 card. Haven't tried yet more aggressive optimizations such as adding 'prefetch' instructions in the time-critical parts of the code.
2011-12-05 12:06:53 +00:00
#ifdef DEV_NETMAP
netmap_detach(adapter->ifp);
#endif /* DEV_NETMAP */
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
ixgbe_free_pci_resources(adapter);
bus_generic_detach(dev);
if_free(adapter->ifp);
ixgbe_free_transmit_structures(adapter);
ixgbe_free_receive_structures(adapter);
A couple fixes got clobbered, putting them back.
2010-11-26 23:57:13 +00:00
free(adapter->mta, M_DEVBUF);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
IXGBE_CORE_LOCK_DESTROY(adapter);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
return (0);
}
/*********************************************************************
*
* Shutdown entry point
*
**********************************************************************/
static int
ixgbe_shutdown(device_t dev)
{
struct adapter *adapter = device_get_softc(dev);
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
int error = 0;
INIT_DEBUGOUT("ixgbe_shutdown: begin");
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
IXGBE_CORE_LOCK(adapter);
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
error = ixgbe_setup_low_power_mode(adapter);
IXGBE_CORE_UNLOCK(adapter);
return (error);
}
/**
* Methods for going from:
* D0 -> D3: ixgbe_suspend
* D3 -> D0: ixgbe_resume
*/
static int
ixgbe_suspend(device_t dev)
{
struct adapter *adapter = device_get_softc(dev);
int error = 0;
INIT_DEBUGOUT("ixgbe_suspend: begin");
IXGBE_CORE_LOCK(adapter);
error = ixgbe_setup_low_power_mode(adapter);
/* Save state and power down */
pci_save_state(dev);
pci_set_powerstate(dev, PCI_POWERSTATE_D3);
IXGBE_CORE_UNLOCK(adapter);
return (error);
}
static int
ixgbe_resume(device_t dev)
{
struct adapter *adapter = device_get_softc(dev);
struct ifnet *ifp = adapter->ifp;
struct ixgbe_hw *hw = &adapter->hw;
u32 wus;
INIT_DEBUGOUT("ixgbe_resume: begin");
IXGBE_CORE_LOCK(adapter);
pci_set_powerstate(dev, PCI_POWERSTATE_D0);
pci_restore_state(dev);
/* Read & clear WUS register */
wus = IXGBE_READ_REG(hw, IXGBE_WUS);
if (wus)
device_printf(dev, "Woken up by (WUS): %#010x\n",
IXGBE_READ_REG(hw, IXGBE_WUS));
IXGBE_WRITE_REG(hw, IXGBE_WUS, 0xffffffff);
/* And clear WUFC until next low-power transition */
IXGBE_WRITE_REG(hw, IXGBE_WUFC, 0);
/*
* Required after D3->D0 transition;
* will re-advertise all previous advertised speeds
*/
if (ifp->if_flags & IFF_UP)
ixgbe_init_locked(adapter);
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
IXGBE_CORE_UNLOCK(adapter);
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
INIT_DEBUGOUT("ixgbe_resume: end");
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
return (0);
}
/*********************************************************************
* Ioctl entry point
*
* ixgbe_ioctl is called when the user wants to configure the
* interface.
*
* return 0 on success, positive on failure
**********************************************************************/
static int
ixgbe_ioctl(struct ifnet * ifp, u_long command, caddr_t data)
{
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
struct adapter *adapter = ifp->if_softc;
struct ifreq *ifr = (struct ifreq *) data;
First off: update the driver README, the old one was horribly crusty, and this still isn't perfect, but its at least a bit more recent. Secondly, a few improvements to the driver from Andrew Boyer, support hint to allow devices to not attach, add VLAN_HWTSO capability so vlans can use TSO, fix in the interrupt handler to make sure the stack TX queue is processed. Oh, and also make sure IPv6 does not cause a re-init in the ioctl routine. Thanks for your efforts Andrew! Thanks to Claudio Jeker for noticing the ixgbe_xmit() routine was not correctly swapping the dma map from the first to the last descriptor in a multi-descriptor transmission, corrected this.
2011-06-02 00:34:57 +00:00
#if defined(INET) || defined(INET6)
- Add the RX refresh changes from igb to ixgbe - Also a couple minor tweaks to the TX code from the same source. - Add the INET ioctl code which has been missing from this driver, and which caused IP aliases to reset the interface. - Last, some minor logic changes that just reflect upcoming hardware support, but have no other functional effect now. MFC after a week
2011-04-25 23:34:21 +00:00
struct ifaddr *ifa = (struct ifaddr *)data;
First off: update the driver README, the old one was horribly crusty, and this still isn't perfect, but its at least a bit more recent. Secondly, a few improvements to the driver from Andrew Boyer, support hint to allow devices to not attach, add VLAN_HWTSO capability so vlans can use TSO, fix in the interrupt handler to make sure the stack TX queue is processed. Oh, and also make sure IPv6 does not cause a re-init in the ioctl routine. Thanks for your efforts Andrew! Thanks to Claudio Jeker for noticing the ixgbe_xmit() routine was not correctly swapping the dma map from the first to the last descriptor in a multi-descriptor transmission, corrected this.
2011-06-02 00:34:57 +00:00
bool avoid_reset = FALSE;
- Add the RX refresh changes from igb to ixgbe - Also a couple minor tweaks to the TX code from the same source. - Add the INET ioctl code which has been missing from this driver, and which caused IP aliases to reset the interface. - Last, some minor logic changes that just reflect upcoming hardware support, but have no other functional effect now. MFC after a week
2011-04-25 23:34:21 +00:00
#endif
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
int error = 0;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
switch (command) {
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
- Add the RX refresh changes from igb to ixgbe - Also a couple minor tweaks to the TX code from the same source. - Add the INET ioctl code which has been missing from this driver, and which caused IP aliases to reset the interface. - Last, some minor logic changes that just reflect upcoming hardware support, but have no other functional effect now. MFC after a week
2011-04-25 23:34:21 +00:00
case SIOCSIFADDR:
#ifdef INET
First off: update the driver README, the old one was horribly crusty, and this still isn't perfect, but its at least a bit more recent. Secondly, a few improvements to the driver from Andrew Boyer, support hint to allow devices to not attach, add VLAN_HWTSO capability so vlans can use TSO, fix in the interrupt handler to make sure the stack TX queue is processed. Oh, and also make sure IPv6 does not cause a re-init in the ioctl routine. Thanks for your efforts Andrew! Thanks to Claudio Jeker for noticing the ixgbe_xmit() routine was not correctly swapping the dma map from the first to the last descriptor in a multi-descriptor transmission, corrected this.
2011-06-02 00:34:57 +00:00
if (ifa->ifa_addr->sa_family == AF_INET)
avoid_reset = TRUE;
#endif
#ifdef INET6
if (ifa->ifa_addr->sa_family == AF_INET6)
avoid_reset = TRUE;
#endif
#if defined(INET) || defined(INET6)
/*
** Calling init results in link renegotiation,
** so we avoid doing it when possible.
*/
if (avoid_reset) {
- Add the RX refresh changes from igb to ixgbe - Also a couple minor tweaks to the TX code from the same source. - Add the INET ioctl code which has been missing from this driver, and which caused IP aliases to reset the interface. - Last, some minor logic changes that just reflect upcoming hardware support, but have no other functional effect now. MFC after a week
2011-04-25 23:34:21 +00:00
ifp->if_flags |= IFF_UP;
First off: update the driver README, the old one was horribly crusty, and this still isn't perfect, but its at least a bit more recent. Secondly, a few improvements to the driver from Andrew Boyer, support hint to allow devices to not attach, add VLAN_HWTSO capability so vlans can use TSO, fix in the interrupt handler to make sure the stack TX queue is processed. Oh, and also make sure IPv6 does not cause a re-init in the ioctl routine. Thanks for your efforts Andrew! Thanks to Claudio Jeker for noticing the ixgbe_xmit() routine was not correctly swapping the dma map from the first to the last descriptor in a multi-descriptor transmission, corrected this.
2011-06-02 00:34:57 +00:00
if (!(ifp->if_drv_flags & IFF_DRV_RUNNING))
ixgbe_init(adapter);
- Add the RX refresh changes from igb to ixgbe - Also a couple minor tweaks to the TX code from the same source. - Add the INET ioctl code which has been missing from this driver, and which caused IP aliases to reset the interface. - Last, some minor logic changes that just reflect upcoming hardware support, but have no other functional effect now. MFC after a week
2011-04-25 23:34:21 +00:00
if (!(ifp->if_flags & IFF_NOARP))
arp_ifinit(ifp, ifa);
} else
error = ether_ioctl(ifp, command, data);
First off: update the driver README, the old one was horribly crusty, and this still isn't perfect, but its at least a bit more recent. Secondly, a few improvements to the driver from Andrew Boyer, support hint to allow devices to not attach, add VLAN_HWTSO capability so vlans can use TSO, fix in the interrupt handler to make sure the stack TX queue is processed. Oh, and also make sure IPv6 does not cause a re-init in the ioctl routine. Thanks for your efforts Andrew! Thanks to Claudio Jeker for noticing the ixgbe_xmit() routine was not correctly swapping the dma map from the first to the last descriptor in a multi-descriptor transmission, corrected this.
2011-06-02 00:34:57 +00:00
#endif
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
break;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
case SIOCSIFMTU:
IOCTL_DEBUGOUT("ioctl: SIOCSIFMTU (Set Interface MTU)");
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
if (ifr->ifr_mtu > IXGBE_MAX_MTU) {
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
error = EINVAL;
} else {
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
IXGBE_CORE_LOCK(adapter);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
ifp->if_mtu = ifr->ifr_mtu;
adapter->max_frame_size =
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
ifp->if_mtu + IXGBE_MTU_HDR;
Fixes for panic experienced in test at Intel, when doing bidirectional stress traffic on 82598. Also a couple bug fixes from Michael Tuexen, thank you!! Add a workaround into the header so that 8 REL can use the driver (adds local copy of ALTQ fix). MFC: in a few days
2010-06-03 00:00:45 +00:00
ixgbe_init_locked(adapter);
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
IXGBE_CORE_UNLOCK(adapter);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
break;
case SIOCSIFFLAGS:
IOCTL_DEBUGOUT("ioctl: SIOCSIFFLAGS (Set Interface Flags)");
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
IXGBE_CORE_LOCK(adapter);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
if (ifp->if_flags & IFF_UP) {
A couple late breaking bugs that testing have turned up. - change include style so build in kernel try OR standalone work. - Limit HWCSUM - I was led to believe that it would work with RSS, but our testing had odd issues which suggests this is false. - A fatfinger error in the ioctl code made ifconfig up not work. Approved by: re
2007-07-12 19:04:11 +00:00
if ((ifp->if_drv_flags & IFF_DRV_RUNNING)) {
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
if ((ifp->if_flags ^ adapter->if_flags) &
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
(IFF_PROMISC | IFF_ALLMULTI)) {
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
ixgbe_set_promisc(adapter);
}
} else
Fixes for panic experienced in test at Intel, when doing bidirectional stress traffic on 82598. Also a couple bug fixes from Michael Tuexen, thank you!! Add a workaround into the header so that 8 REL can use the driver (adds local copy of ALTQ fix). MFC: in a few days
2010-06-03 00:00:45 +00:00
ixgbe_init_locked(adapter);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
} else
if (ifp->if_drv_flags & IFF_DRV_RUNNING)
ixgbe_stop(adapter);
adapter->if_flags = ifp->if_flags;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
IXGBE_CORE_UNLOCK(adapter);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
break;
case SIOCADDMULTI:
case SIOCDELMULTI:
IOCTL_DEBUGOUT("ioctl: SIOC(ADD|DEL)MULTI");
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
IXGBE_CORE_LOCK(adapter);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
ixgbe_disable_intr(adapter);
ixgbe_set_multi(adapter);
ixgbe_enable_intr(adapter);
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
IXGBE_CORE_UNLOCK(adapter);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
break;
case SIOCSIFMEDIA:
case SIOCGIFMEDIA:
IOCTL_DEBUGOUT("ioctl: SIOCxIFMEDIA (Get/Set Interface Media)");
error = ifmedia_ioctl(ifp, ifr, &adapter->media, command);
break;
case SIOCSIFCAP:
{
int mask = ifr->ifr_reqcap ^ ifp->if_capenable;
IOCTL_DEBUGOUT("ioctl: SIOCSIFCAP (Set Capabilities)");
if (mask & IFCAP_HWCSUM)
ifp->if_capenable ^= IFCAP_HWCSUM;
if (mask & IFCAP_TSO4)
ifp->if_capenable ^= IFCAP_TSO4;
MFp4 bz_ipv6_fast: Add TSO6 and LRO/IPv6 support. Fix the module Makefile to at least properly inlcude opt_inet6.h and allow builds without INET or INET6. Sponsored by: The FreeBSD Foundation Sponsored by: iXsystems Reviewed by: gnn (as part of the whole) MFC After: 3 days
2012-05-25 03:02:56 +00:00
if (mask & IFCAP_TSO6)
ifp->if_capenable ^= IFCAP_TSO6;
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
if (mask & IFCAP_LRO)
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
ifp->if_capenable ^= IFCAP_LRO;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
if (mask & IFCAP_VLAN_HWTAGGING)
ifp->if_capenable ^= IFCAP_VLAN_HWTAGGING;
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
if (mask & IFCAP_VLAN_HWFILTER)
ifp->if_capenable ^= IFCAP_VLAN_HWFILTER;
First off: update the driver README, the old one was horribly crusty, and this still isn't perfect, but its at least a bit more recent. Secondly, a few improvements to the driver from Andrew Boyer, support hint to allow devices to not attach, add VLAN_HWTSO capability so vlans can use TSO, fix in the interrupt handler to make sure the stack TX queue is processed. Oh, and also make sure IPv6 does not cause a re-init in the ioctl routine. Thanks for your efforts Andrew! Thanks to Claudio Jeker for noticing the ixgbe_xmit() routine was not correctly swapping the dma map from the first to the last descriptor in a multi-descriptor transmission, corrected this.
2011-06-02 00:34:57 +00:00
if (mask & IFCAP_VLAN_HWTSO)
ifp->if_capenable ^= IFCAP_VLAN_HWTSO;
Thanks to Michael Tuexen for adding SCTP support for 82599, also for finding a one character bug that kept TSO from working. Sometimes with direct attach cables a failure can occur in init, the old method of calling detach was broken, there is no way to return an error to the system from init, so I have changed it to return failure thru the ioctl. And, have fixed the ALTQ code changes of Max Laier, sorry Max :)
2010-03-30 19:09:18 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
IXGBE_CORE_LOCK(adapter);
Fixes for panic experienced in test at Intel, when doing bidirectional stress traffic on 82598. Also a couple bug fixes from Michael Tuexen, thank you!! Add a workaround into the header so that 8 REL can use the driver (adds local copy of ALTQ fix). MFC: in a few days
2010-06-03 00:00:45 +00:00
ixgbe_init_locked(adapter);
Thanks to Michael Tuexen for adding SCTP support for 82599, also for finding a one character bug that kept TSO from working. Sometimes with direct attach cables a failure can occur in init, the old method of calling detach was broken, there is no way to return an error to the system from init, so I have changed it to return failure thru the ioctl. And, have fixed the ALTQ code changes of Max Laier, sorry Max :)
2010-03-30 19:09:18 +00:00
IXGBE_CORE_UNLOCK(adapter);
}
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
VLAN_CAPABILITIES(ifp);
break;
}
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
#if __FreeBSD_version >= 1100036
Patch #3 - Add a new ioctl to access SFP+ module diagnostic data via the I2C routines in shared code.
2012-11-30 22:41:32 +00:00
case SIOCGI2C:
{
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
struct ixgbe_hw *hw = &adapter->hw;
* Add SIOCGI2C driver ioctl used to retrieve i2c info. * Convert ixgbe to use this ioctl * Convert ifconfig to use generic i2c handler for "ix" interfaces. Approved by: Eric Joyner (ixgbe part) MFC after: 2 weeks Sponsored by: Yandex LLC
2014-08-29 18:02:58 +00:00
struct ifi2creq i2c;
int i;
Patch #3 - Add a new ioctl to access SFP+ module diagnostic data via the I2C routines in shared code.
2012-11-30 22:41:32 +00:00
IOCTL_DEBUGOUT("ioctl: SIOCGI2C (Get I2C Data)");
error = copyin(ifr->ifr_data, &i2c, sizeof(i2c));
* Add SIOCGI2C driver ioctl used to retrieve i2c info. * Convert ixgbe to use this ioctl * Convert ifconfig to use generic i2c handler for "ix" interfaces. Approved by: Eric Joyner (ixgbe part) MFC after: 2 weeks Sponsored by: Yandex LLC
2014-08-29 18:02:58 +00:00
if (error != 0)
Patch #3 - Add a new ioctl to access SFP+ module diagnostic data via the I2C routines in shared code.
2012-11-30 22:41:32 +00:00
break;
* Add SIOCGI2C driver ioctl used to retrieve i2c info. * Convert ixgbe to use this ioctl * Convert ifconfig to use generic i2c handler for "ix" interfaces. Approved by: Eric Joyner (ixgbe part) MFC after: 2 weeks Sponsored by: Yandex LLC
2014-08-29 18:02:58 +00:00
if (i2c.dev_addr != 0xA0 && i2c.dev_addr != 0xA2) {
Patch #3 - Add a new ioctl to access SFP+ module diagnostic data via the I2C routines in shared code.
2012-11-30 22:41:32 +00:00
error = EINVAL;
break;
}
* Add SIOCGI2C driver ioctl used to retrieve i2c info. * Convert ixgbe to use this ioctl * Convert ifconfig to use generic i2c handler for "ix" interfaces. Approved by: Eric Joyner (ixgbe part) MFC after: 2 weeks Sponsored by: Yandex LLC
2014-08-29 18:02:58 +00:00
if (i2c.len > sizeof(i2c.data)) {
error = EINVAL;
break;
}
for (i = 0; i < i2c.len; i++)
hw->phy.ops.read_i2c_byte(hw, i2c.offset + i,
i2c.dev_addr, &i2c.data[i]);
Patch #3 - Add a new ioctl to access SFP+ module diagnostic data via the I2C routines in shared code.
2012-11-30 22:41:32 +00:00
error = copyout(&i2c, ifr->ifr_data, sizeof(i2c));
break;
}
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
#endif
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
default:
IOCTL_DEBUGOUT1("ioctl: UNKNOWN (0x%X)\n", (int)command);
error = ether_ioctl(ifp, command, data);
break;
}
return (error);
}
/*********************************************************************
* Init entry point
*
* This routine is used in two ways. It is used by the stack as
* init entry point in network interface structure. It is also used
* by the driver as a hw/sw initialization routine to get to a
* consistent state.
*
* return 0 on success, positive on failure
**********************************************************************/
#define IXGBE_MHADD_MFS_SHIFT 16
Fixes for panic experienced in test at Intel, when doing bidirectional stress traffic on 82598. Also a couple bug fixes from Michael Tuexen, thank you!! Add a workaround into the header so that 8 REL can use the driver (adds local copy of ALTQ fix). MFC: in a few days
2010-06-03 00:00:45 +00:00
static void
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
ixgbe_init_locked(struct adapter *adapter)
{
struct ifnet *ifp = adapter->ifp;
device_t dev = adapter->dev;
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
struct ixgbe_hw *hw = &adapter->hw;
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
u32 k, txdctl, mhadd, gpie;
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
u32 rxdctl, rxctrl;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
mtx_assert(&adapter->core_mtx, MA_OWNED);
A number of important fixes: - mbuf reused after an RX_COPY optimized operation can sometimes have a bogus cached address, resulting in TCP hangs. Add critical save points to the cached address. Thanks to Michael and the team at Verisign for finding this problem. - A couple more spots where the rxbuf->flags member should be cleared just to be sure no incorrect RX_COPY state is left around. Thanks to Adrian for tracking these down. - Remove the rearm_queues function from the driver, this was found to be responsible for some out-of-order packets by Verisign, and was always a bandaid, with the other fixes in this delta the bandaid can finally be removed. - In the other/link interrupt handler the entire state of the EICS register was being writen back into EICR (which clears causes and thus re-enables those interrupts), this was wrong, so now mask off the queue portion of the register value, so we only clear the other/link interrupt we intend. Marc from Verisign found this. - Make the SFP+ unsupported option tuneable now, by customer request. - Finally, just a couple of minor DEBUG string fixes. I want to call out and thank all the participants in the 10G community/Intel calls for helping track down these problems and make the driver better for everyone! MFC after: 3 days, these are critical fixes for 9.2!
2013-08-01 20:10:16 +00:00
INIT_DEBUGOUT("ixgbe_init_locked: begin");
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
hw->adapter_stopped = FALSE;
ixgbe_stop_adapter(hw);
callout_stop(&adapter->timer);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
/* reprogram the RAR[0] in case user changed it. */
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
ixgbe_set_rar(hw, 0, adapter->hw.mac.addr, 0, IXGBE_RAH_AV);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/* Get the latest mac address, User can use a LAA */
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
bcopy(IF_LLADDR(adapter->ifp), hw->mac.addr,
IXGBE_ETH_LENGTH_OF_ADDRESS);
ixgbe_set_rar(hw, 0, hw->mac.addr, 0, 1);
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
hw->addr_ctrl.rar_used_count = 1;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
/* Set the various hardware offload abilities */
ifp->if_hwassist = 0;
MFp4 bz_ipv6_fast: Add TSO6 and LRO/IPv6 support. Fix the module Makefile to at least properly inlcude opt_inet6.h and allow builds without INET or INET6. Sponsored by: The FreeBSD Foundation Sponsored by: iXsystems Reviewed by: gnn (as part of the whole) MFC After: 3 days
2012-05-25 03:02:56 +00:00
if (ifp->if_capenable & IFCAP_TSO)
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
ifp->if_hwassist |= CSUM_TSO;
if (ifp->if_capenable & IFCAP_TXCSUM) {
ifp->if_hwassist |= (CSUM_TCP | CSUM_UDP);
#if __FreeBSD_version >= 800000
- Add the RX refresh changes from igb to ixgbe - Also a couple minor tweaks to the TX code from the same source. - Add the INET ioctl code which has been missing from this driver, and which caused IP aliases to reset the interface. - Last, some minor logic changes that just reflect upcoming hardware support, but have no other functional effect now. MFC after a week
2011-04-25 23:34:21 +00:00
if (hw->mac.type != ixgbe_mac_82598EB)
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
ifp->if_hwassist |= CSUM_SCTP;
#endif
}
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/* Prepare transmit descriptors and buffers */
if (ixgbe_setup_transmit_structures(adapter)) {
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
device_printf(dev, "Could not setup transmit structures\n");
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
ixgbe_stop(adapter);
Fixes for panic experienced in test at Intel, when doing bidirectional stress traffic on 82598. Also a couple bug fixes from Michael Tuexen, thank you!! Add a workaround into the header so that 8 REL can use the driver (adds local copy of ALTQ fix). MFC: in a few days
2010-06-03 00:00:45 +00:00
return;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
Update driver to version 2.3.8: CRITICAL FIX - with stats changes the older 82598 will panic and trash the stack on driver load, FCOE registers ONLY exist in 82599 and must not be read otherwise. kern/153951 - to correct incorrect media type on adapters with pluggable modules I have eliminated the old static table in favor of a new dynamic shared code routine. This also has the benefit of detecting changes when a different module is inserted. Performance/enhancement to the Flow Director code from my linux coworker (the developer of the code). Fixes from Michael Tuexen - a data corruption problem on the 82599 (CRITICAL), fix so the buf size correctly adjusts as the cluster changes, and max descriptors are set properly. Also added 16K clusters for those REALLY big jumbos :) In the RX path, the RX LOCK was not being released, and this causes LOR problems. Add the code that igb already has. Sync with in house shared code, this was necessary for the Flow Director fix. MFC in 2 days
2011-01-19 19:36:27 +00:00
ixgbe_init_hw(hw);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
ixgbe_initialize_transmit_units(adapter);
/* Setup Multicast table */
ixgbe_set_multi(adapter);
/*
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
** Determine the correct mbuf pool
First of a series of 11 patches leading to new ixgbe version 2.5.0 This removes the header split and supporting code from the driver.
2012-11-30 22:19:18 +00:00
** for doing jumbo frames
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
*/
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
if (adapter->max_frame_size <= 2048)
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
adapter->rx_mbuf_sz = MCLBYTES;
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
else if (adapter->max_frame_size <= 4096)
Delta D2489 - Add SRIOV support to the Intel 10G driver. NOTE: This is a technology preview, while it has undergone development tests, Intel has not yet completed full validation of the feature. It is being integrated for early access and customer testing.
2015-06-01 17:15:25 +00:00
adapter->rx_mbuf_sz = MJUMPAGESIZE;
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
else if (adapter->max_frame_size <= 9216)
adapter->rx_mbuf_sz = MJUM9BYTES;
else
adapter->rx_mbuf_sz = MJUM16BYTES;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/* Prepare receive descriptors and buffers */
if (ixgbe_setup_receive_structures(adapter)) {
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
device_printf(dev, "Could not setup receive structures\n");
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
ixgbe_stop(adapter);
Fixes for panic experienced in test at Intel, when doing bidirectional stress traffic on 82598. Also a couple bug fixes from Michael Tuexen, thank you!! Add a workaround into the header so that 8 REL can use the driver (adds local copy of ALTQ fix). MFC: in a few days
2010-06-03 00:00:45 +00:00
return;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
/* Configure RX settings */
ixgbe_initialize_receive_units(adapter);
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
gpie = IXGBE_READ_REG(&adapter->hw, IXGBE_GPIE);
/* Enable Fan Failure Interrupt */
gpie |= IXGBE_SDP1_GPIEN_BY_MAC(hw);
/* Add for Module detection */
if (hw->mac.type == ixgbe_mac_82599EB)
gpie |= IXGBE_SDP2_GPIEN;
/*
* Thermal Failure Detection (X540)
* Link Detection (X552)
*/
if (hw->mac.type == ixgbe_mac_X540 ||
hw->device_id == IXGBE_DEV_ID_X550EM_X_SFP ||
hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T)
gpie |= IXGBE_SDP0_GPIEN_X540;
if (adapter->msix > 1) {
/* Enable Enhanced MSIX mode */
gpie |= IXGBE_GPIE_MSIX_MODE;
gpie |= IXGBE_GPIE_EIAME | IXGBE_GPIE_PBA_SUPPORT |
IXGBE_GPIE_OCD;
}
IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/* Set MTU size */
if (ifp->if_mtu > ETHERMTU) {
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
/* aka IXGBE_MAXFRS on 82599 and newer */
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
mhadd = IXGBE_READ_REG(hw, IXGBE_MHADD);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
mhadd &= ~IXGBE_MHADD_MFS_MASK;
mhadd |= adapter->max_frame_size << IXGBE_MHADD_MFS_SHIFT;
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
/* Now enable all the queues */
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
for (int i = 0; i < adapter->num_queues; i++) {
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
txdctl = IXGBE_READ_REG(hw, IXGBE_TXDCTL(i));
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
txdctl |= IXGBE_TXDCTL_ENABLE;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
/* Set WTHRESH to 8, burst writeback */
txdctl |= (8 << 16);
Enable prefetching of descriptors on the TX ring, using the same values as in the Intel driver 3.8.21 for linux. The fact that it is standard in the above driver suggests that it has no bad side effects. But of course there must be a reason for enabling features, not just "it does not harm", so here it is a good one: Prefetching enables full line rate even using a single queue (14.88 Mpps, compared to ~12 Mpps without prefetch). This in turn is terribly useful when one wants to schedule traffic. For obvious reasons the difference is only visible with netmap or other high speed solutions, but presumably the advantage should be in the order of a fraction of a microsecond when starting transmission on an empty queue. Discussed with Jack Vogel. MFC after: 1 week
2012-04-11 15:02:14 +00:00
/*
* When the internal queue falls below PTHRESH (32),
* start prefetching as long as there are at least
* HTHRESH (1) buffers ready. The values are taken
* from the Intel linux driver 3.8.21.
* Prefetching enables tx line rate even with 1 queue.
*/
Correct typo(?) and actually set PTHRESH to 32 and not 16 as per Intel Linux driver 3.8.21. MFC after: 1 week
2012-06-07 22:57:26 +00:00
txdctl |= (32 << 0) | (1 << 8);
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(i), txdctl);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
for (int i = 0; i < adapter->num_queues; i++) {
rxdctl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
if (hw->mac.type == ixgbe_mac_82598EB) {
/*
** PTHRESH = 21
** HTHRESH = 4
** WTHRESH = 8
*/
rxdctl &= ~0x3FFFFF;
rxdctl |= 0x080420;
}
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
rxdctl |= IXGBE_RXDCTL_ENABLE;
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(i), rxdctl);
for (k = 0; k < 10; k++) {
if (IXGBE_READ_REG(hw, IXGBE_RXDCTL(i)) &
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
IXGBE_RXDCTL_ENABLE)
break;
else
msec_delay(1);
}
wmb();
netmap-related changes: 1. correct the initialization of RDT when there is an ixgbe_init() while a netmap client is active. This code was previously in ixgbe_initialize_receive_units() but RDT is overwritten shortly afterwards in ixgbe_init_locked() 2. add code (not active yet) to disable CRCSTRIP while in netmap mode. From all evidence i could gather, it seems that when the 82599 has to write a data block that is not a full cache line, it first reads the line (64 bytes) and then writes back the updated version. This hurts reception of min-sized frames, which are only 60 bytes if the CRC is stripped: i could never get above 11Mpps (received from one queue) with CRCSTRIP enabled, whyle 64+4-byte packets reach 14.2 Mpps (the theoretical maximum). Leaving the CRC in gets us 14.88Mpps for 60+4 byte frames, (and penalizes 64+4). The min-size case is important not just because it looks good in benchmarks, but also because this is the size of pure acks. Note we cannot leave CRCSTRIP on by default because it is incompatible with some other features (LRO etc.)
2012-01-19 09:36:19 +00:00
#ifdef DEV_NETMAP
/*
* In netmap mode, we must preserve the buffers made
* available to userspace before the if_init()
* (this is true by default on the TX side, because
* init makes all buffers available to userspace).
*
* netmap_reset() and the device specific routines
* (e.g. ixgbe_setup_receive_rings()) map these
* buffers at the end of the NIC ring, so here we
* must set the RDT (tail) register to make sure
* they are not overwritten.
*
* In this driver the NIC ring starts at RDH = 0,
* RDT points to the last slot available for reception (?),
* so RDT = num_rx_desc - 1 means the whole ring is available.
*/
if (ifp->if_capenable & IFCAP_NETMAP) {
struct netmap_adapter *na = NA(adapter->ifp);
struct netmap_kring *kring = &na->rx_rings[i];
It is 2014 and we have a new version of netmap. Most relevant features: - netmap emulation on any NIC, even those without native netmap support. On the ixgbe we have measured about 4Mpps/core/queue in this mode, which is still a lot more than with sockets/bpf. - seamless interconnection of VALE switch, NICs and host stack. If you disable accelerations on your NIC (say em0) ifconfig em0 -txcsum -txcsum you can use the VALE switch to connect the NIC and the host stack: vale-ctl -h valeXX:em0 allowing sharing the NIC with other netmap clients. - THE USER API HAS SLIGHTLY CHANGED (head/cur/tail pointers instead of pointers/count as before). This was unavoidable to support, in the future, multiple threads operating on the same rings. Netmap clients require very small source code changes to compile again. On the plus side, the new API should be easier to understand and the internals are a lot simpler. The manual page has been updated extensively to reflect the current features and give some examples. This is the result of work of several people including Giuseppe Lettieri, Vincenzo Maffione, Michio Honda and myself, and has been financially supported by EU projects CHANGE and OPENLAB, from NetApp University Research Fund, NEC, and of course the Universita` di Pisa.
2014-01-06 12:53:15 +00:00
int t = na->num_rx_desc - 1 - nm_kr_rxspace(kring);
netmap-related changes: 1. correct the initialization of RDT when there is an ixgbe_init() while a netmap client is active. This code was previously in ixgbe_initialize_receive_units() but RDT is overwritten shortly afterwards in ixgbe_init_locked() 2. add code (not active yet) to disable CRCSTRIP while in netmap mode. From all evidence i could gather, it seems that when the 82599 has to write a data block that is not a full cache line, it first reads the line (64 bytes) and then writes back the updated version. This hurts reception of min-sized frames, which are only 60 bytes if the CRC is stripped: i could never get above 11Mpps (received from one queue) with CRCSTRIP enabled, whyle 64+4-byte packets reach 14.2 Mpps (the theoretical maximum). Leaving the CRC in gets us 14.88Mpps for 60+4 byte frames, (and penalizes 64+4). The min-size case is important not just because it looks good in benchmarks, but also because this is the size of pure acks. Note we cannot leave CRCSTRIP on by default because it is incompatible with some other features (LRO etc.)
2012-01-19 09:36:19 +00:00
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
IXGBE_WRITE_REG(hw, IXGBE_RDT(i), t);
netmap-related changes: 1. correct the initialization of RDT when there is an ixgbe_init() while a netmap client is active. This code was previously in ixgbe_initialize_receive_units() but RDT is overwritten shortly afterwards in ixgbe_init_locked() 2. add code (not active yet) to disable CRCSTRIP while in netmap mode. From all evidence i could gather, it seems that when the 82599 has to write a data block that is not a full cache line, it first reads the line (64 bytes) and then writes back the updated version. This hurts reception of min-sized frames, which are only 60 bytes if the CRC is stripped: i could never get above 11Mpps (received from one queue) with CRCSTRIP enabled, whyle 64+4-byte packets reach 14.2 Mpps (the theoretical maximum). Leaving the CRC in gets us 14.88Mpps for 60+4 byte frames, (and penalizes 64+4). The min-size case is important not just because it looks good in benchmarks, but also because this is the size of pure acks. Note we cannot leave CRCSTRIP on by default because it is incompatible with some other features (LRO etc.)
2012-01-19 09:36:19 +00:00
} else
#endif /* DEV_NETMAP */
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
IXGBE_WRITE_REG(hw, IXGBE_RDT(i), adapter->num_rx_desc - 1);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
/* Enable Receive engine */
rxctrl = IXGBE_READ_REG(hw, IXGBE_RXCTRL);
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
if (hw->mac.type == ixgbe_mac_82598EB)
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
rxctrl |= IXGBE_RXCTRL_DMBYPS;
rxctrl |= IXGBE_RXCTRL_RXEN;
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
ixgbe_enable_rx_dma(hw, rxctrl);
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
callout_reset(&adapter->timer, hz, ixgbe_local_timer, adapter);
/* Set up MSI/X routing */
A few changes: When not defining header split do not allocate mbufs, this can be a BIG savings in the mbuf memory pool. Also keep seperate dma maps for the header and payload pieces when doing header split. The basis of this code was a patch done a while ago by yongari, thank you :) A number of white space changes. MFC: in a few days
2010-05-19 00:03:48 +00:00
if (ixgbe_enable_msix) {
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
ixgbe_configure_ivars(adapter);
A few changes: When not defining header split do not allocate mbufs, this can be a BIG savings in the mbuf memory pool. Also keep seperate dma maps for the header and payload pieces when doing header split. The basis of this code was a patch done a while ago by yongari, thank you :) A number of white space changes. MFC: in a few days
2010-05-19 00:03:48 +00:00
/* Set up auto-mask */
if (hw->mac.type == ixgbe_mac_82598EB)
IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
else {
IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
}
} else { /* Simple settings for Legacy/MSI */
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
ixgbe_set_ivar(adapter, 0, 0, 0);
ixgbe_set_ivar(adapter, 0, 0, 1);
A few changes: When not defining header split do not allocate mbufs, this can be a BIG savings in the mbuf memory pool. Also keep seperate dma maps for the header and payload pieces when doing header split. The basis of this code was a patch done a while ago by yongari, thank you :) A number of white space changes. MFC: in a few days
2010-05-19 00:03:48 +00:00
IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
}
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
#ifdef IXGBE_FDIR
/* Init Flow director */
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
if (hw->mac.type != ixgbe_mac_82598EB) {
Update to the ixgbe driver: - Add a couple of new devices - Flow control changes in shared and core code - Bug fix to Flow Director for 82598 - Shared code sync to internal with required core change Thanks to those helping in the testing and improvements to this driver! MFC after:5 days
2012-07-05 20:51:44 +00:00
u32 hdrm = 32 << fdir_pballoc;
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
hw->mac.ops.setup_rxpba(hw, 0, hdrm, PBA_STRATEGY_EQUAL);
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
ixgbe_init_fdir_signature_82599(&adapter->hw, fdir_pballoc);
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
}
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
#endif
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
/*
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
** Check on any SFP devices that
** need to be kick-started
*/
Thanks to Michael Tuexen for adding SCTP support for 82599, also for finding a one character bug that kept TSO from working. Sometimes with direct attach cables a failure can occur in init, the old method of calling detach was broken, there is no way to return an error to the system from init, so I have changed it to return failure thru the ioctl. And, have fixed the ALTQ code changes of Max Laier, sorry Max :)
2010-03-30 19:09:18 +00:00
if (hw->phy.type == ixgbe_phy_none) {
int err = hw->phy.ops.identify(hw);
if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
device_printf(dev,
"Unsupported SFP+ module type was detected.\n");
Fixes for panic experienced in test at Intel, when doing bidirectional stress traffic on 82598. Also a couple bug fixes from Michael Tuexen, thank you!! Add a workaround into the header so that 8 REL can use the driver (adds local copy of ALTQ fix). MFC: in a few days
2010-06-03 00:00:45 +00:00
return;
Thanks to Michael Tuexen for adding SCTP support for 82599, also for finding a one character bug that kept TSO from working. Sometimes with direct attach cables a failure can occur in init, the old method of calling detach was broken, there is no way to return an error to the system from init, so I have changed it to return failure thru the ioctl. And, have fixed the ALTQ code changes of Max Laier, sorry Max :)
2010-03-30 19:09:18 +00:00
}
}
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
A few minor fixes: - add a moderation value to the Link vector - allow disabling HW RSC on the 82599 if LRO is not enabled. - correct error in the stats code - change optic type on the 82598 DA device Thanks to Andrew Boyer for the changes.
2010-05-14 22:00:37 +00:00
/* Set moderation on the Link interrupt */
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
IXGBE_WRITE_REG(hw, IXGBE_EITR(adapter->vector), IXGBE_LINK_ITR);
A few minor fixes: - add a moderation value to the Link vector - allow disabling HW RSC on the 82599 if LRO is not enabled. - correct error in the stats code - change optic type on the 82598 DA device Thanks to Andrew Boyer for the changes.
2010-05-14 22:00:37 +00:00
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
/* Configure Energy Efficient Ethernet for supported devices */
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
if (adapter->eee_support)
ixgbe_setup_eee(hw, adapter->eee_enabled);
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
/* Config/Enable Link */
ixgbe_config_link(adapter);
Update to the ixgbe driver: - Add a couple of new devices - Flow control changes in shared and core code - Bug fix to Flow Director for 82598 - Shared code sync to internal with required core change Thanks to those helping in the testing and improvements to this driver! MFC after:5 days
2012-07-05 20:51:44 +00:00
/* Hardware Packet Buffer & Flow Control setup */
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
ixgbe_config_delay_values(adapter);
Update to the ixgbe driver: - Add a couple of new devices - Flow control changes in shared and core code - Bug fix to Flow Director for 82598 - Shared code sync to internal with required core change Thanks to those helping in the testing and improvements to this driver! MFC after:5 days
2012-07-05 20:51:44 +00:00
/* Initialize the FC settings */
ixgbe_start_hw(hw);
- Fix link loss on vlan reconfiguration. - Fix issues with 'vlanhwfilter'. MFC after: 1 week Silence from: jfv 5 weeks
2013-11-05 09:46:01 +00:00
/* Set up VLAN support and filter */
ixgbe_setup_vlan_hw_support(adapter);
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
/* Setup DMA Coalescing */
ixgbe_config_dmac(adapter);
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
/* And now turn on interrupts */
ixgbe_enable_intr(adapter);
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/* Now inform the stack we're ready */
ifp->if_drv_flags |= IFF_DRV_RUNNING;
Fixes for panic experienced in test at Intel, when doing bidirectional stress traffic on 82598. Also a couple bug fixes from Michael Tuexen, thank you!! Add a workaround into the header so that 8 REL can use the driver (adds local copy of ALTQ fix). MFC: in a few days
2010-06-03 00:00:45 +00:00
return;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
static void
ixgbe_init(void *arg)
{
struct adapter *adapter = arg;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
IXGBE_CORE_LOCK(adapter);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
ixgbe_init_locked(adapter);
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
IXGBE_CORE_UNLOCK(adapter);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
return;
}
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
static void
ixgbe_config_delay_values(struct adapter *adapter)
{
struct ixgbe_hw *hw = &adapter->hw;
u32 rxpb, frame, size, tmp;
frame = adapter->max_frame_size;
/* Calculate High Water */
switch (hw->mac.type) {
case ixgbe_mac_X540:
case ixgbe_mac_X550:
case ixgbe_mac_X550EM_x:
tmp = IXGBE_DV_X540(frame, frame);
break;
default:
tmp = IXGBE_DV(frame, frame);
break;
}
size = IXGBE_BT2KB(tmp);
rxpb = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0)) >> 10;
hw->fc.high_water[0] = rxpb - size;
/* Now calculate Low Water */
switch (hw->mac.type) {
case ixgbe_mac_X540:
case ixgbe_mac_X550:
case ixgbe_mac_X550EM_x:
tmp = IXGBE_LOW_DV_X540(frame);
break;
default:
tmp = IXGBE_LOW_DV(frame);
break;
}
hw->fc.low_water[0] = IXGBE_BT2KB(tmp);
hw->fc.requested_mode = adapter->fc;
hw->fc.pause_time = IXGBE_FC_PAUSE;
hw->fc.send_xon = TRUE;
}
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
/*
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
**
** MSIX Interrupt Handlers and Tasklets
**
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
*/
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
static inline void
ixgbe_enable_queue(struct adapter *adapter, u32 vector)
{
struct ixgbe_hw *hw = &adapter->hw;
u64 queue = (u64)(1 << vector);
u32 mask;
if (hw->mac.type == ixgbe_mac_82598EB) {
mask = (IXGBE_EIMS_RTX_QUEUE & queue);
IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
} else {
mask = (queue & 0xFFFFFFFF);
if (mask)
IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
mask = (queue >> 32);
if (mask)
IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
}
}
static inline void
ixgbe_disable_queue(struct adapter *adapter, u32 vector)
{
struct ixgbe_hw *hw = &adapter->hw;
u64 queue = (u64)(1 << vector);
u32 mask;
if (hw->mac.type == ixgbe_mac_82598EB) {
mask = (IXGBE_EIMS_RTX_QUEUE & queue);
IXGBE_WRITE_REG(hw, IXGBE_EIMC, mask);
} else {
mask = (queue & 0xFFFFFFFF);
if (mask)
IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), mask);
mask = (queue >> 32);
if (mask)
IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), mask);
}
}
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
static void
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
ixgbe_handle_que(void *context, int pending)
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
{
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
struct ix_queue *que = context;
struct adapter *adapter = que->adapter;
struct tx_ring *txr = que->txr;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
struct ifnet *ifp = adapter->ifp;
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
bool more;
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
more = ixgbe_rxeof(que);
Remove the tx queue selection based on the cpu whe no flowid is present, this was causing some bad reordering, now just use 0. Also, add a few watchdog bits, and tx handler bits that were corrected in igb.
2010-04-16 16:33:05 +00:00
IXGBE_TX_LOCK(txr);
ixgbe_txeof(txr);
Patch #6 Whitespace cleanup, and removal of some very old defines (at Gleb's request). Also, change the defines around the old transmit code to IXGBE_LEGACY_TX, I do this to make it possible to define this regardless of the OS level (it is not defined by default). There are also a couple changed comments for clarity.
2012-11-30 23:13:56 +00:00
#ifndef IXGBE_LEGACY_TX
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
if (!drbr_empty(ifp, txr->br))
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
ixgbe_mq_start_locked(ifp, txr);
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
#else
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
ixgbe_start_locked(txr, ifp);
#endif
Remove the tx queue selection based on the cpu whe no flowid is present, this was causing some bad reordering, now just use 0. Also, add a few watchdog bits, and tx handler bits that were corrected in igb.
2010-04-16 16:33:05 +00:00
IXGBE_TX_UNLOCK(txr);
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
}
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
/* Reenable this interrupt */
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
if (que->res != NULL)
ixgbe_enable_queue(adapter, que->msix);
else
ixgbe_enable_intr(adapter);
Remove the tx queue selection based on the cpu whe no flowid is present, this was causing some bad reordering, now just use 0. Also, add a few watchdog bits, and tx handler bits that were corrected in igb.
2010-04-16 16:33:05 +00:00
return;
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
}
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/*********************************************************************
*
* Legacy Interrupt Service routine
*
**********************************************************************/
static void
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
ixgbe_legacy_irq(void *arg)
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
{
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
struct ix_queue *que = arg;
struct adapter *adapter = que->adapter;
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
struct ixgbe_hw *hw = &adapter->hw;
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
struct ifnet *ifp = adapter->ifp;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
struct tx_ring *txr = adapter->tx_rings;
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
bool more;
u32 reg_eicr;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
reg_eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
++que->irqs;
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
if (reg_eicr == 0) {
ixgbe_enable_intr(adapter);
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
return;
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
}
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
more = ixgbe_rxeof(que);
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
IXGBE_TX_LOCK(txr);
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
ixgbe_txeof(txr);
#ifdef IXGBE_LEGACY_TX
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
ixgbe_start_locked(txr, ifp);
#else
if (!drbr_empty(ifp, txr->br))
ixgbe_mq_start_locked(ifp, txr);
#endif
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
IXGBE_TX_UNLOCK(txr);
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
/* Check for fan failure */
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
if ((hw->phy.media_type == ixgbe_media_type_copper) &&
(reg_eicr & IXGBE_EICR_GPI_SDP1_BY_MAC(hw))) {
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
device_printf(adapter->dev, "\nCRITICAL: FAN FAILURE!! "
"REPLACE IMMEDIATELY!!\n");
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EICR_GPI_SDP1_BY_MAC(hw));
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
}
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
/* Link status change */
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
if (reg_eicr & IXGBE_EICR_LSC)
taskqueue_enqueue(adapter->tq, &adapter->link_task);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
/* External PHY interrupt */
if (hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T &&
(reg_eicr & IXGBE_EICR_GPI_SDP0_X540))
taskqueue_enqueue(adapter->tq, &adapter->phy_task);
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
if (more)
taskqueue_enqueue(que->tq, &que->que_task);
else
ixgbe_enable_intr(adapter);
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
return;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
/*********************************************************************
*
First off: update the driver README, the old one was horribly crusty, and this still isn't perfect, but its at least a bit more recent. Secondly, a few improvements to the driver from Andrew Boyer, support hint to allow devices to not attach, add VLAN_HWTSO capability so vlans can use TSO, fix in the interrupt handler to make sure the stack TX queue is processed. Oh, and also make sure IPv6 does not cause a re-init in the ioctl routine. Thanks for your efforts Andrew! Thanks to Claudio Jeker for noticing the ixgbe_xmit() routine was not correctly swapping the dma map from the first to the last descriptor in a multi-descriptor transmission, corrected this.
2011-06-02 00:34:57 +00:00
* MSIX Queue Interrupt Service routine
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
*
**********************************************************************/
void
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
ixgbe_msix_que(void *arg)
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
{
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
struct ix_queue *que = arg;
struct adapter *adapter = que->adapter;
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
struct ifnet *ifp = adapter->ifp;
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
struct tx_ring *txr = que->txr;
struct rx_ring *rxr = que->rxr;
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
bool more;
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
u32 newitr = 0;
A number of important fixes: - mbuf reused after an RX_COPY optimized operation can sometimes have a bogus cached address, resulting in TCP hangs. Add critical save points to the cached address. Thanks to Michael and the team at Verisign for finding this problem. - A couple more spots where the rxbuf->flags member should be cleared just to be sure no incorrect RX_COPY state is left around. Thanks to Adrian for tracking these down. - Remove the rearm_queues function from the driver, this was found to be responsible for some out-of-order packets by Verisign, and was always a bandaid, with the other fixes in this delta the bandaid can finally be removed. - In the other/link interrupt handler the entire state of the EICS register was being writen back into EICR (which clears causes and thus re-enables those interrupts), this was wrong, so now mask off the queue portion of the register value, so we only clear the other/link interrupt we intend. Marc from Verisign found this. - Make the SFP+ unsupported option tuneable now, by customer request. - Finally, just a couple of minor DEBUG string fixes. I want to call out and thank all the participants in the 10G community/Intel calls for helping track down these problems and make the driver better for everyone! MFC after: 3 days, these are critical fixes for 9.2!
2013-08-01 20:10:16 +00:00
/* Protect against spurious interrupts */
if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
return;
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
ixgbe_disable_queue(adapter, que->msix);
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
++que->irqs;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
more = ixgbe_rxeof(que);
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
This updates the ixgbe driver to Intel internal version 1.4.7 Shared code changes, core driver fixes, vlan event/filter support Also Kip Macy's fix to allow any number of queues, thanks Kip!
2008-07-30 18:15:18 +00:00
IXGBE_TX_LOCK(txr);
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
ixgbe_txeof(txr);
Patch #6 Whitespace cleanup, and removal of some very old defines (at Gleb's request). Also, change the defines around the old transmit code to IXGBE_LEGACY_TX, I do this to make it possible to define this regardless of the OS level (it is not defined by default). There are also a couple changed comments for clarity.
2012-11-30 23:13:56 +00:00
#ifdef IXGBE_LEGACY_TX
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
if (!IFQ_DRV_IS_EMPTY(ifp->if_snd))
ixgbe_start_locked(txr, ifp);
First off: update the driver README, the old one was horribly crusty, and this still isn't perfect, but its at least a bit more recent. Secondly, a few improvements to the driver from Andrew Boyer, support hint to allow devices to not attach, add VLAN_HWTSO capability so vlans can use TSO, fix in the interrupt handler to make sure the stack TX queue is processed. Oh, and also make sure IPv6 does not cause a re-init in the ioctl routine. Thanks for your efforts Andrew! Thanks to Claudio Jeker for noticing the ixgbe_xmit() routine was not correctly swapping the dma map from the first to the last descriptor in a multi-descriptor transmission, corrected this.
2011-06-02 00:34:57 +00:00
#else
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
if (!drbr_empty(ifp, txr->br))
ixgbe_mq_start_locked(ifp, txr);
First off: update the driver README, the old one was horribly crusty, and this still isn't perfect, but its at least a bit more recent. Secondly, a few improvements to the driver from Andrew Boyer, support hint to allow devices to not attach, add VLAN_HWTSO capability so vlans can use TSO, fix in the interrupt handler to make sure the stack TX queue is processed. Oh, and also make sure IPv6 does not cause a re-init in the ioctl routine. Thanks for your efforts Andrew! Thanks to Claudio Jeker for noticing the ixgbe_xmit() routine was not correctly swapping the dma map from the first to the last descriptor in a multi-descriptor transmission, corrected this.
2011-06-02 00:34:57 +00:00
#endif
This updates the ixgbe driver to Intel internal version 1.4.7 Shared code changes, core driver fixes, vlan event/filter support Also Kip Macy's fix to allow any number of queues, thanks Kip!
2008-07-30 18:15:18 +00:00
IXGBE_TX_UNLOCK(txr);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
/* Do AIM now? */
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
if (ixgbe_enable_aim == FALSE)
goto no_calc;
/*
** Do Adaptive Interrupt Moderation:
** - Write out last calculated setting
** - Calculate based on average size over
** the last interval.
*/
if (que->eitr_setting)
IXGBE_WRITE_REG(&adapter->hw,
IXGBE_EITR(que->msix), que->eitr_setting);
que->eitr_setting = 0;
/* Idle, do nothing */
if ((txr->bytes == 0) && (rxr->bytes == 0))
goto no_calc;
if ((txr->bytes) && (txr->packets))
newitr = txr->bytes/txr->packets;
if ((rxr->bytes) && (rxr->packets))
newitr = max(newitr,
(rxr->bytes / rxr->packets));
newitr += 24; /* account for hardware frame, crc */
/* set an upper boundary */
newitr = min(newitr, 3000);
/* Be nice to the mid range */
if ((newitr > 300) && (newitr < 1200))
newitr = (newitr / 3);
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
else
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
newitr = (newitr / 2);
if (adapter->hw.mac.type == ixgbe_mac_82598EB)
newitr |= newitr << 16;
else
newitr |= IXGBE_EITR_CNT_WDIS;
/* save for next interrupt */
que->eitr_setting = newitr;
/* Reset state */
txr->bytes = 0;
txr->packets = 0;
rxr->bytes = 0;
rxr->packets = 0;
no_calc:
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
if (more)
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
taskqueue_enqueue(que->tq, &que->que_task);
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
else
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
ixgbe_enable_queue(adapter, que->msix);
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
return;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
static void
ixgbe_msix_link(void *arg)
{
struct adapter *adapter = arg;
struct ixgbe_hw *hw = &adapter->hw;
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
u32 reg_eicr, mod_mask;
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
++adapter->link_irq;
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
/* First get the cause */
reg_eicr = IXGBE_READ_REG(hw, IXGBE_EICS);
A number of important fixes: - mbuf reused after an RX_COPY optimized operation can sometimes have a bogus cached address, resulting in TCP hangs. Add critical save points to the cached address. Thanks to Michael and the team at Verisign for finding this problem. - A couple more spots where the rxbuf->flags member should be cleared just to be sure no incorrect RX_COPY state is left around. Thanks to Adrian for tracking these down. - Remove the rearm_queues function from the driver, this was found to be responsible for some out-of-order packets by Verisign, and was always a bandaid, with the other fixes in this delta the bandaid can finally be removed. - In the other/link interrupt handler the entire state of the EICS register was being writen back into EICR (which clears causes and thus re-enables those interrupts), this was wrong, so now mask off the queue portion of the register value, so we only clear the other/link interrupt we intend. Marc from Verisign found this. - Make the SFP+ unsupported option tuneable now, by customer request. - Finally, just a couple of minor DEBUG string fixes. I want to call out and thank all the participants in the 10G community/Intel calls for helping track down these problems and make the driver better for everyone! MFC after: 3 days, these are critical fixes for 9.2!
2013-08-01 20:10:16 +00:00
/* Be sure the queue bits are not cleared */
Correct a fat-finger in the last delta. MFC after: ASAP
2013-08-05 16:16:50 +00:00
reg_eicr &= ~IXGBE_EICR_RTX_QUEUE;
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
/* Clear interrupt with write */
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
IXGBE_WRITE_REG(hw, IXGBE_EICR, reg_eicr);
/* Link status change */
if (reg_eicr & IXGBE_EICR_LSC)
taskqueue_enqueue(adapter->tq, &adapter->link_task);
- Add the RX refresh changes from igb to ixgbe - Also a couple minor tweaks to the TX code from the same source. - Add the INET ioctl code which has been missing from this driver, and which caused IP aliases to reset the interface. - Last, some minor logic changes that just reflect upcoming hardware support, but have no other functional effect now. MFC after a week
2011-04-25 23:34:21 +00:00
if (adapter->hw.mac.type != ixgbe_mac_82598EB) {
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
#ifdef IXGBE_FDIR
if (reg_eicr & IXGBE_EICR_FLOW_DIR) {
/* This is probably overkill :) */
if (!atomic_cmpset_int(&adapter->fdir_reinit, 0, 1))
return;
Update to the ixgbe driver: - Add a couple of new devices - Flow control changes in shared and core code - Bug fix to Flow Director for 82598 - Shared code sync to internal with required core change Thanks to those helping in the testing and improvements to this driver! MFC after:5 days
2012-07-05 20:51:44 +00:00
/* Disable the interrupt */
IXGBE_WRITE_REG(hw, IXGBE_EIMC, IXGBE_EICR_FLOW_DIR);
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
taskqueue_enqueue(adapter->tq, &adapter->fdir_task);
} else
#endif
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
if (reg_eicr & IXGBE_EICR_ECC) {
device_printf(adapter->dev, "\nCRITICAL: ECC ERROR!! "
"Please Reboot!!\n");
IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_ECC);
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
}
/* Check for over temp condition */
if (reg_eicr & IXGBE_EICR_TS) {
device_printf(adapter->dev, "\nCRITICAL: OVER TEMP!! "
"PHY IS SHUT DOWN!!\n");
device_printf(adapter->dev, "System shutdown required!\n");
IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_TS);
}
}
/* Pluggable optics-related interrupt */
if (hw->device_id == IXGBE_DEV_ID_X550EM_X_SFP)
mod_mask = IXGBE_EICR_GPI_SDP0_X540;
else
mod_mask = IXGBE_EICR_GPI_SDP2_BY_MAC(hw);
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
if (ixgbe_is_sfp(hw)) {
if (reg_eicr & IXGBE_EICR_GPI_SDP1_BY_MAC(hw)) {
IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1_BY_MAC(hw));
taskqueue_enqueue(adapter->tq, &adapter->msf_task);
} else if (reg_eicr & mod_mask) {
IXGBE_WRITE_REG(hw, IXGBE_EICR, mod_mask);
taskqueue_enqueue(adapter->tq, &adapter->mod_task);
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
}
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
}
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
/* Check for fan failure */
if ((hw->device_id == IXGBE_DEV_ID_82598AT) &&
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
(reg_eicr & IXGBE_EICR_GPI_SDP1)) {
IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP1);
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
device_printf(adapter->dev, "\nCRITICAL: FAN FAILURE!! "
"REPLACE IMMEDIATELY!!\n");
}
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
/* External PHY interrupt */
if (hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T &&
(reg_eicr & IXGBE_EICR_GPI_SDP0_X540)) {
IXGBE_WRITE_REG(hw, IXGBE_EICR, IXGBE_EICR_GPI_SDP0_X540);
taskqueue_enqueue(adapter->tq, &adapter->phy_task);
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
}
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, IXGBE_EIMS_OTHER);
return;
}
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/*********************************************************************
*
* Media Ioctl callback
*
* This routine is called whenever the user queries the status of
* the interface using ifconfig.
*
**********************************************************************/
static void
ixgbe_media_status(struct ifnet * ifp, struct ifmediareq * ifmr)
{
struct adapter *adapter = ifp->if_softc;
Fix style(9) from my previous commit r279803. Spotted by: kevlo
2015-03-09 10:29:15 +00:00
struct ixgbe_hw *hw = &adapter->hw;
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
int layer;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
INIT_DEBUGOUT("ixgbe_media_status: begin");
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
IXGBE_CORE_LOCK(adapter);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
ixgbe_update_link_status(adapter);
ifmr->ifm_status = IFM_AVALID;
ifmr->ifm_active = IFM_ETHER;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
if (!adapter->link_active) {
IXGBE_CORE_UNLOCK(adapter);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
return;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
}
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
ifmr->ifm_status |= IFM_ACTIVE;
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
layer = ixgbe_get_supported_physical_layer(hw);
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_T ||
layer & IXGBE_PHYSICAL_LAYER_1000BASE_T ||
layer & IXGBE_PHYSICAL_LAYER_100BASE_TX)
switch (adapter->link_speed) {
case IXGBE_LINK_SPEED_10GB_FULL:
ifmr->ifm_active |= IFM_10G_T | IFM_FDX;
break;
case IXGBE_LINK_SPEED_1GB_FULL:
Fix style(9) from my previous commit r279803. Spotted by: kevlo
2015-03-09 10:29:15 +00:00
ifmr->ifm_active |= IFM_1000_T | IFM_FDX;
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
break;
case IXGBE_LINK_SPEED_100_FULL:
Fix style(9) from my previous commit r279803. Spotted by: kevlo
2015-03-09 10:29:15 +00:00
ifmr->ifm_active |= IFM_100_TX | IFM_FDX;
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
break;
}
if (layer & IXGBE_PHYSICAL_LAYER_SFP_PLUS_CU ||
layer & IXGBE_PHYSICAL_LAYER_SFP_ACTIVE_DA)
switch (adapter->link_speed) {
case IXGBE_LINK_SPEED_10GB_FULL:
ifmr->ifm_active |= IFM_10G_TWINAX | IFM_FDX;
break;
}
if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_LR)
switch (adapter->link_speed) {
case IXGBE_LINK_SPEED_10GB_FULL:
ifmr->ifm_active |= IFM_10G_LR | IFM_FDX;
break;
case IXGBE_LINK_SPEED_1GB_FULL:
ifmr->ifm_active |= IFM_1000_LX | IFM_FDX;
break;
}
if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_LRM)
switch (adapter->link_speed) {
case IXGBE_LINK_SPEED_10GB_FULL:
ifmr->ifm_active |= IFM_10G_LRM | IFM_FDX;
break;
case IXGBE_LINK_SPEED_1GB_FULL:
ifmr->ifm_active |= IFM_1000_LX | IFM_FDX;
break;
}
if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_SR ||
layer & IXGBE_PHYSICAL_LAYER_1000BASE_SX)
switch (adapter->link_speed) {
case IXGBE_LINK_SPEED_10GB_FULL:
ifmr->ifm_active |= IFM_10G_SR | IFM_FDX;
break;
case IXGBE_LINK_SPEED_1GB_FULL:
Fix style(9) from my previous commit r279803. Spotted by: kevlo
2015-03-09 10:29:15 +00:00
ifmr->ifm_active |= IFM_1000_SX | IFM_FDX;
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
break;
}
if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_CX4)
switch (adapter->link_speed) {
case IXGBE_LINK_SPEED_10GB_FULL:
ifmr->ifm_active |= IFM_10G_CX4 | IFM_FDX;
break;
}
/*
** XXX: These need to use the proper media types once
** they're added.
*/
if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KR)
switch (adapter->link_speed) {
case IXGBE_LINK_SPEED_10GB_FULL:
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
ifmr->ifm_active |= IFM_10G_SR | IFM_FDX;
break;
case IXGBE_LINK_SPEED_2_5GB_FULL:
ifmr->ifm_active |= IFM_2500_SX | IFM_FDX;
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
break;
case IXGBE_LINK_SPEED_1GB_FULL:
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
ifmr->ifm_active |= IFM_1000_CX | IFM_FDX;
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
break;
}
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
else if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KX4
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
|| layer & IXGBE_PHYSICAL_LAYER_1000BASE_KX)
switch (adapter->link_speed) {
case IXGBE_LINK_SPEED_10GB_FULL:
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
ifmr->ifm_active |= IFM_10G_CX4 | IFM_FDX;
break;
case IXGBE_LINK_SPEED_2_5GB_FULL:
ifmr->ifm_active |= IFM_2500_SX | IFM_FDX;
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
break;
case IXGBE_LINK_SPEED_1GB_FULL:
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
ifmr->ifm_active |= IFM_1000_CX | IFM_FDX;
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
break;
}
/* If nothing is recognized... */
if (IFM_SUBTYPE(ifmr->ifm_active) == 0)
ifmr->ifm_active |= IFM_UNKNOWN;
#if __FreeBSD_version >= 900025
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
/* Display current flow control setting used on link */
if (hw->fc.current_mode == ixgbe_fc_rx_pause ||
hw->fc.current_mode == ixgbe_fc_full)
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
ifmr->ifm_active |= IFM_ETH_RXPAUSE;
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
if (hw->fc.current_mode == ixgbe_fc_tx_pause ||
hw->fc.current_mode == ixgbe_fc_full)
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
ifmr->ifm_active |= IFM_ETH_TXPAUSE;
#endif
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
IXGBE_CORE_UNLOCK(adapter);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
return;
}
/*********************************************************************
*
* Media Ioctl callback
*
* This routine is called when the user changes speed/duplex using
* media/mediopt option with ifconfig.
*
**********************************************************************/
static int
ixgbe_media_change(struct ifnet * ifp)
{
struct adapter *adapter = ifp->if_softc;
struct ifmedia *ifm = &adapter->media;
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
struct ixgbe_hw *hw = &adapter->hw;
ixgbe_link_speed speed = 0;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
INIT_DEBUGOUT("ixgbe_media_change: begin");
if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
return (EINVAL);
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
if (hw->phy.media_type == ixgbe_media_type_backplane)
return (EPERM);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/*
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
** We don't actually need to check against the supported
** media types of the adapter; ifmedia will take care of
** that for us.
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
*/
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
switch (IFM_SUBTYPE(ifm->ifm_media)) {
case IFM_AUTO:
case IFM_10G_T:
speed |= IXGBE_LINK_SPEED_100_FULL;
case IFM_10G_LRM:
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
case IFM_10G_SR: /* KR, too */
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
case IFM_10G_LR:
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
case IFM_10G_CX4: /* KX4 */
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
speed |= IXGBE_LINK_SPEED_1GB_FULL;
case IFM_10G_TWINAX:
speed |= IXGBE_LINK_SPEED_10GB_FULL;
break;
case IFM_1000_T:
speed |= IXGBE_LINK_SPEED_100_FULL;
case IFM_1000_LX:
case IFM_1000_SX:
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
case IFM_1000_CX: /* KX */
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
speed |= IXGBE_LINK_SPEED_1GB_FULL;
break;
case IFM_100_TX:
speed |= IXGBE_LINK_SPEED_100_FULL;
break;
default:
goto invalid;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
hw->mac.autotry_restart = TRUE;
hw->mac.ops.setup_link(hw, speed, TRUE);
adapter->advertise =
((speed & IXGBE_LINK_SPEED_10GB_FULL) << 2) |
((speed & IXGBE_LINK_SPEED_1GB_FULL) << 1) |
((speed & IXGBE_LINK_SPEED_100_FULL) << 0);
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
return (0);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
invalid:
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
device_printf(adapter->dev, "Invalid media type!\n");
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
return (EINVAL);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
static void
ixgbe_set_promisc(struct adapter *adapter)
{
u_int32_t reg_rctl;
struct ifnet *ifp = adapter->ifp;
Two small fixes: Set promiscuous code was unconditionally turning off multicast when turning off promiscuous mode, this should only be done when there are less than MAX groups. Thanks to Mike Karels for this correction. Second, the overtmp interrupt setup/detection was wrong, correcting it. MFC after: one week
2013-03-29 18:03:00 +00:00
int mcnt = 0;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
reg_rctl = IXGBE_READ_REG(&adapter->hw, IXGBE_FCTRL);
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
reg_rctl &= (~IXGBE_FCTRL_UPE);
Two small fixes: Set promiscuous code was unconditionally turning off multicast when turning off promiscuous mode, this should only be done when there are less than MAX groups. Thanks to Mike Karels for this correction. Second, the overtmp interrupt setup/detection was wrong, correcting it. MFC after: one week
2013-03-29 18:03:00 +00:00
if (ifp->if_flags & IFF_ALLMULTI)
mcnt = MAX_NUM_MULTICAST_ADDRESSES;
else {
struct ifmultiaddr *ifma;
#if __FreeBSD_version < 800000
IF_ADDR_LOCK(ifp);
#else
if_maddr_rlock(ifp);
#endif
TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
if (ifma->ifma_addr->sa_family != AF_LINK)
continue;
if (mcnt == MAX_NUM_MULTICAST_ADDRESSES)
break;
mcnt++;
}
#if __FreeBSD_version < 800000
IF_ADDR_UNLOCK(ifp);
#else
if_maddr_runlock(ifp);
#endif
}
if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
reg_rctl &= (~IXGBE_FCTRL_MPE);
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
IXGBE_WRITE_REG(&adapter->hw, IXGBE_FCTRL, reg_rctl);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
if (ifp->if_flags & IFF_PROMISC) {
reg_rctl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
IXGBE_WRITE_REG(&adapter->hw, IXGBE_FCTRL, reg_rctl);
} else if (ifp->if_flags & IFF_ALLMULTI) {
reg_rctl |= IXGBE_FCTRL_MPE;
reg_rctl &= ~IXGBE_FCTRL_UPE;
IXGBE_WRITE_REG(&adapter->hw, IXGBE_FCTRL, reg_rctl);
}
return;
}
/*********************************************************************
* Multicast Update
*
* This routine is called whenever multicast address list is updated.
*
**********************************************************************/
#define IXGBE_RAR_ENTRIES 16
static void
ixgbe_set_multi(struct adapter *adapter)
{
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
u32 fctrl;
u8 *mta;
u8 *update_ptr;
struct ifmultiaddr *ifma;
int mcnt = 0;
struct ifnet *ifp = adapter->ifp;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
IOCTL_DEBUGOUT("ixgbe_set_multi: begin");
A couple fixes got clobbered, putting them back.
2010-11-26 23:57:13 +00:00
mta = adapter->mta;
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
bzero(mta, sizeof(u8) * IXGBE_ETH_LENGTH_OF_ADDRESS *
MAX_NUM_MULTICAST_ADDRESSES);
A couple fixes got clobbered, putting them back.
2010-11-26 23:57:13 +00:00
A couple of changes: first, make header split on by default and remove from its dependency on LRO, my tests have shown that its always beneficial, even when doing bridging. Second, fix up a few problems in the statistics code, the adapter dependencies had gotten lost so some code that should only run on 82599 was always running, this resulted in bogus flow control numbers on 82598.
2009-08-24 20:41:51 +00:00
#if __FreeBSD_version < 800000
IF_ADDR_LOCK(ifp);
#else
Use if_maddr_rlock()/if_maddr_runlock() rather than IF_ADDR_LOCK()/ IF_ADDR_UNLOCK() across network device drivers when accessing the per-interface multicast address list, if_multiaddrs. This will allow us to change the locking strategy without affecting our driver programming interface or binary interface. For two wireless drivers, remove unnecessary locking, since they don't actually access the multicast address list. Approved by: re (kib) MFC after: 6 weeks
2009-06-26 11:45:06 +00:00
if_maddr_rlock(ifp);
A couple of changes: first, make header split on by default and remove from its dependency on LRO, my tests have shown that its always beneficial, even when doing bridging. Second, fix up a few problems in the statistics code, the adapter dependencies had gotten lost so some code that should only run on 82599 was always running, this resulted in bogus flow control numbers on 82598.
2009-08-24 20:41:51 +00:00
#endif
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
if (ifma->ifma_addr->sa_family != AF_LINK)
continue;
First, sync to internal shared code, and then Fixes: - flow control - don't override user value on re-init - fix to make 1G optics work correctly - change to interrupt enabling - some bits were incorrect for certain hardware. - certain stats fixes, remove a duplicate increment of ierror, thanks to Scott Long for pointing these out. - shared code link interface changed, requiring some core code changes to accomodate this. - add an m_adj() to ETHER_ALIGN on the recieve side, this was requested by Mike Karels, thanks Mike. - Multicast code corrections also thanks to Mike Karels.
2013-03-04 23:07:40 +00:00
if (mcnt == MAX_NUM_MULTICAST_ADDRESSES)
break;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
bcopy(LLADDR((struct sockaddr_dl *) ifma->ifma_addr),
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
&mta[mcnt * IXGBE_ETH_LENGTH_OF_ADDRESS],
IXGBE_ETH_LENGTH_OF_ADDRESS);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
mcnt++;
}
A couple of changes: first, make header split on by default and remove from its dependency on LRO, my tests have shown that its always beneficial, even when doing bridging. Second, fix up a few problems in the statistics code, the adapter dependencies had gotten lost so some code that should only run on 82599 was always running, this resulted in bogus flow control numbers on 82598.
2009-08-24 20:41:51 +00:00
#if __FreeBSD_version < 800000
IF_ADDR_UNLOCK(ifp);
#else
Use if_maddr_rlock()/if_maddr_runlock() rather than IF_ADDR_LOCK()/ IF_ADDR_UNLOCK() across network device drivers when accessing the per-interface multicast address list, if_multiaddrs. This will allow us to change the locking strategy without affecting our driver programming interface or binary interface. For two wireless drivers, remove unnecessary locking, since they don't actually access the multicast address list. Approved by: re (kib) MFC after: 6 weeks
2009-06-26 11:45:06 +00:00
if_maddr_runlock(ifp);
A couple of changes: first, make header split on by default and remove from its dependency on LRO, my tests have shown that its always beneficial, even when doing bridging. Second, fix up a few problems in the statistics code, the adapter dependencies had gotten lost so some code that should only run on 82599 was always running, this resulted in bogus flow control numbers on 82598.
2009-08-24 20:41:51 +00:00
#endif
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
First, sync to internal shared code, and then Fixes: - flow control - don't override user value on re-init - fix to make 1G optics work correctly - change to interrupt enabling - some bits were incorrect for certain hardware. - certain stats fixes, remove a duplicate increment of ierror, thanks to Scott Long for pointing these out. - shared code link interface changed, requiring some core code changes to accomodate this. - add an m_adj() to ETHER_ALIGN on the recieve side, this was requested by Mike Karels, thanks Mike. - Multicast code corrections also thanks to Mike Karels.
2013-03-04 23:07:40 +00:00
fctrl = IXGBE_READ_REG(&adapter->hw, IXGBE_FCTRL);
fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
if (ifp->if_flags & IFF_PROMISC)
fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
else if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES ||
ifp->if_flags & IFF_ALLMULTI) {
fctrl |= IXGBE_FCTRL_MPE;
fctrl &= ~IXGBE_FCTRL_UPE;
} else
fctrl &= ~(IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
IXGBE_WRITE_REG(&adapter->hw, IXGBE_FCTRL, fctrl);
if (mcnt < MAX_NUM_MULTICAST_ADDRESSES) {
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
update_ptr = mta;
First, sync to internal shared code, and then Fixes: - flow control - don't override user value on re-init - fix to make 1G optics work correctly - change to interrupt enabling - some bits were incorrect for certain hardware. - certain stats fixes, remove a duplicate increment of ierror, thanks to Scott Long for pointing these out. - shared code link interface changed, requiring some core code changes to accomodate this. - add an m_adj() to ETHER_ALIGN on the recieve side, this was requested by Mike Karels, thanks Mike. - Multicast code corrections also thanks to Mike Karels.
2013-03-04 23:07:40 +00:00
ixgbe_update_mc_addr_list(&adapter->hw,
update_ptr, mcnt, ixgbe_mc_array_itr, TRUE);
}
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
return;
}
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
/*
* This is an iterator function now needed by the multicast
* shared code. It simply feeds the shared code routine the
* addresses in the array of ixgbe_set_multi() one by one.
*/
static u8 *
ixgbe_mc_array_itr(struct ixgbe_hw *hw, u8 **update_ptr, u32 *vmdq)
{
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
u8 *addr = *update_ptr;
u8 *newptr;
*vmdq = 0;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
newptr = addr + IXGBE_ETH_LENGTH_OF_ADDRESS;
*update_ptr = newptr;
return addr;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
}
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/*********************************************************************
* Timer routine
*
* This routine checks for link status,updates statistics,
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
* and runs the watchdog check.
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
*
**********************************************************************/
static void
ixgbe_local_timer(void *arg)
{
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
struct adapter *adapter = arg;
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
device_t dev = adapter->dev;
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
struct ix_queue *que = adapter->queues;
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
u64 queues = 0;
int hung = 0;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
mtx_assert(&adapter->core_mtx, MA_OWNED);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
/* Check for pluggable optics */
if (adapter->sfp_probe)
if (!ixgbe_sfp_probe(adapter))
goto out; /* Nothing to do */
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
ixgbe_update_link_status(adapter);
ixgbe_update_stats_counters(adapter);
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/*
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
** Check the TX queues status
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
** - mark hung queues so we don't schedule on them
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
** - watchdog only if all queues show hung
*/
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
for (int i = 0; i < adapter->num_queues; i++, que++) {
/* Keep track of queues with work for soft irq */
if (que->txr->busy)
queues |= ((u64)1 << que->me);
/*
** Each time txeof runs without cleaning, but there
** are uncleaned descriptors it increments busy. If
** we get to the MAX we declare it hung.
*/
if (que->busy == IXGBE_QUEUE_HUNG) {
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
++hung;
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
/* Mark the queue as inactive */
adapter->active_queues &= ~((u64)1 << que->me);
continue;
} else {
/* Check if we've come back from hung */
if ((adapter->active_queues & ((u64)1 << que->me)) == 0)
adapter->active_queues |= ((u64)1 << que->me);
}
if (que->busy >= IXGBE_MAX_TX_BUSY) {
device_printf(dev,"Warning queue %d "
"appears to be hung!\n", i);
que->txr->busy = IXGBE_QUEUE_HUNG;
++hung;
}
}
/* Only truly watchdog if all queues show hung */
if (hung == adapter->num_queues)
goto watchdog;
else if (queues != 0) { /* Force an IRQ on queues with work */
ixgbe_rearm_queues(adapter, queues);
}
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
out:
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
callout_reset(&adapter->timer, hz, ixgbe_local_timer, adapter);
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
return;
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
watchdog:
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
device_printf(adapter->dev, "Watchdog timeout -- resetting\n");
adapter->ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
adapter->watchdog_events++;
ixgbe_init_locked(adapter);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
/*
** Note: this routine updates the OS on the link state
** the real check of the hardware only happens with
** a link interrupt.
*/
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
static void
ixgbe_update_link_status(struct adapter *adapter)
{
struct ifnet *ifp = adapter->ifp;
device_t dev = adapter->dev;
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
if (adapter->link_up){
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
if (adapter->link_active == FALSE) {
if (bootverbose)
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
device_printf(dev,"Link is up %d Gbps %s \n",
((adapter->link_speed == 128)? 10:1),
"Full Duplex");
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
adapter->link_active = TRUE;
Update to the ixgbe driver: - Add a couple of new devices - Flow control changes in shared and core code - Bug fix to Flow Director for 82598 - Shared code sync to internal with required core change Thanks to those helping in the testing and improvements to this driver! MFC after:5 days
2012-07-05 20:51:44 +00:00
/* Update any Flow Control changes */
ixgbe_fc_enable(&adapter->hw);
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
/* Update DMA coalescing config */
ixgbe_config_dmac(adapter);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
if_link_state_change(ifp, LINK_STATE_UP);
}
} else { /* Link down */
if (adapter->link_active == TRUE) {
if (bootverbose)
device_printf(dev,"Link is Down\n");
if_link_state_change(ifp, LINK_STATE_DOWN);
adapter->link_active = FALSE;
}
}
return;
}
/*********************************************************************
*
* This routine disables all traffic on the adapter by issuing a
* global reset on the MAC and deallocates TX/RX buffers.
*
**********************************************************************/
static void
ixgbe_stop(void *arg)
{
struct ifnet *ifp;
struct adapter *adapter = arg;
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
struct ixgbe_hw *hw = &adapter->hw;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
ifp = adapter->ifp;
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
mtx_assert(&adapter->core_mtx, MA_OWNED);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
INIT_DEBUGOUT("ixgbe_stop: begin\n");
ixgbe_disable_intr(adapter);
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
callout_stop(&adapter->timer);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
/* Let the stack know...*/
ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
ixgbe_reset_hw(hw);
hw->adapter_stopped = FALSE;
ixgbe_stop_adapter(hw);
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
if (hw->mac.type == ixgbe_mac_82599EB)
ixgbe_stop_mac_link_on_d3_82599(hw);
/* Turn off the laser - noop with no optics */
ixgbe_disable_tx_laser(hw);
/* Update the stack */
adapter->link_up = FALSE;
ixgbe_update_link_status(adapter);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
/* reprogram the RAR[0] in case user changed it. */
ixgbe_set_rar(&adapter->hw, 0, adapter->hw.mac.addr, 0, IXGBE_RAH_AV);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
return;
}
/*********************************************************************
*
* Determine hardware revision.
*
**********************************************************************/
static void
ixgbe_identify_hardware(struct adapter *adapter)
{
device_t dev = adapter->dev;
Update driver to version 2.3.8: CRITICAL FIX - with stats changes the older 82598 will panic and trash the stack on driver load, FCOE registers ONLY exist in 82599 and must not be read otherwise. kern/153951 - to correct incorrect media type on adapters with pluggable modules I have eliminated the old static table in favor of a new dynamic shared code routine. This also has the benefit of detecting changes when a different module is inserted. Performance/enhancement to the Flow Director code from my linux coworker (the developer of the code). Fixes from Michael Tuexen - a data corruption problem on the 82599 (CRITICAL), fix so the buf size correctly adjusts as the cluster changes, and max descriptors are set properly. Also added 16K clusters for those REALLY big jumbos :) In the RX path, the RX LOCK was not being released, and this causes LOR problems. Add the code that igb already has. Sync with in house shared code, this was necessary for the Flow Director fix. MFC in 2 days
2011-01-19 19:36:27 +00:00
struct ixgbe_hw *hw = &adapter->hw;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/* Save off the information about this board */
Update driver to version 2.3.8: CRITICAL FIX - with stats changes the older 82598 will panic and trash the stack on driver load, FCOE registers ONLY exist in 82599 and must not be read otherwise. kern/153951 - to correct incorrect media type on adapters with pluggable modules I have eliminated the old static table in favor of a new dynamic shared code routine. This also has the benefit of detecting changes when a different module is inserted. Performance/enhancement to the Flow Director code from my linux coworker (the developer of the code). Fixes from Michael Tuexen - a data corruption problem on the 82599 (CRITICAL), fix so the buf size correctly adjusts as the cluster changes, and max descriptors are set properly. Also added 16K clusters for those REALLY big jumbos :) In the RX path, the RX LOCK was not being released, and this causes LOR problems. Add the code that igb already has. Sync with in house shared code, this was necessary for the Flow Director fix. MFC in 2 days
2011-01-19 19:36:27 +00:00
hw->vendor_id = pci_get_vendor(dev);
hw->device_id = pci_get_device(dev);
hw->revision_id = pci_read_config(dev, PCIR_REVID, 1);
hw->subsystem_vendor_id =
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
pci_read_config(dev, PCIR_SUBVEND_0, 2);
Update driver to version 2.3.8: CRITICAL FIX - with stats changes the older 82598 will panic and trash the stack on driver load, FCOE registers ONLY exist in 82599 and must not be read otherwise. kern/153951 - to correct incorrect media type on adapters with pluggable modules I have eliminated the old static table in favor of a new dynamic shared code routine. This also has the benefit of detecting changes when a different module is inserted. Performance/enhancement to the Flow Director code from my linux coworker (the developer of the code). Fixes from Michael Tuexen - a data corruption problem on the 82599 (CRITICAL), fix so the buf size correctly adjusts as the cluster changes, and max descriptors are set properly. Also added 16K clusters for those REALLY big jumbos :) In the RX path, the RX LOCK was not being released, and this causes LOR problems. Add the code that igb already has. Sync with in house shared code, this was necessary for the Flow Director fix. MFC in 2 days
2011-01-19 19:36:27 +00:00
hw->subsystem_device_id =
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
pci_read_config(dev, PCIR_SUBDEV_0, 2);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
/*
** Make sure BUSMASTER is set
*/
pci_enable_busmaster(dev);
Update driver to version 2.3.8: CRITICAL FIX - with stats changes the older 82598 will panic and trash the stack on driver load, FCOE registers ONLY exist in 82599 and must not be read otherwise. kern/153951 - to correct incorrect media type on adapters with pluggable modules I have eliminated the old static table in favor of a new dynamic shared code routine. This also has the benefit of detecting changes when a different module is inserted. Performance/enhancement to the Flow Director code from my linux coworker (the developer of the code). Fixes from Michael Tuexen - a data corruption problem on the 82599 (CRITICAL), fix so the buf size correctly adjusts as the cluster changes, and max descriptors are set properly. Also added 16K clusters for those REALLY big jumbos :) In the RX path, the RX LOCK was not being released, and this causes LOR problems. Add the code that igb already has. Sync with in house shared code, this was necessary for the Flow Director fix. MFC in 2 days
2011-01-19 19:36:27 +00:00
/* We need this here to set the num_segs below */
ixgbe_set_mac_type(hw);
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
/* Pick up the 82599 settings */
Update driver to version 2.3.8: CRITICAL FIX - with stats changes the older 82598 will panic and trash the stack on driver load, FCOE registers ONLY exist in 82599 and must not be read otherwise. kern/153951 - to correct incorrect media type on adapters with pluggable modules I have eliminated the old static table in favor of a new dynamic shared code routine. This also has the benefit of detecting changes when a different module is inserted. Performance/enhancement to the Flow Director code from my linux coworker (the developer of the code). Fixes from Michael Tuexen - a data corruption problem on the 82599 (CRITICAL), fix so the buf size correctly adjusts as the cluster changes, and max descriptors are set properly. Also added 16K clusters for those REALLY big jumbos :) In the RX path, the RX LOCK was not being released, and this causes LOR problems. Add the code that igb already has. Sync with in house shared code, this was necessary for the Flow Director fix. MFC in 2 days
2011-01-19 19:36:27 +00:00
if (hw->mac.type != ixgbe_mac_82598EB) {
hw->phy.smart_speed = ixgbe_smart_speed;
adapter->num_segs = IXGBE_82599_SCATTER;
} else
adapter->num_segs = IXGBE_82598_SCATTER;
return;
}
/*********************************************************************
*
* Determine optic type
*
**********************************************************************/
static void
ixgbe_setup_optics(struct adapter *adapter)
{
struct ixgbe_hw *hw = &adapter->hw;
int layer;
First, sync to internal shared code, and then Fixes: - flow control - don't override user value on re-init - fix to make 1G optics work correctly - change to interrupt enabling - some bits were incorrect for certain hardware. - certain stats fixes, remove a duplicate increment of ierror, thanks to Scott Long for pointing these out. - shared code link interface changed, requiring some core code changes to accomodate this. - add an m_adj() to ETHER_ALIGN on the recieve side, this was requested by Mike Karels, thanks Mike. - Multicast code corrections also thanks to Mike Karels.
2013-03-04 23:07:40 +00:00
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
layer = ixgbe_get_supported_physical_layer(hw);
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_T) {
adapter->optics = IFM_10G_T;
return;
}
if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_T) {
adapter->optics = IFM_1000_T;
return;
Update driver to version 2.3.8: CRITICAL FIX - with stats changes the older 82598 will panic and trash the stack on driver load, FCOE registers ONLY exist in 82599 and must not be read otherwise. kern/153951 - to correct incorrect media type on adapters with pluggable modules I have eliminated the old static table in favor of a new dynamic shared code routine. This also has the benefit of detecting changes when a different module is inserted. Performance/enhancement to the Flow Director code from my linux coworker (the developer of the code). Fixes from Michael Tuexen - a data corruption problem on the 82599 (CRITICAL), fix so the buf size correctly adjusts as the cluster changes, and max descriptors are set properly. Also added 16K clusters for those REALLY big jumbos :) In the RX path, the RX LOCK was not being released, and this causes LOR problems. Add the code that igb already has. Sync with in house shared code, this was necessary for the Flow Director fix. MFC in 2 days
2011-01-19 19:36:27 +00:00
}
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
A few important fixes: - Testing TSO6 has led me to discover that HW RSC is a problematic feature, it is ONLY designed to work with IPv4 in the first place, and if IP forwarding is done it can't be disabled as LRO in the stack, also initial testing we've done at Intel shows an equal performance using TSO[46] on the TX and LRO on RX, if you ran older code on 82599 or later hardware you actually could have detrimental performance for this reason. So I am disabling the feature by default and all our adapters will now use LRO instead. - If you have flow control off and multiple queues it was possible when the buffer of one queue becomes full that all RX movement is stalled, to eliminate this problem a feature bit is now set that will allow packets to be dropped when full rather than stall. Note, the default is to have flow control on, and this keeps this from happening. - Because of the recent fixes in the stack, LRO is now auto-disabled when problematic, so I have decided to enable it by default in the capabilities in the driver. - There are some 1G modules used by some customers, a couple small tweaks to properly support those in the media code. - A note: we have now done some testing of TSO6 and using LRO with IPv6 and it all works great!! Seeing line rate in both directions in best cases. Thanks bz for your excellent work!!
2012-10-31 23:50:36 +00:00
if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_SX) {
adapter->optics = IFM_1000_SX;
return;
}
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
if (layer & (IXGBE_PHYSICAL_LAYER_10GBASE_LR |
IXGBE_PHYSICAL_LAYER_10GBASE_LRM)) {
adapter->optics = IFM_10G_LR;
return;
}
if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_SR) {
adapter->optics = IFM_10G_SR;
return;
}
if (layer & IXGBE_PHYSICAL_LAYER_SFP_PLUS_CU) {
adapter->optics = IFM_10G_TWINAX;
return;
}
if (layer & (IXGBE_PHYSICAL_LAYER_10GBASE_KX4 |
IXGBE_PHYSICAL_LAYER_10GBASE_CX4)) {
adapter->optics = IFM_10G_CX4;
return;
}
/* If we get here just set the default */
adapter->optics = IFM_ETHER | IFM_AUTO;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
return;
}
/*********************************************************************
*
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
* Setup the Legacy or MSI Interrupt handler
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
*
**********************************************************************/
static int
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
ixgbe_allocate_legacy(struct adapter *adapter)
{
Patch #6 Whitespace cleanup, and removal of some very old defines (at Gleb's request). Also, change the defines around the old transmit code to IXGBE_LEGACY_TX, I do this to make it possible to define this regardless of the OS level (it is not defined by default). There are also a couple changed comments for clarity.
2012-11-30 23:13:56 +00:00
device_t dev = adapter->dev;
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
struct ix_queue *que = adapter->queues;
Patch #6 Whitespace cleanup, and removal of some very old defines (at Gleb's request). Also, change the defines around the old transmit code to IXGBE_LEGACY_TX, I do this to make it possible to define this regardless of the OS level (it is not defined by default). There are also a couple changed comments for clarity.
2012-11-30 23:13:56 +00:00
#ifndef IXGBE_LEGACY_TX
Merge similar fixes from 223198 from igb to ixgbe: - Use a dedicated task to handle deferred transmits from the if_transmit method instead of reusing the existing per-queue interrupt task. Reusing the per-queue interrupt task could result in both an interrupt thread and the taskqueue thread trying to handle received packets on a single queue resulting in out-of-order packet processing and lock contention. - Don't define ixgbe_start() at all where if_transmit is used. Tested by: Vijay Singh Reviewed by: jfv MFC after: 2 weeks
2012-09-26 18:11:43 +00:00
struct tx_ring *txr = adapter->tx_rings;
#endif
Patch #6 Whitespace cleanup, and removal of some very old defines (at Gleb's request). Also, change the defines around the old transmit code to IXGBE_LEGACY_TX, I do this to make it possible to define this regardless of the OS level (it is not defined by default). There are also a couple changed comments for clarity.
2012-11-30 23:13:56 +00:00
int error, rid = 0;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
/* MSI RID at 1 */
if (adapter->msix == 1)
rid = 1;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
/* We allocate a single interrupt resource */
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
adapter->res = bus_alloc_resource_any(dev,
SYS_RES_IRQ, &rid, RF_SHAREABLE | RF_ACTIVE);
if (adapter->res == NULL) {
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
device_printf(dev, "Unable to allocate bus resource: "
"interrupt\n");
return (ENXIO);
}
/*
* Try allocating a fast interrupt and the associated deferred
* processing contexts.
*/
Patch #6 Whitespace cleanup, and removal of some very old defines (at Gleb's request). Also, change the defines around the old transmit code to IXGBE_LEGACY_TX, I do this to make it possible to define this regardless of the OS level (it is not defined by default). There are also a couple changed comments for clarity.
2012-11-30 23:13:56 +00:00
#ifndef IXGBE_LEGACY_TX
Merge similar fixes from 223198 from igb to ixgbe: - Use a dedicated task to handle deferred transmits from the if_transmit method instead of reusing the existing per-queue interrupt task. Reusing the per-queue interrupt task could result in both an interrupt thread and the taskqueue thread trying to handle received packets on a single queue resulting in out-of-order packet processing and lock contention. - Don't define ixgbe_start() at all where if_transmit is used. Tested by: Vijay Singh Reviewed by: jfv MFC after: 2 weeks
2012-09-26 18:11:43 +00:00
TASK_INIT(&txr->txq_task, 0, ixgbe_deferred_mq_start, txr);
#endif
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
TASK_INIT(&que->que_task, 0, ixgbe_handle_que, que);
que->tq = taskqueue_create_fast("ixgbe_que", M_NOWAIT,
taskqueue_thread_enqueue, &que->tq);
taskqueue_start_threads(&que->tq, 1, PI_NET, "%s ixq",
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
device_get_nameunit(adapter->dev));
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
/* Tasklets for Link, SFP and Multispeed Fiber */
TASK_INIT(&adapter->link_task, 0, ixgbe_handle_link, adapter);
TASK_INIT(&adapter->mod_task, 0, ixgbe_handle_mod, adapter);
TASK_INIT(&adapter->msf_task, 0, ixgbe_handle_msf, adapter);
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
TASK_INIT(&adapter->phy_task, 0, ixgbe_handle_phy, adapter);
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
#ifdef IXGBE_FDIR
TASK_INIT(&adapter->fdir_task, 0, ixgbe_reinit_fdir, adapter);
#endif
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
adapter->tq = taskqueue_create_fast("ixgbe_link", M_NOWAIT,
taskqueue_thread_enqueue, &adapter->tq);
taskqueue_start_threads(&adapter->tq, 1, PI_NET, "%s linkq",
device_get_nameunit(adapter->dev));
if ((error = bus_setup_intr(dev, adapter->res,
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
INTR_TYPE_NET | INTR_MPSAFE, NULL, ixgbe_legacy_irq,
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
que, &adapter->tag)) != 0) {
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
device_printf(dev, "Failed to register fast interrupt "
"handler: %d\n", error);
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
taskqueue_free(que->tq);
taskqueue_free(adapter->tq);
que->tq = NULL;
adapter->tq = NULL;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
return (error);
}
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
/* For simplicity in the handlers */
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
adapter->active_queues = IXGBE_EIMS_ENABLE_MASK;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
return (0);
}
/*********************************************************************
*
* Setup MSIX Interrupt resources and handlers
*
**********************************************************************/
static int
ixgbe_allocate_msix(struct adapter *adapter)
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
{
device_t dev = adapter->dev;
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
struct ix_queue *que = adapter->queues;
Merge similar fixes from 223198 from igb to ixgbe: - Use a dedicated task to handle deferred transmits from the if_transmit method instead of reusing the existing per-queue interrupt task. Reusing the per-queue interrupt task could result in both an interrupt thread and the taskqueue thread trying to handle received packets on a single queue resulting in out-of-order packet processing and lock contention. - Don't define ixgbe_start() at all where if_transmit is used. Tested by: Vijay Singh Reviewed by: jfv MFC after: 2 weeks
2012-09-26 18:11:43 +00:00
struct tx_ring *txr = adapter->tx_rings;
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
int error, rid, vector = 0;
Initialise these variables so gcc doesn't complain. Submitted by: luigi
2014-06-30 23:34:36 +00:00
int cpu_id = 0;
Fix ixgbe(4) to compile - with RSS; with ix+ixv in the kernel. * Fix the multiple same-named devclasses; the duplicate name trips up the linker. * Re-do the taskqueue stuff to use the new cpuset API, not the old pinned API. * Add includes for the new location of the RSS configuration routines. This allows ixgbe to compile as a module /and/ linked into the kernel, along with RSS working. Sponsored by: Norse Corp, Inc.
2015-03-18 05:05:30 +00:00
#ifdef RSS
cpuset_t cpu_mask;
#endif
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
#ifdef RSS
/*
* If we're doing RSS, the number of queues needs to
* match the number of RSS buckets that are configured.
*
* + If there's more queues than RSS buckets, we'll end
* up with queues that get no traffic.
*
* + If there's more RSS buckets than queues, we'll end
* up having multiple RSS buckets map to the same queue,
* so there'll be some contention.
*/
if (adapter->num_queues != rss_getnumbuckets()) {
device_printf(dev,
"%s: number of queues (%d) != number of RSS buckets (%d)"
"; performance will be impacted.\n",
__func__,
adapter->num_queues,
rss_getnumbuckets());
}
#endif
Merge similar fixes from 223198 from igb to ixgbe: - Use a dedicated task to handle deferred transmits from the if_transmit method instead of reusing the existing per-queue interrupt task. Reusing the per-queue interrupt task could result in both an interrupt thread and the taskqueue thread trying to handle received packets on a single queue resulting in out-of-order packet processing and lock contention. - Don't define ixgbe_start() at all where if_transmit is used. Tested by: Vijay Singh Reviewed by: jfv MFC after: 2 weeks
2012-09-26 18:11:43 +00:00
for (int i = 0; i < adapter->num_queues; i++, vector++, que++, txr++) {
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
rid = vector + 1;
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
que->res = bus_alloc_resource_any(dev, SYS_RES_IRQ, &rid,
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
RF_SHAREABLE | RF_ACTIVE);
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
if (que->res == NULL) {
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
device_printf(dev,"Unable to allocate"
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
" bus resource: que interrupt [%d]\n", vector);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
return (ENXIO);
}
/* Set the handler function */
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
error = bus_setup_intr(dev, que->res,
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
INTR_TYPE_NET | INTR_MPSAFE, NULL,
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
ixgbe_msix_que, que, &que->tag);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
if (error) {
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
que->res = NULL;
device_printf(dev, "Failed to register QUE handler");
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
return (error);
}
Remove a disable_queue from the beginning of the interrupt handler, automask handles it. Also, add in msix vector descriptions. MFC for 8.1 asap
2010-06-11 19:03:59 +00:00
#if __FreeBSD_version >= 800504
bus_describe_intr(dev, que->res, que->tag, "que %d", i);
#endif
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
que->msix = vector;
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
adapter->active_queues |= (u64)(1 << que->msix);
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
#ifdef RSS
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
/*
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
* The queue ID is used as the RSS layer bucket ID.
* We look up the queue ID -> RSS CPU ID and select
* that.
*/
cpu_id = rss_getcpu(i % rss_getnumbuckets());
#else
/*
* Bind the msix vector, and thus the
* rings to the corresponding cpu.
*
* This just happens to match the default RSS round-robin
* bucket -> queue -> CPU allocation.
*/
Decided to limit the interrupt bind to multiqueue config as done in igb.
2009-06-25 18:40:27 +00:00
if (adapter->num_queues > 1)
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
cpu_id = i;
#endif
if (adapter->num_queues > 1)
bus_bind_intr(dev, que->res, cpu_id);
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
#ifdef RSS
device_printf(dev,
"Bound RSS bucket %d to CPU %d\n",
i, cpu_id);
#else
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
if (bootverbose)
device_printf(dev,
"Bound queue %d to cpu %d\n",
i, cpu_id);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
#endif
Remove the sysctl process_limit interface, after some thought I've decided its overkill,a simple tuneable for each RX and TX limit, and then init sets the ring values based on that, should be sufficient. More importantly, fix a bug causing a panic, when changing the define style to IXGBE_LEGACY_TX a taskqueue init was inadvertently set #ifdef when it should be #ifndef.
2012-12-03 21:38:02 +00:00
#ifndef IXGBE_LEGACY_TX
Merge similar fixes from 223198 from igb to ixgbe: - Use a dedicated task to handle deferred transmits from the if_transmit method instead of reusing the existing per-queue interrupt task. Reusing the per-queue interrupt task could result in both an interrupt thread and the taskqueue thread trying to handle received packets on a single queue resulting in out-of-order packet processing and lock contention. - Don't define ixgbe_start() at all where if_transmit is used. Tested by: Vijay Singh Reviewed by: jfv MFC after: 2 weeks
2012-09-26 18:11:43 +00:00
TASK_INIT(&txr->txq_task, 0, ixgbe_deferred_mq_start, txr);
#endif
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
TASK_INIT(&que->que_task, 0, ixgbe_handle_que, que);
que->tq = taskqueue_create_fast("ixgbe_que", M_NOWAIT,
taskqueue_thread_enqueue, &que->tq);
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
#ifdef RSS
Fix ixgbe(4) to compile - with RSS; with ix+ixv in the kernel. * Fix the multiple same-named devclasses; the duplicate name trips up the linker. * Re-do the taskqueue stuff to use the new cpuset API, not the old pinned API. * Add includes for the new location of the RSS configuration routines. This allows ixgbe to compile as a module /and/ linked into the kernel, along with RSS working. Sponsored by: Norse Corp, Inc.
2015-03-18 05:05:30 +00:00
CPU_SETOF(cpu_id, &cpu_mask);
taskqueue_start_threads_cpuset(&que->tq, 1, PI_NET,
&cpu_mask,
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
"%s (bucket %d)",
device_get_nameunit(adapter->dev),
cpu_id);
#else
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
taskqueue_start_threads(&que->tq, 1, PI_NET, "%s que",
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
device_get_nameunit(adapter->dev));
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
#endif
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
/* and Link */
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
rid = vector + 1;
adapter->res = bus_alloc_resource_any(dev,
SYS_RES_IRQ, &rid, RF_SHAREABLE | RF_ACTIVE);
if (!adapter->res) {
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
device_printf(dev,"Unable to allocate"
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
" bus resource: Link interrupt [%d]\n", rid);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
return (ENXIO);
}
/* Set the link handler function */
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
error = bus_setup_intr(dev, adapter->res,
INTR_TYPE_NET | INTR_MPSAFE, NULL,
ixgbe_msix_link, adapter, &adapter->tag);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
if (error) {
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
adapter->res = NULL;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
device_printf(dev, "Failed to register LINK handler");
return (error);
}
Remove a disable_queue from the beginning of the interrupt handler, automask handles it. Also, add in msix vector descriptions. MFC for 8.1 asap
2010-06-11 19:03:59 +00:00
#if __FreeBSD_version >= 800504
bus_describe_intr(dev, adapter->res, adapter->tag, "link");
#endif
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
adapter->vector = vector;
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
/* Tasklets for Link, SFP and Multispeed Fiber */
TASK_INIT(&adapter->link_task, 0, ixgbe_handle_link, adapter);
TASK_INIT(&adapter->mod_task, 0, ixgbe_handle_mod, adapter);
TASK_INIT(&adapter->msf_task, 0, ixgbe_handle_msf, adapter);
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
TASK_INIT(&adapter->phy_task, 0, ixgbe_handle_phy, adapter);
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
#ifdef IXGBE_FDIR
TASK_INIT(&adapter->fdir_task, 0, ixgbe_reinit_fdir, adapter);
#endif
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
adapter->tq = taskqueue_create_fast("ixgbe_link", M_NOWAIT,
taskqueue_thread_enqueue, &adapter->tq);
taskqueue_start_threads(&adapter->tq, 1, PI_NET, "%s linkq",
device_get_nameunit(adapter->dev));
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
return (0);
}
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
/*
* Setup Either MSI/X or MSI
*/
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
static int
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
ixgbe_setup_msix(struct adapter *adapter)
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
{
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
device_t dev = adapter->dev;
int rid, want, queues, msgs;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
/* Override by tuneable */
if (ixgbe_enable_msix == 0)
goto msi;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
/* First try MSI/X */
Make the various driver MSIX setup routines fallback to MSI more gracefully. This change was suggested by Marius Strobl, thank you. PR: kern/181016 MFC after: ASAP
2013-08-06 21:01:38 +00:00
msgs = pci_msix_count(dev);
if (msgs == 0)
goto msi;
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
rid = PCIR_BAR(MSIX_82598_BAR);
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
adapter->msix_mem = bus_alloc_resource_any(dev,
SYS_RES_MEMORY, &rid, RF_ACTIVE);
Make the various driver MSIX setup routines fallback to MSI more gracefully. This change was suggested by Marius Strobl, thank you. PR: kern/181016 MFC after: ASAP
2013-08-06 21:01:38 +00:00
if (adapter->msix_mem == NULL) {
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
rid += 4; /* 82599 maps in higher BAR */
adapter->msix_mem = bus_alloc_resource_any(dev,
SYS_RES_MEMORY, &rid, RF_ACTIVE);
}
Make the various driver MSIX setup routines fallback to MSI more gracefully. This change was suggested by Marius Strobl, thank you. PR: kern/181016 MFC after: ASAP
2013-08-06 21:01:38 +00:00
if (adapter->msix_mem == NULL) {
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
/* May not be enabled */
device_printf(adapter->dev,
"Unable to map MSIX table \n");
goto msi;
}
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
/* Figure out a reasonable auto config value */
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
queues = (mp_ncpus > (msgs-1)) ? (msgs-1) : mp_ncpus;
Revert r275136, it was not approved, it was sloppy, if a feature like this is needed please resubmit for Intel's approval.
2014-12-02 23:02:57 +00:00
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
#ifdef RSS
/* If we're doing RSS, clamp at the number of RSS buckets */
if (queues > rss_getnumbuckets())
queues = rss_getnumbuckets();
#endif
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
if (ixgbe_num_queues != 0)
queues = ixgbe_num_queues;
Revert r275136, it was not approved, it was sloppy, if a feature like this is needed please resubmit for Intel's approval.
2014-12-02 23:02:57 +00:00
/* reflect correct sysctl value */
ixgbe_num_queues = queues;
A few changes: When not defining header split do not allocate mbufs, this can be a BIG savings in the mbuf memory pool. Also keep seperate dma maps for the header and payload pieces when doing header split. The basis of this code was a patch done a while ago by yongari, thank you :) A number of white space changes. MFC: in a few days
2010-05-19 00:03:48 +00:00
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
/*
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
** Want one vector (RX/TX pair) per queue
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
** plus an additional for Link.
*/
A few changes: When not defining header split do not allocate mbufs, this can be a BIG savings in the mbuf memory pool. Also keep seperate dma maps for the header and payload pieces when doing header split. The basis of this code was a patch done a while ago by yongari, thank you :) A number of white space changes. MFC: in a few days
2010-05-19 00:03:48 +00:00
want = queues + 1;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
if (msgs >= want)
msgs = want;
else {
device_printf(adapter->dev,
"MSIX Configuration Problem, "
"%d vectors but %d queues wanted!\n",
msgs, want);
Make the various driver MSIX setup routines fallback to MSI more gracefully. This change was suggested by Marius Strobl, thank you. PR: kern/181016 MFC after: ASAP
2013-08-06 21:01:38 +00:00
goto msi;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
Improve the MSIX setup code in the drivers, thanks to Marius for the changes. Make sure that pci_alloc_msix() does give us the vectors we need and fall back to MSI when it doesn't, also release any that were allocated when insufficient. MFC after: 3 days
2013-08-12 22:54:38 +00:00
if ((pci_alloc_msix(dev, &msgs) == 0) && (msgs == want)) {
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
device_printf(adapter->dev,
"Using MSIX interrupts with %d vectors\n", msgs);
A few changes: When not defining header split do not allocate mbufs, this can be a BIG savings in the mbuf memory pool. Also keep seperate dma maps for the header and payload pieces when doing header split. The basis of this code was a patch done a while ago by yongari, thank you :) A number of white space changes. MFC: in a few days
2010-05-19 00:03:48 +00:00
adapter->num_queues = queues;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
return (msgs);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
Improve the MSIX setup code in the drivers, thanks to Marius for the changes. Make sure that pci_alloc_msix() does give us the vectors we need and fall back to MSI when it doesn't, also release any that were allocated when insufficient. MFC after: 3 days
2013-08-12 22:54:38 +00:00
/*
** If MSIX alloc failed or provided us with
** less than needed, free and fall through to MSI
*/
pci_release_msi(dev);
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
msi:
Make the various driver MSIX setup routines fallback to MSI more gracefully. This change was suggested by Marius Strobl, thank you. PR: kern/181016 MFC after: ASAP
2013-08-06 21:01:38 +00:00
if (adapter->msix_mem != NULL) {
bus_release_resource(dev, SYS_RES_MEMORY,
rid, adapter->msix_mem);
adapter->msix_mem = NULL;
}
msgs = 1;
if (pci_alloc_msi(dev, &msgs) == 0) {
First off: update the driver README, the old one was horribly crusty, and this still isn't perfect, but its at least a bit more recent. Secondly, a few improvements to the driver from Andrew Boyer, support hint to allow devices to not attach, add VLAN_HWTSO capability so vlans can use TSO, fix in the interrupt handler to make sure the stack TX queue is processed. Oh, and also make sure IPv6 does not cause a re-init in the ioctl routine. Thanks for your efforts Andrew! Thanks to Claudio Jeker for noticing the ixgbe_xmit() routine was not correctly swapping the dma map from the first to the last descriptor in a multi-descriptor transmission, corrected this.
2011-06-02 00:34:57 +00:00
device_printf(adapter->dev,"Using an MSI interrupt\n");
Make the various driver MSIX setup routines fallback to MSI more gracefully. This change was suggested by Marius Strobl, thank you. PR: kern/181016 MFC after: ASAP
2013-08-06 21:01:38 +00:00
return (msgs);
}
device_printf(adapter->dev,"Using a Legacy interrupt\n");
return (0);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
static int
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
ixgbe_allocate_pci_resources(struct adapter *adapter)
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
{
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
int rid;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
device_t dev = adapter->dev;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
rid = PCIR_BAR(0);
adapter->pci_mem = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
&rid, RF_ACTIVE);
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
if (!(adapter->pci_mem)) {
device_printf(dev,"Unable to allocate bus resource: memory\n");
return (ENXIO);
}
adapter->osdep.mem_bus_space_tag =
rman_get_bustag(adapter->pci_mem);
adapter->osdep.mem_bus_space_handle =
rman_get_bushandle(adapter->pci_mem);
adapter->hw.hw_addr = (u8 *) &adapter->osdep.mem_bus_space_handle;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
/* Legacy defaults */
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
adapter->num_queues = 1;
adapter->hw.back = &adapter->osdep;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
/*
** Now setup MSI or MSI/X, should
** return us the number of supported
** vectors. (Will be 1 for MSI)
*/
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
adapter->msix = ixgbe_setup_msix(adapter);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
return (0);
}
static void
ixgbe_free_pci_resources(struct adapter * adapter)
{
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
struct ix_queue *que = adapter->queues;
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
device_t dev = adapter->dev;
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
int rid, memrid;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
if (adapter->hw.mac.type == ixgbe_mac_82598EB)
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
memrid = PCIR_BAR(MSIX_82598_BAR);
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
else
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
memrid = PCIR_BAR(MSIX_82599_BAR);
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/*
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
** There is a slight possibility of a failure mode
** in attach that will result in entering this function
** before interrupt resources have been initialized, and
** in that case we do not want to execute the loops below
** We can detect this reliably by the state of the adapter
** res pointer.
*/
if (adapter->res == NULL)
goto mem;
/*
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
** Release all msix queue resources:
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
*/
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
for (int i = 0; i < adapter->num_queues; i++, que++) {
rid = que->msix + 1;
if (que->tag != NULL) {
bus_teardown_intr(dev, que->res, que->tag);
que->tag = NULL;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
if (que->res != NULL)
bus_release_resource(dev, SYS_RES_IRQ, rid, que->res);
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
}
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
/* Clean the Legacy or Link interrupt last */
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
if (adapter->vector) /* we are doing MSIX */
rid = adapter->vector + 1;
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
else
(adapter->msix != 0) ? (rid = 1):(rid = 0);
if (adapter->tag != NULL) {
bus_teardown_intr(dev, adapter->res, adapter->tag);
adapter->tag = NULL;
}
if (adapter->res != NULL)
bus_release_resource(dev, SYS_RES_IRQ, rid, adapter->res);
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
mem:
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
if (adapter->msix)
pci_release_msi(dev);
if (adapter->msix_mem != NULL)
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
bus_release_resource(dev, SYS_RES_MEMORY,
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
memrid, adapter->msix_mem);
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
if (adapter->pci_mem != NULL)
bus_release_resource(dev, SYS_RES_MEMORY,
PCIR_BAR(0), adapter->pci_mem);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
return;
}
/*********************************************************************
*
* Setup networking device structure and register an interface.
*
**********************************************************************/
A couple fixes got clobbered, putting them back.
2010-11-26 23:57:13 +00:00
static int
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
ixgbe_setup_interface(device_t dev, struct adapter *adapter)
{
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
struct ifnet *ifp;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
INIT_DEBUGOUT("ixgbe_setup_interface: begin");
ifp = adapter->ifp = if_alloc(IFT_ETHER);
A couple fixes got clobbered, putting them back.
2010-11-26 23:57:13 +00:00
if (ifp == NULL) {
device_printf(dev, "can not allocate ifnet structure\n");
return (-1);
}
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
if_initname(ifp, device_get_name(dev), device_get_unit(dev));
Since 32-bit if_baudrate isn't enough to describe a baud rate of a 10 Gbit interface, in the r241616 a crutch was provided. It didn't work well, and finally we decided that it is time to break ABI and simply make if_baudrate a 64-bit value. Meanwhile, the entire struct if_data was reviewed. o Remove the if_baudrate_pf crutch. o Make all fields of struct if_data fixed machine independent size. The notion of data (packet counters, etc) are by no means MD. And it is a bug that on amd64 we've got a 64-bit counters, while on i386 32-bit, which at modern speeds overflow within a second. This also removes quite a lot of COMPAT_FREEBSD32 code. o Give 16 bit for the ifi_datalen field. This field was provided to make future changes to if_data less ABI breaking. Unfortunately the 8 bit size of it had effectively limited sizeof if_data to 256 bytes. o Give 32 bits to ifi_mtu and ifi_metric. o Give 64 bits to the rest of fields, since they are counters. __FreeBSD_version bumped. Discussed with: emax Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-03-13 03:42:24 +00:00
ifp->if_baudrate = IF_Gbps(10);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
ifp->if_init = ixgbe_init;
ifp->if_softc = adapter;
ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
ifp->if_ioctl = ixgbe_ioctl;
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
#if __FreeBSD_version >= 1100036
if_setgetcounterfn(ifp, ixgbe_get_counter);
#endif
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
#if __FreeBSD_version >= 1100045
/* TSO parameters */
ifp->if_hw_tsomax = 65518;
ifp->if_hw_tsomaxsegcount = IXGBE_82599_SCATTER;
ifp->if_hw_tsomaxsegsize = 2048;
#endif
Patch #6 Whitespace cleanup, and removal of some very old defines (at Gleb's request). Also, change the defines around the old transmit code to IXGBE_LEGACY_TX, I do this to make it possible to define this regardless of the OS level (it is not defined by default). There are also a couple changed comments for clarity.
2012-11-30 23:13:56 +00:00
#ifndef IXGBE_LEGACY_TX
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
ifp->if_transmit = ixgbe_mq_start;
ifp->if_qflush = ixgbe_qflush;
Merge similar fixes from 223198 from igb to ixgbe: - Use a dedicated task to handle deferred transmits from the if_transmit method instead of reusing the existing per-queue interrupt task. Reusing the per-queue interrupt task could result in both an interrupt thread and the taskqueue thread trying to handle received packets on a single queue resulting in out-of-order packet processing and lock contention. - Don't define ixgbe_start() at all where if_transmit is used. Tested by: Vijay Singh Reviewed by: jfv MFC after: 2 weeks
2012-09-26 18:11:43 +00:00
#else
ifp->if_start = ixgbe_start;
IFQ_SET_MAXLEN(&ifp->if_snd, adapter->num_tx_desc - 2);
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
ifp->if_snd.ifq_drv_maxlen = adapter->num_tx_desc - 2;
IFQ_SET_READY(&ifp->if_snd);
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
#endif
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
ether_ifattach(ifp, adapter->hw.mac.addr);
adapter->max_frame_size =
ifp->if_mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
/*
* Tell the upper layer(s) we support long frames.
*/
Use define from if_var.h to access a field inside struct if_data, that resides in struct ifnet. Sponsored by: Nginx, Inc.
2014-08-30 19:55:54 +00:00
ifp->if_hdrlen = sizeof(struct ether_vlan_header);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
MFp4 bz_ipv6_fast: Add TSO6 and LRO/IPv6 support. Fix the module Makefile to at least properly inlcude opt_inet6.h and allow builds without INET or INET6. Sponsored by: The FreeBSD Foundation Sponsored by: iXsystems Reviewed by: gnn (as part of the whole) MFC After: 3 days
2012-05-25 03:02:56 +00:00
ifp->if_capabilities |= IFCAP_HWCSUM | IFCAP_TSO | IFCAP_VLAN_HWCSUM;
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
ifp->if_capabilities |= IFCAP_JUMBO_MTU;
A few important fixes: - Testing TSO6 has led me to discover that HW RSC is a problematic feature, it is ONLY designed to work with IPv4 in the first place, and if IP forwarding is done it can't be disabled as LRO in the stack, also initial testing we've done at Intel shows an equal performance using TSO[46] on the TX and LRO on RX, if you ran older code on 82599 or later hardware you actually could have detrimental performance for this reason. So I am disabling the feature by default and all our adapters will now use LRO instead. - If you have flow control off and multiple queues it was possible when the buffer of one queue becomes full that all RX movement is stalled, to eliminate this problem a feature bit is now set that will allow packets to be dropped when full rather than stall. Note, the default is to have flow control on, and this keeps this from happening. - Because of the recent fixes in the stack, LRO is now auto-disabled when problematic, so I have decided to enable it by default in the capabilities in the driver. - There are some 1G modules used by some customers, a couple small tweaks to properly support those in the media code. - A note: we have now done some testing of TSO6 and using LRO with IPv6 and it all works great!! Seeing line rate in both directions in best cases. Thanks bz for your excellent work!!
2012-10-31 23:50:36 +00:00
ifp->if_capabilities |= IFCAP_LRO;
First off: update the driver README, the old one was horribly crusty, and this still isn't perfect, but its at least a bit more recent. Secondly, a few improvements to the driver from Andrew Boyer, support hint to allow devices to not attach, add VLAN_HWTSO capability so vlans can use TSO, fix in the interrupt handler to make sure the stack TX queue is processed. Oh, and also make sure IPv6 does not cause a re-init in the ioctl routine. Thanks for your efforts Andrew! Thanks to Claudio Jeker for noticing the ixgbe_xmit() routine was not correctly swapping the dma map from the first to the last descriptor in a multi-descriptor transmission, corrected this.
2011-06-02 00:34:57 +00:00
ifp->if_capabilities |= IFCAP_VLAN_HWTAGGING
| IFCAP_VLAN_HWTSO
There are some high performance NICs that count statistics in hardware, and there are ifnets, that do that via counter(9). Provide a flag that would skip cache line trashing '+=' operation in ether_input(). Sponsored by: Netflix Sponsored by: Nginx, Inc. Reviewed by: melifaro, adrian Approved by: re (marius)
2013-10-09 19:04:40 +00:00
| IFCAP_VLAN_MTU
| IFCAP_HWSTATS;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
ifp->if_capenable = ifp->if_capabilities;
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
/*
First off: update the driver README, the old one was horribly crusty, and this still isn't perfect, but its at least a bit more recent. Secondly, a few improvements to the driver from Andrew Boyer, support hint to allow devices to not attach, add VLAN_HWTSO capability so vlans can use TSO, fix in the interrupt handler to make sure the stack TX queue is processed. Oh, and also make sure IPv6 does not cause a re-init in the ioctl routine. Thanks for your efforts Andrew! Thanks to Claudio Jeker for noticing the ixgbe_xmit() routine was not correctly swapping the dma map from the first to the last descriptor in a multi-descriptor transmission, corrected this.
2011-06-02 00:34:57 +00:00
** Don't turn this on by default, if vlans are
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
** created on another pseudo device (eg. lagg)
** then vlan events are not passed thru, breaking
** operation, but with HW FILTER off it works. If
First off: update the driver README, the old one was horribly crusty, and this still isn't perfect, but its at least a bit more recent. Secondly, a few improvements to the driver from Andrew Boyer, support hint to allow devices to not attach, add VLAN_HWTSO capability so vlans can use TSO, fix in the interrupt handler to make sure the stack TX queue is processed. Oh, and also make sure IPv6 does not cause a re-init in the ioctl routine. Thanks for your efforts Andrew! Thanks to Claudio Jeker for noticing the ixgbe_xmit() routine was not correctly swapping the dma map from the first to the last descriptor in a multi-descriptor transmission, corrected this.
2011-06-02 00:34:57 +00:00
** using vlans directly on the ixgbe driver you can
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
** enable this and get full hardware tag filtering.
*/
ifp->if_capabilities |= IFCAP_VLAN_HWFILTER;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/*
* Specify the media types supported by this adapter and register
* callbacks to update media and link information
*/
ifmedia_init(&adapter->media, IFM_IMASK, ixgbe_media_change,
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
ixgbe_media_status);
ixgbe_add_media_types(adapter);
/* Autoselect media by default */
ifmedia_set(&adapter->media, IFM_ETHER | IFM_AUTO);
return (0);
}
static void
ixgbe_add_media_types(struct adapter *adapter)
{
struct ixgbe_hw *hw = &adapter->hw;
device_t dev = adapter->dev;
int layer;
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
layer = ixgbe_get_supported_physical_layer(hw);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
/* Media types with matching FreeBSD media defines */
if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_T)
ifmedia_add(&adapter->media, IFM_ETHER | IFM_10G_T, 0, NULL);
if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_T)
ifmedia_add(&adapter->media, IFM_ETHER | IFM_1000_T, 0, NULL);
if (layer & IXGBE_PHYSICAL_LAYER_100BASE_TX)
ifmedia_add(&adapter->media, IFM_ETHER | IFM_100_TX, 0, NULL);
if (layer & IXGBE_PHYSICAL_LAYER_SFP_PLUS_CU ||
layer & IXGBE_PHYSICAL_LAYER_SFP_ACTIVE_DA)
ifmedia_add(&adapter->media, IFM_ETHER | IFM_10G_TWINAX, 0, NULL);
if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_LR)
ifmedia_add(&adapter->media, IFM_ETHER | IFM_10G_LR, 0, NULL);
if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_SR)
ifmedia_add(&adapter->media, IFM_ETHER | IFM_10G_SR, 0, NULL);
if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_CX4)
ifmedia_add(&adapter->media, IFM_ETHER | IFM_10G_CX4, 0, NULL);
if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_SX)
ifmedia_add(&adapter->media, IFM_ETHER | IFM_1000_SX, 0, NULL);
/*
** Other (no matching FreeBSD media type):
** To workaround this, we'll assign these completely
** inappropriate media types.
*/
if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KR) {
device_printf(dev, "Media supported: 10GbaseKR\n");
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
device_printf(dev, "10GbaseKR mapped to 10GbaseSR\n");
ifmedia_add(&adapter->media, IFM_ETHER | IFM_10G_SR, 0, NULL);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
}
if (layer & IXGBE_PHYSICAL_LAYER_10GBASE_KX4) {
device_printf(dev, "Media supported: 10GbaseKX4\n");
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
device_printf(dev, "10GbaseKX4 mapped to 10GbaseCX4\n");
ifmedia_add(&adapter->media, IFM_ETHER | IFM_10G_CX4, 0, NULL);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
}
if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_KX) {
device_printf(dev, "Media supported: 1000baseKX\n");
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
device_printf(dev, "1000baseKX mapped to 1000baseCX\n");
ifmedia_add(&adapter->media, IFM_ETHER | IFM_1000_CX, 0, NULL);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
}
if (layer & IXGBE_PHYSICAL_LAYER_1000BASE_BX) {
/* Someday, someone will care about you... */
device_printf(dev, "Media supported: 1000baseBX\n");
}
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
if (hw->device_id == IXGBE_DEV_ID_82598AT) {
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
ifmedia_add(&adapter->media,
IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
ifmedia_add(&adapter->media,
IFM_ETHER | IFM_1000_T, 0, NULL);
}
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
ifmedia_add(&adapter->media, IFM_ETHER | IFM_AUTO, 0, NULL);
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
}
static void
ixgbe_config_link(struct adapter *adapter)
{
struct ixgbe_hw *hw = &adapter->hw;
u32 autoneg, err = 0;
bool sfp, negotiate;
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
sfp = ixgbe_is_sfp(hw);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
if (sfp) {
if (hw->phy.multispeed_fiber) {
hw->mac.ops.setup_sfp(hw);
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
ixgbe_enable_tx_laser(hw);
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
taskqueue_enqueue(adapter->tq, &adapter->msf_task);
} else
taskqueue_enqueue(adapter->tq, &adapter->mod_task);
} else {
if (hw->mac.ops.check_link)
First, sync to internal shared code, and then Fixes: - flow control - don't override user value on re-init - fix to make 1G optics work correctly - change to interrupt enabling - some bits were incorrect for certain hardware. - certain stats fixes, remove a duplicate increment of ierror, thanks to Scott Long for pointing these out. - shared code link interface changed, requiring some core code changes to accomodate this. - add an m_adj() to ETHER_ALIGN on the recieve side, this was requested by Mike Karels, thanks Mike. - Multicast code corrections also thanks to Mike Karels.
2013-03-04 23:07:40 +00:00
err = ixgbe_check_link(hw, &adapter->link_speed,
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
&adapter->link_up, FALSE);
if (err)
goto out;
autoneg = hw->phy.autoneg_advertised;
if ((!autoneg) && (hw->mac.ops.get_link_capabilities))
err = hw->mac.ops.get_link_capabilities(hw,
&autoneg, &negotiate);
if (err)
goto out;
if (hw->mac.ops.setup_link)
First, sync to internal shared code, and then Fixes: - flow control - don't override user value on re-init - fix to make 1G optics work correctly - change to interrupt enabling - some bits were incorrect for certain hardware. - certain stats fixes, remove a duplicate increment of ierror, thanks to Scott Long for pointing these out. - shared code link interface changed, requiring some core code changes to accomodate this. - add an m_adj() to ETHER_ALIGN on the recieve side, this was requested by Mike Karels, thanks Mike. - Multicast code corrections also thanks to Mike Karels.
2013-03-04 23:07:40 +00:00
err = hw->mac.ops.setup_link(hw,
autoneg, adapter->link_up);
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
}
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
out:
return;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/*********************************************************************
*
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
* Enable transmit units.
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
*
**********************************************************************/
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
static void
ixgbe_initialize_transmit_units(struct adapter *adapter)
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
{
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
struct tx_ring *txr = adapter->tx_rings;
struct ixgbe_hw *hw = &adapter->hw;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
/* Setup the Base and Length of the Tx Descriptor Ring */
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
for (int i = 0; i < adapter->num_queues; i++, txr++) {
u64 tdba = txr->txdma.dma_paddr;
u32 txctrl = 0;
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
IXGBE_WRITE_REG(hw, IXGBE_TDBAL(i),
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
(tdba & 0x00000000ffffffffULL));
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
IXGBE_WRITE_REG(hw, IXGBE_TDBAH(i), (tdba >> 32));
IXGBE_WRITE_REG(hw, IXGBE_TDLEN(i),
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
adapter->num_tx_desc * sizeof(union ixgbe_adv_tx_desc));
/* Setup the HW Tx Head and Tail descriptor pointers */
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
IXGBE_WRITE_REG(hw, IXGBE_TDH(i), 0);
IXGBE_WRITE_REG(hw, IXGBE_TDT(i), 0);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
/* Cache the tail address */
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
txr->tail = IXGBE_TDT(txr->me);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
/* Set the processing limit */
txr->process_limit = ixgbe_tx_process_limit;
/* Disable Head Writeback */
switch (hw->mac.type) {
case ixgbe_mac_82598EB:
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
txctrl = IXGBE_READ_REG(hw, IXGBE_DCA_TXCTRL(i));
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
break;
case ixgbe_mac_82599EB:
case ixgbe_mac_X540:
default:
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
txctrl = IXGBE_READ_REG(hw, IXGBE_DCA_TXCTRL_82599(i));
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
break;
}
txctrl &= ~IXGBE_DCA_TXCTRL_DESC_WRO_EN;
switch (hw->mac.type) {
case ixgbe_mac_82598EB:
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
IXGBE_WRITE_REG(hw, IXGBE_DCA_TXCTRL(i), txctrl);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
break;
case ixgbe_mac_82599EB:
case ixgbe_mac_X540:
default:
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
IXGBE_WRITE_REG(hw, IXGBE_DCA_TXCTRL_82599(i), txctrl);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
break;
}
}
if (hw->mac.type != ixgbe_mac_82598EB) {
u32 dmatxctl, rttdcs;
dmatxctl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
dmatxctl |= IXGBE_DMATXCTL_TE;
IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, dmatxctl);
/* Disable arbiter to set MTQC */
rttdcs = IXGBE_READ_REG(hw, IXGBE_RTTDCS);
rttdcs |= IXGBE_RTTDCS_ARBDIS;
IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
IXGBE_WRITE_REG(hw, IXGBE_MTQC, IXGBE_MTQC_64Q_1PB);
rttdcs &= ~IXGBE_RTTDCS_ARBDIS;
IXGBE_WRITE_REG(hw, IXGBE_RTTDCS, rttdcs);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
return;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
static void
ixgbe_initialise_rss_mapping(struct adapter *adapter)
{
struct ixgbe_hw *hw = &adapter->hw;
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
uint32_t reta;
int i, j, queue_id, table_size;
int index_mult;
uint32_t rss_key[10];
uint32_t mrqc;
Teach ixgbe(4) about rss_gethashconfig(). If RSS is enabled, ixgbe(4) will query the RSS API for the types of hashes which should be used. It'll then only enable hashes that are exposed via the RSS layer. This way it won't try to do things like enable UDP hashing if RSS explicitly states that it isn't supported in lookups. Tested: * 82599EB ixgbe(4) NIC
2014-07-20 07:45:48 +00:00
#ifdef RSS
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
uint32_t rss_hash_config;
Teach ixgbe(4) about rss_gethashconfig(). If RSS is enabled, ixgbe(4) will query the RSS API for the types of hashes which should be used. It'll then only enable hashes that are exposed via the RSS layer. This way it won't try to do things like enable UDP hashing if RSS explicitly states that it isn't supported in lookups. Tested: * 82599EB ixgbe(4) NIC
2014-07-20 07:45:48 +00:00
#endif
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
/* Setup RSS */
reta = 0;
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
#ifdef RSS
/* Fetch the configured RSS key */
rss_getkey((uint8_t *) &rss_key);
#else
/* set up random bits */
arc4rand(&rss_key, sizeof(rss_key), 0);
#endif
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
/* Set multiplier for RETA setup and table size based on MAC */
index_mult = 0x1;
table_size = 128;
switch (adapter->hw.mac.type) {
case ixgbe_mac_82598EB:
index_mult = 0x11;
break;
case ixgbe_mac_X550:
case ixgbe_mac_X550EM_x:
table_size = 512;
break;
default:
break;
}
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
/* Set up the redirection table */
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
for (i = 0, j = 0; i < table_size; i++, j++) {
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
if (j == adapter->num_queues) j = 0;
#ifdef RSS
/*
* Fetch the RSS bucket id for the given indirection entry.
* Cap it at the number of configured buckets (which is
* num_queues.)
*/
queue_id = rss_get_indirection_to_bucket(i);
queue_id = queue_id % adapter->num_queues;
#else
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
queue_id = (j * index_mult);
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
#endif
Correctly program the RSS redirection table entries. Without this, the RSS bucket assignments aren't correct - they're DCBA instead of ABCD in each DWORD. Tested: 82599EB ixgbe(4), TCP and UDP RSS
2014-07-20 04:11:18 +00:00
/*
* The low 8 bits are for hash value (n+0);
* The next 8 bits are for hash value (n+1), etc.
*/
reta = reta >> 8;
reta = reta | ( ((uint32_t) queue_id) << 24);
if ((i & 3) == 3) {
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
if (i < 128)
IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
else
IXGBE_WRITE_REG(hw, IXGBE_ERETA((i >> 2) - 32), reta);
Correctly program the RSS redirection table entries. Without this, the RSS bucket assignments aren't correct - they're DCBA instead of ABCD in each DWORD. Tested: 82599EB ixgbe(4), TCP and UDP RSS
2014-07-20 04:11:18 +00:00
reta = 0;
}
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
}
/* Now fill our hash function seeds */
for (int i = 0; i < 10; i++)
IXGBE_WRITE_REG(hw, IXGBE_RSSRK(i), rss_key[i]);
/* Perform hash on these packet types */
Teach ixgbe(4) about rss_gethashconfig(). If RSS is enabled, ixgbe(4) will query the RSS API for the types of hashes which should be used. It'll then only enable hashes that are exposed via the RSS layer. This way it won't try to do things like enable UDP hashing if RSS explicitly states that it isn't supported in lookups. Tested: * 82599EB ixgbe(4) NIC
2014-07-20 07:45:48 +00:00
#ifdef RSS
mrqc = IXGBE_MRQC_RSSEN;
rss_hash_config = rss_gethashconfig();
if (rss_hash_config & RSS_HASHTYPE_RSS_IPV4)
mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4;
if (rss_hash_config & RSS_HASHTYPE_RSS_TCP_IPV4)
mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4_TCP;
if (rss_hash_config & RSS_HASHTYPE_RSS_IPV6)
mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6;
if (rss_hash_config & RSS_HASHTYPE_RSS_TCP_IPV6)
mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_TCP;
if (rss_hash_config & RSS_HASHTYPE_RSS_IPV6_EX)
mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_EX;
if (rss_hash_config & RSS_HASHTYPE_RSS_TCP_IPV6_EX)
mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_EX_TCP;
if (rss_hash_config & RSS_HASHTYPE_RSS_UDP_IPV4)
mrqc |= IXGBE_MRQC_RSS_FIELD_IPV4_UDP;
if (rss_hash_config & RSS_HASHTYPE_RSS_UDP_IPV4_EX)
device_printf(adapter->dev,
"%s: RSS_HASHTYPE_RSS_UDP_IPV4_EX defined, "
"but not supported\n", __func__);
if (rss_hash_config & RSS_HASHTYPE_RSS_UDP_IPV6)
mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_UDP;
if (rss_hash_config & RSS_HASHTYPE_RSS_UDP_IPV6_EX)
mrqc |= IXGBE_MRQC_RSS_FIELD_IPV6_EX_UDP;
#else
Disable the ixgbe(4) UDP 4-tuple hashing for the time being. A mix of fragmented and non-fragmented UDP in a single stream will end up being hashed differently, resulting in out-of-order behaviour in the receive path. This was done in the linux e1000 driver in 2011. Discussed with: jfv
2014-07-20 07:43:41 +00:00
/*
* Disable UDP - IP fragments aren't currently being handled
* and so we end up with a mix of 2-tuple and 4-tuple
* traffic.
*/
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
mrqc = IXGBE_MRQC_RSSEN
| IXGBE_MRQC_RSS_FIELD_IPV4
| IXGBE_MRQC_RSS_FIELD_IPV4_TCP
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
#if 0
| IXGBE_MRQC_RSS_FIELD_IPV4_UDP
#endif
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
| IXGBE_MRQC_RSS_FIELD_IPV6_EX_TCP
| IXGBE_MRQC_RSS_FIELD_IPV6_EX
| IXGBE_MRQC_RSS_FIELD_IPV6
| IXGBE_MRQC_RSS_FIELD_IPV6_TCP
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
#if 0
| IXGBE_MRQC_RSS_FIELD_IPV6_UDP
| IXGBE_MRQC_RSS_FIELD_IPV6_EX_UDP
#endif
Disable the ixgbe(4) UDP 4-tuple hashing for the time being. A mix of fragmented and non-fragmented UDP in a single stream will end up being hashed differently, resulting in out-of-order behaviour in the receive path. This was done in the linux e1000 driver in 2011. Discussed with: jfv
2014-07-20 07:43:41 +00:00
;
Teach ixgbe(4) about rss_gethashconfig(). If RSS is enabled, ixgbe(4) will query the RSS API for the types of hashes which should be used. It'll then only enable hashes that are exposed via the RSS layer. This way it won't try to do things like enable UDP hashing if RSS explicitly states that it isn't supported in lookups. Tested: * 82599EB ixgbe(4) NIC
2014-07-20 07:45:48 +00:00
#endif /* RSS */
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
IXGBE_WRITE_REG(hw, IXGBE_MRQC, mrqc);
}
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/*********************************************************************
*
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
* Setup receive registers and features.
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
*
**********************************************************************/
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
#define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT 2
The maximum read size of incoming packets is done in 1024-byte increments. The current code was rounding down the maximum frame size instead of routing up, resulting in a read size of 1024 bytes, in the non-jumbo frame case, and splitting the packets across multiple mbufs. Consequently the above problem exposed another issue, which is when packets were splitted across multiple mbufs, and all of the mbufs in the chain have the M_PKTHDR flag set. Submitted by: original patch by Ray Ruvinskiy at BlueCoat dot com Reviewed by: jfv, kmacy, rwatson Approved by: re (rwatson) MFC after: 5 days
2011-09-05 17:54:19 +00:00
#define BSIZEPKT_ROUNDUP ((1<<IXGBE_SRRCTL_BSIZEPKT_SHIFT)-1)
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
static void
ixgbe_initialize_receive_units(struct adapter *adapter)
{
struct rx_ring *rxr = adapter->rx_rings;
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
struct ixgbe_hw *hw = &adapter->hw;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
struct ifnet *ifp = adapter->ifp;
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
u32 bufsz, fctrl, srrctl, rxcsum;
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
u32 hlreg;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/*
* Make sure receives are disabled while
* setting up the descriptor ring
*/
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
ixgbe_disable_rx(hw);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/* Enable broadcasts */
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
fctrl |= IXGBE_FCTRL_BAM;
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
fctrl |= IXGBE_FCTRL_DPF;
fctrl |= IXGBE_FCTRL_PMCF;
}
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
/* Set for Jumbo Frames? */
Multiqueue RX is not correctly enabled on the new 82599 adapter, the SRRCTL register needs to be setup per queue. Approved by: re
2009-07-01 16:13:01 +00:00
hlreg = IXGBE_READ_REG(hw, IXGBE_HLREG0);
Update driver to version 2.3.8: CRITICAL FIX - with stats changes the older 82598 will panic and trash the stack on driver load, FCOE registers ONLY exist in 82599 and must not be read otherwise. kern/153951 - to correct incorrect media type on adapters with pluggable modules I have eliminated the old static table in favor of a new dynamic shared code routine. This also has the benefit of detecting changes when a different module is inserted. Performance/enhancement to the Flow Director code from my linux coworker (the developer of the code). Fixes from Michael Tuexen - a data corruption problem on the 82599 (CRITICAL), fix so the buf size correctly adjusts as the cluster changes, and max descriptors are set properly. Also added 16K clusters for those REALLY big jumbos :) In the RX path, the RX LOCK was not being released, and this causes LOR problems. Add the code that igb already has. Sync with in house shared code, this was necessary for the Flow Director fix. MFC in 2 days
2011-01-19 19:36:27 +00:00
if (ifp->if_mtu > ETHERMTU)
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
hlreg |= IXGBE_HLREG0_JUMBOEN;
Update driver to version 2.3.8: CRITICAL FIX - with stats changes the older 82598 will panic and trash the stack on driver load, FCOE registers ONLY exist in 82599 and must not be read otherwise. kern/153951 - to correct incorrect media type on adapters with pluggable modules I have eliminated the old static table in favor of a new dynamic shared code routine. This also has the benefit of detecting changes when a different module is inserted. Performance/enhancement to the Flow Director code from my linux coworker (the developer of the code). Fixes from Michael Tuexen - a data corruption problem on the 82599 (CRITICAL), fix so the buf size correctly adjusts as the cluster changes, and max descriptors are set properly. Also added 16K clusters for those REALLY big jumbos :) In the RX path, the RX LOCK was not being released, and this causes LOR problems. Add the code that igb already has. Sync with in house shared code, this was necessary for the Flow Director fix. MFC in 2 days
2011-01-19 19:36:27 +00:00
else
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
hlreg &= ~IXGBE_HLREG0_JUMBOEN;
Properly disable crc stripping when operating in netmap mode. Contrarily to what i wrote in my previous commit, the 82599 does include the CRC in the length. The operating mode is reset in ixgbe_init_locked() and so we need to hook into the places where the two registers (HLREG0 and RDRXCTL) are modified.
2012-04-13 16:42:54 +00:00
#ifdef DEV_NETMAP
/* crcstrip is conditional in netmap (in RDRXCTL too ?) */
if (ifp->if_capenable & IFCAP_NETMAP && !ix_crcstrip)
hlreg &= ~IXGBE_HLREG0_RXCRCSTRP;
else
hlreg |= IXGBE_HLREG0_RXCRCSTRP;
#endif /* DEV_NETMAP */
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
bufsz = (adapter->rx_mbuf_sz +
BSIZEPKT_ROUNDUP) >> IXGBE_SRRCTL_BSIZEPKT_SHIFT;
Update driver to version 2.3.8: CRITICAL FIX - with stats changes the older 82598 will panic and trash the stack on driver load, FCOE registers ONLY exist in 82599 and must not be read otherwise. kern/153951 - to correct incorrect media type on adapters with pluggable modules I have eliminated the old static table in favor of a new dynamic shared code routine. This also has the benefit of detecting changes when a different module is inserted. Performance/enhancement to the Flow Director code from my linux coworker (the developer of the code). Fixes from Michael Tuexen - a data corruption problem on the 82599 (CRITICAL), fix so the buf size correctly adjusts as the cluster changes, and max descriptors are set properly. Also added 16K clusters for those REALLY big jumbos :) In the RX path, the RX LOCK was not being released, and this causes LOR problems. Add the code that igb already has. Sync with in house shared code, this was necessary for the Flow Director fix. MFC in 2 days
2011-01-19 19:36:27 +00:00
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
for (int i = 0; i < adapter->num_queues; i++, rxr++) {
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
u64 rdba = rxr->rxdma.dma_paddr;
Multiqueue RX is not correctly enabled on the new 82599 adapter, the SRRCTL register needs to be setup per queue. Approved by: re
2009-07-01 16:13:01 +00:00
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/* Setup the Base and Length of the Rx Descriptor Ring */
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
IXGBE_WRITE_REG(hw, IXGBE_RDBAL(i),
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
(rdba & 0x00000000ffffffffULL));
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
IXGBE_WRITE_REG(hw, IXGBE_RDBAH(i), (rdba >> 32));
IXGBE_WRITE_REG(hw, IXGBE_RDLEN(i),
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
adapter->num_rx_desc * sizeof(union ixgbe_adv_rx_desc));
Multiqueue RX is not correctly enabled on the new 82599 adapter, the SRRCTL register needs to be setup per queue. Approved by: re
2009-07-01 16:13:01 +00:00
/* Set up the SRRCTL register */
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
Multiqueue RX is not correctly enabled on the new 82599 adapter, the SRRCTL register needs to be setup per queue. Approved by: re
2009-07-01 16:13:01 +00:00
srrctl &= ~IXGBE_SRRCTL_BSIZEHDR_MASK;
srrctl &= ~IXGBE_SRRCTL_BSIZEPKT_MASK;
srrctl |= bufsz;
First of a series of 11 patches leading to new ixgbe version 2.5.0 This removes the header split and supporting code from the driver.
2012-11-30 22:19:18 +00:00
srrctl |= IXGBE_SRRCTL_DESCTYPE_ADV_ONEBUF;
Set the DROP_EN bit before the RX queue is brought up and active. He noticed issues setting this bit in SRRCTL after the queue was up, so doing it from the sysctl handler isn't enough and may not actually work correctly. This commit doesn't remove the sysctl path or try to change its behaviour. I'll talk with others about how to finish fixing that before I tackle that. PR: kern/194311 Submitted by: luigi MFC after: 3 days Sponsored by: Norse Corp, Inc
2014-10-15 01:22:56 +00:00
/*
* Set DROP_EN iff we have no flow control and >1 queue.
* Note that srrctl was cleared shortly before during reset,
* so we do not need to clear the bit, but do it just in case
* this code is moved elsewhere.
*/
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
if (adapter->num_queues > 1 &&
adapter->hw.fc.requested_mode == ixgbe_fc_none) {
Set the DROP_EN bit before the RX queue is brought up and active. He noticed issues setting this bit in SRRCTL after the queue was up, so doing it from the sysctl handler isn't enough and may not actually work correctly. This commit doesn't remove the sysctl path or try to change its behaviour. I'll talk with others about how to finish fixing that before I tackle that. PR: kern/194311 Submitted by: luigi MFC after: 3 days Sponsored by: Norse Corp, Inc
2014-10-15 01:22:56 +00:00
srrctl |= IXGBE_SRRCTL_DROP_EN;
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
} else {
Set the DROP_EN bit before the RX queue is brought up and active. He noticed issues setting this bit in SRRCTL after the queue was up, so doing it from the sysctl handler isn't enough and may not actually work correctly. This commit doesn't remove the sysctl path or try to change its behaviour. I'll talk with others about how to finish fixing that before I tackle that. PR: kern/194311 Submitted by: luigi MFC after: 3 days Sponsored by: Norse Corp, Inc
2014-10-15 01:22:56 +00:00
srrctl &= ~IXGBE_SRRCTL_DROP_EN;
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
}
Set the DROP_EN bit before the RX queue is brought up and active. He noticed issues setting this bit in SRRCTL after the queue was up, so doing it from the sysctl handler isn't enough and may not actually work correctly. This commit doesn't remove the sysctl path or try to change its behaviour. I'll talk with others about how to finish fixing that before I tackle that. PR: kern/194311 Submitted by: luigi MFC after: 3 days Sponsored by: Norse Corp, Inc
2014-10-15 01:22:56 +00:00
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(i), srrctl);
Multiqueue RX is not correctly enabled on the new 82599 adapter, the SRRCTL register needs to be setup per queue. Approved by: re
2009-07-01 16:13:01 +00:00
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/* Setup the HW Rx Head and Tail Descriptor Pointers */
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
IXGBE_WRITE_REG(hw, IXGBE_RDH(i), 0);
IXGBE_WRITE_REG(hw, IXGBE_RDT(i), 0);
Remove the sysctl process_limit interface, after some thought I've decided its overkill,a simple tuneable for each RX and TX limit, and then init sets the ring values based on that, should be sufficient. More importantly, fix a bug causing a panic, when changing the define style to IXGBE_LEGACY_TX a taskqueue init was inadvertently set #ifdef when it should be #ifndef.
2012-12-03 21:38:02 +00:00
/* Set the processing limit */
rxr->process_limit = ixgbe_rx_process_limit;
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
/* Set the driver rx tail address */
rxr->tail = IXGBE_RDT(rxr->me);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
- Add the RX refresh changes from igb to ixgbe - Also a couple minor tweaks to the TX code from the same source. - Add the INET ioctl code which has been missing from this driver, and which caused IP aliases to reset the interface. - Last, some minor logic changes that just reflect upcoming hardware support, but have no other functional effect now. MFC after a week
2011-04-25 23:34:21 +00:00
if (adapter->hw.mac.type != ixgbe_mac_82598EB) {
A few changes: When not defining header split do not allocate mbufs, this can be a BIG savings in the mbuf memory pool. Also keep seperate dma maps for the header and payload pieces when doing header split. The basis of this code was a patch done a while ago by yongari, thank you :) A number of white space changes. MFC: in a few days
2010-05-19 00:03:48 +00:00
u32 psrtype = IXGBE_PSRTYPE_TCPHDR |
IXGBE_PSRTYPE_UDPHDR |
IXGBE_PSRTYPE_IPV4HDR |
IXGBE_PSRTYPE_IPV6HDR;
IXGBE_WRITE_REG(hw, IXGBE_PSRTYPE(0), psrtype);
}
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
rxcsum = IXGBE_READ_REG(hw, IXGBE_RXCSUM);
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
ixgbe_initialise_rss_mapping(adapter);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Add initial RSS awareness to the ixgbe(4) driver. The ixgbe(4) hardware is capable of RSS hashing RX packets and doing RSS queue selection for up to 8 queues. However, even if multi-queue is enabled for ixgbe(4), the RX path doesn't use the RSS flowid from the received descriptor. It just uses the MSIX queue id. This patch does a handful of things if RSS is enabled: * Instead of using a random key at boot, fetch the RSS key from the RSS code and program that in to the RSS redirection table. That whole chunk of code should be double checked for endian correctness. * Use the RSS queue mapping to CPU ID to figure out where to thread pin the RX swi thread and the taskqueue threads for each queue. * The software queue is now really an "RSS bucket". * When programming the RSS indirection table, use the RSS code to figure out which RSS bucket each slot in the indirection table maps to. * When transmitting, use the flowid RSS mapping if the mbuf has an RSS aware hash. The existing method wasn't guaranteed to align correctly with the destination RSS bucket (and thus CPU ID.) This code warns if the number of RSS buckets isn't the same as the automatically configured number of hardware queues. The administrator will have to tweak one of them for better performance. There's currently no way to re-balance the RSS indirection table after startup. I'll worry about that later. Additionally, it may be worthwhile to always use the full 32 bit flowid if multi-queue is enabled. It'll make things like lagg(4) behave better with respect to traffic distribution.
2014-06-30 04:38:29 +00:00
if (adapter->num_queues > 1) {
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/* RSS and RX IPP Checksum are mutually exclusive */
rxcsum |= IXGBE_RXCSUM_PCSD;
}
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
if (ifp->if_capenable & IFCAP_RXCSUM)
rxcsum |= IXGBE_RXCSUM_PCSD;
if (!(rxcsum & IXGBE_RXCSUM_PCSD))
rxcsum |= IXGBE_RXCSUM_IPPCSE;
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
IXGBE_WRITE_REG(hw, IXGBE_RXCSUM, rxcsum);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
return;
}
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
This updates the ixgbe driver to Intel internal version 1.4.7 Shared code changes, core driver fixes, vlan event/filter support Also Kip Macy's fix to allow any number of queues, thanks Kip!
2008-07-30 18:15:18 +00:00
/*
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
** This routine is run via an vlan config EVENT,
** it enables us to use the HW Filter table since
** we can get the vlan id. This just creates the
** entry in the soft version of the VFTA, init will
** repopulate the real table.
*/
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
static void
Improvement on the last change, this gives a precise way to tell the one and only interface that a vlan event is for. Thanks to John Baldwin for the patch. Approved by: re
2009-07-24 21:35:52 +00:00
ixgbe_register_vlan(void *arg, struct ifnet *ifp, u16 vtag)
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
{
This updates the ixgbe driver to Intel internal version 1.4.7 Shared code changes, core driver fixes, vlan event/filter support Also Kip Macy's fix to allow any number of queues, thanks Kip!
2008-07-30 18:15:18 +00:00
struct adapter *adapter = ifp->if_softc;
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
u16 index, bit;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Improvement on the last change, this gives a precise way to tell the one and only interface that a vlan event is for. Thanks to John Baldwin for the patch. Approved by: re
2009-07-24 21:35:52 +00:00
if (ifp->if_softc != arg) /* Not our event */
This delta fixes two bugs: - When a vlan event occurs a check was not made that the event was actually for the interface, thus resulting in a panic. All three drivers have this vulnerability. Add a check for this condition. - Secondly, there was a duplicate buf_ring free in the em driver resulting in a panic on unload. Remove. Approved by: re
2009-07-24 16:57:49 +00:00
return;
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
if ((vtag == 0) || (vtag > 4095)) /* Invalid */
return;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
IXGBE_CORE_LOCK(adapter);
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
index = (vtag >> 5) & 0x7F;
bit = vtag & 0x1F;
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
adapter->shadow_vfta[index] |= (1 << bit);
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
++adapter->num_vlans;
- Fix link loss on vlan reconfiguration. - Fix issues with 'vlanhwfilter'. MFC after: 1 week Silence from: jfv 5 weeks
2013-11-05 09:46:01 +00:00
ixgbe_setup_vlan_hw_support(adapter);
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
IXGBE_CORE_UNLOCK(adapter);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
This updates the ixgbe driver to Intel internal version 1.4.7 Shared code changes, core driver fixes, vlan event/filter support Also Kip Macy's fix to allow any number of queues, thanks Kip!
2008-07-30 18:15:18 +00:00
/*
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
** This routine is run via an vlan
** unconfig EVENT, remove our entry
** in the soft vfta.
*/
This updates the ixgbe driver to Intel internal version 1.4.7 Shared code changes, core driver fixes, vlan event/filter support Also Kip Macy's fix to allow any number of queues, thanks Kip!
2008-07-30 18:15:18 +00:00
static void
Improvement on the last change, this gives a precise way to tell the one and only interface that a vlan event is for. Thanks to John Baldwin for the patch. Approved by: re
2009-07-24 21:35:52 +00:00
ixgbe_unregister_vlan(void *arg, struct ifnet *ifp, u16 vtag)
This updates the ixgbe driver to Intel internal version 1.4.7 Shared code changes, core driver fixes, vlan event/filter support Also Kip Macy's fix to allow any number of queues, thanks Kip!
2008-07-30 18:15:18 +00:00
{
struct adapter *adapter = ifp->if_softc;
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
u16 index, bit;
This updates the ixgbe driver to Intel internal version 1.4.7 Shared code changes, core driver fixes, vlan event/filter support Also Kip Macy's fix to allow any number of queues, thanks Kip!
2008-07-30 18:15:18 +00:00
Improvement on the last change, this gives a precise way to tell the one and only interface that a vlan event is for. Thanks to John Baldwin for the patch. Approved by: re
2009-07-24 21:35:52 +00:00
if (ifp->if_softc != arg)
This delta fixes two bugs: - When a vlan event occurs a check was not made that the event was actually for the interface, thus resulting in a panic. All three drivers have this vulnerability. Add a check for this condition. - Secondly, there was a duplicate buf_ring free in the em driver resulting in a panic on unload. Remove. Approved by: re
2009-07-24 16:57:49 +00:00
return;
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
if ((vtag == 0) || (vtag > 4095)) /* Invalid */
return;
This updates the ixgbe driver to Intel internal version 1.4.7 Shared code changes, core driver fixes, vlan event/filter support Also Kip Macy's fix to allow any number of queues, thanks Kip!
2008-07-30 18:15:18 +00:00
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
IXGBE_CORE_LOCK(adapter);
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
index = (vtag >> 5) & 0x7F;
bit = vtag & 0x1F;
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
adapter->shadow_vfta[index] &= ~(1 << bit);
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
--adapter->num_vlans;
/* Re-init to load the changes */
- Fix link loss on vlan reconfiguration. - Fix issues with 'vlanhwfilter'. MFC after: 1 week Silence from: jfv 5 weeks
2013-11-05 09:46:01 +00:00
ixgbe_setup_vlan_hw_support(adapter);
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
IXGBE_CORE_UNLOCK(adapter);
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
}
static void
ixgbe_setup_vlan_hw_support(struct adapter *adapter)
{
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
struct ifnet *ifp = adapter->ifp;
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
struct ixgbe_hw *hw = &adapter->hw;
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
struct rx_ring *rxr;
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
u32 ctrl;
/*
** We get here thru init_locked, meaning
** a soft reset, this has already cleared
** the VFTA and other state, so if there
** have been no vlan's registered do nothing.
*/
if (adapter->num_vlans == 0)
return;
- Fix link loss on vlan reconfiguration. - Fix issues with 'vlanhwfilter'. MFC after: 1 week Silence from: jfv 5 weeks
2013-11-05 09:46:01 +00:00
/* Setup the queues for vlans */
for (int i = 0; i < adapter->num_queues; i++) {
rxr = &adapter->rx_rings[i];
/* On 82599 the VLAN enable is per/queue in RXDCTL */
if (hw->mac.type != ixgbe_mac_82598EB) {
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
- Fix link loss on vlan reconfiguration. - Fix issues with 'vlanhwfilter'. MFC after: 1 week Silence from: jfv 5 weeks
2013-11-05 09:46:01 +00:00
ctrl |= IXGBE_RXDCTL_VME;
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(i), ctrl);
- Fix link loss on vlan reconfiguration. - Fix issues with 'vlanhwfilter'. MFC after: 1 week Silence from: jfv 5 weeks
2013-11-05 09:46:01 +00:00
}
rxr->vtag_strip = TRUE;
}
if ((ifp->if_capenable & IFCAP_VLAN_HWFILTER) == 0)
return;
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
/*
** A soft reset zero's out the VFTA, so
** we need to repopulate it now.
*/
for (int i = 0; i < IXGBE_VFTA_SIZE; i++)
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
if (adapter->shadow_vfta[i] != 0)
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
IXGBE_WRITE_REG(hw, IXGBE_VFTA(i),
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
adapter->shadow_vfta[i]);
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
Update ixgbe driver to verion 2.3.6 - This adds a VM SRIOV interface, ixv, it is however transparent to the user, it links with the ixgbe.ko, but when ixgbe is loaded in a virtualized guest with SRIOV configured this will be detected. - Sync shared code to latest - Many bug fixes and improvements, thanks to everyone who has been using the driver and reporting issues.
2010-11-26 22:46:32 +00:00
/* Enable the Filter Table if enabled */
if (ifp->if_capenable & IFCAP_VLAN_HWFILTER) {
ctrl &= ~IXGBE_VLNCTRL_CFIEN;
ctrl |= IXGBE_VLNCTRL_VFE;
}
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
if (hw->mac.type == ixgbe_mac_82598EB)
ctrl |= IXGBE_VLNCTRL_VME;
IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
This updates the ixgbe driver to Intel internal version 1.4.7 Shared code changes, core driver fixes, vlan event/filter support Also Kip Macy's fix to allow any number of queues, thanks Kip!
2008-07-30 18:15:18 +00:00
}
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
static void
ixgbe_enable_intr(struct adapter *adapter)
{
Two small fixes: Set promiscuous code was unconditionally turning off multicast when turning off promiscuous mode, this should only be done when there are less than MAX groups. Thanks to Mike Karels for this correction. Second, the overtmp interrupt setup/detection was wrong, correcting it. MFC after: one week
2013-03-29 18:03:00 +00:00
struct ixgbe_hw *hw = &adapter->hw;
struct ix_queue *que = adapter->queues;
u32 mask, fwsm;
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
Two small fixes: Set promiscuous code was unconditionally turning off multicast when turning off promiscuous mode, this should only be done when there are less than MAX groups. Thanks to Mike Karels for this correction. Second, the overtmp interrupt setup/detection was wrong, correcting it. MFC after: one week
2013-03-29 18:03:00 +00:00
mask = (IXGBE_EIMS_ENABLE_MASK & ~IXGBE_EIMS_RTX_QUEUE);
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
/* Enable Fan Failure detection */
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
if (hw->device_id == IXGBE_DEV_ID_82598AT)
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
mask |= IXGBE_EIMS_GPI_SDP1;
First, sync to internal shared code, and then Fixes: - flow control - don't override user value on re-init - fix to make 1G optics work correctly - change to interrupt enabling - some bits were incorrect for certain hardware. - certain stats fixes, remove a duplicate increment of ierror, thanks to Scott Long for pointing these out. - shared code link interface changed, requiring some core code changes to accomodate this. - add an m_adj() to ETHER_ALIGN on the recieve side, this was requested by Mike Karels, thanks Mike. - Multicast code corrections also thanks to Mike Karels.
2013-03-04 23:07:40 +00:00
switch (adapter->hw.mac.type) {
case ixgbe_mac_82599EB:
mask |= IXGBE_EIMS_ECC;
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
/* Temperature sensor on some adapters */
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
mask |= IXGBE_EIMS_GPI_SDP0;
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
/* SFP+ (RX_LOS_N & MOD_ABS_N) */
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
mask |= IXGBE_EIMS_GPI_SDP1;
mask |= IXGBE_EIMS_GPI_SDP2;
First, sync to internal shared code, and then Fixes: - flow control - don't override user value on re-init - fix to make 1G optics work correctly - change to interrupt enabling - some bits were incorrect for certain hardware. - certain stats fixes, remove a duplicate increment of ierror, thanks to Scott Long for pointing these out. - shared code link interface changed, requiring some core code changes to accomodate this. - add an m_adj() to ETHER_ALIGN on the recieve side, this was requested by Mike Karels, thanks Mike. - Multicast code corrections also thanks to Mike Karels.
2013-03-04 23:07:40 +00:00
#ifdef IXGBE_FDIR
mask |= IXGBE_EIMS_FLOW_DIR;
#endif
break;
case ixgbe_mac_X540:
Two small fixes: Set promiscuous code was unconditionally turning off multicast when turning off promiscuous mode, this should only be done when there are less than MAX groups. Thanks to Mike Karels for this correction. Second, the overtmp interrupt setup/detection was wrong, correcting it. MFC after: one week
2013-03-29 18:03:00 +00:00
/* Detect if Thermal Sensor is enabled */
fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM);
if (fwsm & IXGBE_FWSM_TS_ENABLED)
mask |= IXGBE_EIMS_TS;
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
mask |= IXGBE_EIMS_ECC;
#ifdef IXGBE_FDIR
mask |= IXGBE_EIMS_FLOW_DIR;
#endif
break;
case ixgbe_mac_X550:
case ixgbe_mac_X550EM_x:
/* MAC thermal sensor is automatically enabled */
mask |= IXGBE_EIMS_TS;
/* Some devices use SDP0 for important information */
if (hw->device_id == IXGBE_DEV_ID_X550EM_X_SFP ||
hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T)
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
mask |= IXGBE_EIMS_GPI_SDP0_BY_MAC(hw);
mask |= IXGBE_EIMS_ECC;
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
#ifdef IXGBE_FDIR
First, sync to internal shared code, and then Fixes: - flow control - don't override user value on re-init - fix to make 1G optics work correctly - change to interrupt enabling - some bits were incorrect for certain hardware. - certain stats fixes, remove a duplicate increment of ierror, thanks to Scott Long for pointing these out. - shared code link interface changed, requiring some core code changes to accomodate this. - add an m_adj() to ETHER_ALIGN on the recieve side, this was requested by Mike Karels, thanks Mike. - Multicast code corrections also thanks to Mike Karels.
2013-03-04 23:07:40 +00:00
mask |= IXGBE_EIMS_FLOW_DIR;
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
#endif
First, sync to internal shared code, and then Fixes: - flow control - don't override user value on re-init - fix to make 1G optics work correctly - change to interrupt enabling - some bits were incorrect for certain hardware. - certain stats fixes, remove a duplicate increment of ierror, thanks to Scott Long for pointing these out. - shared code link interface changed, requiring some core code changes to accomodate this. - add an m_adj() to ETHER_ALIGN on the recieve side, this was requested by Mike Karels, thanks Mike. - Multicast code corrections also thanks to Mike Karels.
2013-03-04 23:07:40 +00:00
/* falls through */
default:
break;
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
}
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
IXGBE_WRITE_REG(hw, IXGBE_EIMS, mask);
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
/* With MSI-X we use auto clear */
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
if (adapter->msix_mem) {
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
mask = IXGBE_EIMS_ENABLE_MASK;
A couple fixes got clobbered, putting them back.
2010-11-26 23:57:13 +00:00
/* Don't autoclear Link */
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
mask &= ~IXGBE_EIMS_OTHER;
mask &= ~IXGBE_EIMS_LSC;
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
IXGBE_WRITE_REG(hw, IXGBE_EIAC, mask);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
/*
A couple fixes got clobbered, putting them back.
2010-11-26 23:57:13 +00:00
** Now enable all queues, this is done separately to
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
** allow for handling the extended (beyond 32) MSIX
** vectors that can be used by 82599
*/
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
for (int i = 0; i < adapter->num_queues; i++, que++)
ixgbe_enable_queue(adapter, que->msix);
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
IXGBE_WRITE_FLUSH(hw);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
return;
}
static void
ixgbe_disable_intr(struct adapter *adapter)
{
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
if (adapter->msix_mem)
IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIAC, 0);
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
if (adapter->hw.mac.type == ixgbe_mac_82598EB) {
IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, ~0);
} else {
IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC, 0xFFFF0000);
IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(0), ~0);
IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMC_EX(1), ~0);
}
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
IXGBE_WRITE_FLUSH(&adapter->hw);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
return;
}
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
/*
** Get the width and transaction speed of
** the slot this adapter is plugged into.
*/
static void
ixgbe_get_slot_info(struct ixgbe_hw *hw)
{
device_t dev = ((struct ixgbe_osdep *)hw->back)->dev;
struct ixgbe_mac_info *mac = &hw->mac;
u16 link;
u32 offset;
/* For most devices simply call the shared code routine */
if (hw->device_id != IXGBE_DEV_ID_82599_SFP_SF_QP) {
ixgbe_get_bus_info(hw);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
/* These devices don't use PCI-E */
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
switch (hw->mac.type) {
case ixgbe_mac_X550EM_x:
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
return;
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
default:
goto display;
}
Add quad port probe support, this gives the admin proper information about the slot (which should be a PCIE Gen 3 slot for this adapter) by looking back thru the PCI parent devices to the slot device. The fix above also corrects the bandwidth display to GT/s rather than the incorrect Gb/s Next, allow the use of ALTQ if you select the compile option IXGBE_LEGACY_TX. Allow the use of 'unsupported' optic modules by a compile option as well. Add a phy reset capability into the stop code, this is so a static configured driver will still behave properly when taken down (not being able to unload it). This revision synchronizes the shared code with Intel internal current code, and note that it now includes DCB supporting code, this was necessitated by some internal changes with the code, but it also will provide the opportunity to develop this feature in the core driver down the road. I have edited the README to get rid of some of the worse anachronisms in it as well, its by no means as robust as I might wish at this point however. Oh, I also have included some conditional stuff in the code so it will be compatible in both the 9.X and 10 environments. Performance has been a focus in recent changes and I believe this revision driver will perform very well in most workloads. MFC after: 2 weeks
2013-06-18 21:28:19 +00:00
}
/*
** For the Quad port adapter we need to parse back
** up the PCI tree to find the speed of the expansion
** slot into which this adapter is plugged. A bit more work.
*/
dev = device_get_parent(device_get_parent(dev));
#ifdef IXGBE_DEBUG
device_printf(dev, "parent pcib = %x,%x,%x\n",
pci_get_bus(dev), pci_get_slot(dev), pci_get_function(dev));
#endif
dev = device_get_parent(device_get_parent(dev));
#ifdef IXGBE_DEBUG
device_printf(dev, "slot pcib = %x,%x,%x\n",
pci_get_bus(dev), pci_get_slot(dev), pci_get_function(dev));
#endif
/* Now get the PCI Express Capabilities offset */
pci_find_cap(dev, PCIY_EXPRESS, &offset);
/* ...and read the Link Status Register */
link = pci_read_config(dev, offset + PCIER_LINK_STA, 2);
switch (link & IXGBE_PCI_LINK_WIDTH) {
case IXGBE_PCI_LINK_WIDTH_1:
hw->bus.width = ixgbe_bus_width_pcie_x1;
break;
case IXGBE_PCI_LINK_WIDTH_2:
hw->bus.width = ixgbe_bus_width_pcie_x2;
break;
case IXGBE_PCI_LINK_WIDTH_4:
hw->bus.width = ixgbe_bus_width_pcie_x4;
break;
case IXGBE_PCI_LINK_WIDTH_8:
hw->bus.width = ixgbe_bus_width_pcie_x8;
break;
default:
hw->bus.width = ixgbe_bus_width_unknown;
break;
}
switch (link & IXGBE_PCI_LINK_SPEED) {
case IXGBE_PCI_LINK_SPEED_2500:
hw->bus.speed = ixgbe_bus_speed_2500;
break;
case IXGBE_PCI_LINK_SPEED_5000:
hw->bus.speed = ixgbe_bus_speed_5000;
break;
case IXGBE_PCI_LINK_SPEED_8000:
hw->bus.speed = ixgbe_bus_speed_8000;
break;
default:
hw->bus.speed = ixgbe_bus_speed_unknown;
break;
}
mac->ops.set_lan_id(hw);
display:
device_printf(dev,"PCI Express Bus: Speed %s %s\n",
((hw->bus.speed == ixgbe_bus_speed_8000) ? "8.0GT/s":
(hw->bus.speed == ixgbe_bus_speed_5000) ? "5.0GT/s":
(hw->bus.speed == ixgbe_bus_speed_2500) ? "2.5GT/s":"Unknown"),
(hw->bus.width == ixgbe_bus_width_pcie_x8) ? "Width x8" :
(hw->bus.width == ixgbe_bus_width_pcie_x4) ? "Width x4" :
(hw->bus.width == ixgbe_bus_width_pcie_x1) ? "Width x1" :
("Unknown"));
if ((hw->device_id != IXGBE_DEV_ID_82599_SFP_SF_QP) &&
((hw->bus.width <= ixgbe_bus_width_pcie_x4) &&
(hw->bus.speed == ixgbe_bus_speed_2500))) {
device_printf(dev, "PCI-Express bandwidth available"
" for this card\n is not sufficient for"
" optimal performance.\n");
device_printf(dev, "For optimal performance a x8 "
"PCIE, or x4 PCIE Gen2 slot is required.\n");
}
if ((hw->device_id == IXGBE_DEV_ID_82599_SFP_SF_QP) &&
((hw->bus.width <= ixgbe_bus_width_pcie_x8) &&
(hw->bus.speed < ixgbe_bus_speed_8000))) {
device_printf(dev, "PCI-Express bandwidth available"
" for this card\n is not sufficient for"
" optimal performance.\n");
device_printf(dev, "For optimal performance a x8 "
"PCIE Gen3 slot is required.\n");
}
return;
}
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
/*
** Setup the correct IVAR register for a particular MSIX interrupt
** (yes this is all very magic and confusing :)
** - entry is the register array entry
** - vector is the MSIX vector for this queue
** - type is RX/TX/MISC
*/
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
static void
Update for the Intel 10G driver, this adds support for newest hardware, adds multiqueue tx interface, infrastructure cleanup to allow up to 32 MSIX vectors on newer Nehalem systems. Bug fixes, etc.
2009-06-24 18:27:07 +00:00
ixgbe_set_ivar(struct adapter *adapter, u8 entry, u8 vector, s8 type)
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
{
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
struct ixgbe_hw *hw = &adapter->hw;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
u32 ivar, index;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
vector |= IXGBE_IVAR_ALLOC_VAL;
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
switch (hw->mac.type) {
case ixgbe_mac_82598EB:
if (type == -1)
entry = IXGBE_IVAR_OTHER_CAUSES_INDEX;
else
entry += (type * 64);
index = (entry >> 2) & 0x1F;
ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(index));
ivar &= ~(0xFF << (8 * (entry & 0x3)));
ivar |= (vector << (8 * (entry & 0x3)));
IXGBE_WRITE_REG(&adapter->hw, IXGBE_IVAR(index), ivar);
break;
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
case ixgbe_mac_82599EB:
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
case ixgbe_mac_X540:
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
case ixgbe_mac_X550:
case ixgbe_mac_X550EM_x:
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
if (type == -1) { /* MISC IVAR */
index = (entry & 1) * 8;
ivar = IXGBE_READ_REG(hw, IXGBE_IVAR_MISC);
ivar &= ~(0xFF << index);
ivar |= (vector << index);
IXGBE_WRITE_REG(hw, IXGBE_IVAR_MISC, ivar);
} else { /* RX/TX IVARS */
index = (16 * (entry & 1)) + (8 * type);
ivar = IXGBE_READ_REG(hw, IXGBE_IVAR(entry >> 1));
ivar &= ~(0xFF << index);
ivar |= (vector << index);
IXGBE_WRITE_REG(hw, IXGBE_IVAR(entry >> 1), ivar);
}
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
default:
break;
}
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
static void
ixgbe_configure_ivars(struct adapter *adapter)
{
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
struct ix_queue *que = adapter->queues;
u32 newitr;
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
if (ixgbe_max_interrupt_rate > 0)
ixgbe changes: - remove experimental code for disabling CRC - use the correct constant for conversion between interrupt rate and EITR values (the previous values were off by a factor of 2) - make dev.ix.N.queueM.interrupt_rate a RW sysctl variable. Changing individual values affects the queue immediately, and propagates to all interfaces at the next reinit. - add dev.ix.N.queueM.irqs rdonly sysctl, to export the actual interrupt counts Netmap-related changes for ixgbe: - use the "new" format for TX descriptors in netmap mode. - pass interrupt mitigation delays to the user process doing poll() on a netmap file descriptor. On the RX side this means we will not check the ring more than once per interrupt. This gives the process a chance to sleep and process packets in larger batches, thus reducing CPU usage. On the TX side we take this even further: completed transmissions are reclaimed every half ring even if the NIC interrupts more often. This saves even more CPU without any additional tx delays. Generic Netmap-related changes: - align the netmap_kring to cache lines so that there is no false sharing (possibly useful for multiqueue NICs and MSIX interrupts, which are handled by different cores). It's a minor improvement but it does not cost anything. Reviewed by: Jack Vogel Approved by: Jack Vogel
2012-01-26 09:55:16 +00:00
newitr = (4000000 / ixgbe_max_interrupt_rate) & 0x0FF8;
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
else {
/*
** Disable DMA coalescing if interrupt moderation is
** disabled.
*/
adapter->dmac = 0;
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
newitr = 0;
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
}
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
for (int i = 0; i < adapter->num_queues; i++, que++) {
/* First the RX queue entry */
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
ixgbe_set_ivar(adapter, i, que->msix, 0);
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
/* ... and the TX */
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
ixgbe_set_ivar(adapter, i, que->msix, 1);
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
/* Set an Initial EITR value */
IXGBE_WRITE_REG(&adapter->hw,
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
IXGBE_EITR(que->msix), newitr);
Update the driver to Intel version 2.1.6 - add some new hardware support for 82599 - Big change to interrupt architecture, it now uses a queue which contains an RX/TX pair as the recipient of the interrupt. This will reduce overall system interrupts/msix usage. - Improved RX mbuf handling: the old get_buf routine is no longer synchronized with rxeof, this allows the elimination of packet discards due to mbuf allocation failure. - Much simplified and improved AIM code, it now happens in the queue interrupt context and takes into account both the traffic on the RX AND TX side. - variety of small tweaks, like ring size, that have been seen as performance improvements. - Thanks to those that provided feedback or suggested changes, I hope I've caught all of them.
2010-03-27 00:21:40 +00:00
}
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/* For the Link interrupt */
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
ixgbe_set_ivar(adapter, 1, adapter->vector, -1);
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
}
/*
** ixgbe_sfp_probe - called in the local timer to
** determine if a port had optics inserted.
*/
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
static bool ixgbe_sfp_probe(struct adapter *adapter)
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
{
struct ixgbe_hw *hw = &adapter->hw;
device_t dev = adapter->dev;
bool result = FALSE;
if ((hw->phy.type == ixgbe_phy_nl) &&
(hw->phy.sfp_type == ixgbe_sfp_type_not_present)) {
s32 ret = hw->phy.ops.identify_sfp(hw);
if (ret)
goto out;
ret = hw->phy.ops.reset(hw);
if (ret == IXGBE_ERR_SFP_NOT_SUPPORTED) {
device_printf(dev,"Unsupported SFP+ module detected!");
printf(" Reload driver with supported module.\n");
adapter->sfp_probe = FALSE;
goto out;
} else
device_printf(dev,"SFP+ module detected!\n");
/* We now have supported optics */
adapter->sfp_probe = FALSE;
Update driver to version 2.3.8: CRITICAL FIX - with stats changes the older 82598 will panic and trash the stack on driver load, FCOE registers ONLY exist in 82599 and must not be read otherwise. kern/153951 - to correct incorrect media type on adapters with pluggable modules I have eliminated the old static table in favor of a new dynamic shared code routine. This also has the benefit of detecting changes when a different module is inserted. Performance/enhancement to the Flow Director code from my linux coworker (the developer of the code). Fixes from Michael Tuexen - a data corruption problem on the 82599 (CRITICAL), fix so the buf size correctly adjusts as the cluster changes, and max descriptors are set properly. Also added 16K clusters for those REALLY big jumbos :) In the RX path, the RX LOCK was not being released, and this causes LOR problems. Add the code that igb already has. Sync with in house shared code, this was necessary for the Flow Director fix. MFC in 2 days
2011-01-19 19:36:27 +00:00
/* Set the optics type so system reports correctly */
ixgbe_setup_optics(adapter);
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
result = TRUE;
}
out:
return (result);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
/*
** Tasklet handler for MSIX Link interrupts
** - do outside interrupt since it might sleep
*/
static void
ixgbe_handle_link(void *context, int pending)
{
struct adapter *adapter = context;
ixgbe_check_link(&adapter->hw,
&adapter->link_speed, &adapter->link_up, 0);
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
ixgbe_update_link_status(adapter);
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
}
/*
** Tasklet for handling SFP module interrupts
*/
static void
ixgbe_handle_mod(void *context, int pending)
{
struct adapter *adapter = context;
struct ixgbe_hw *hw = &adapter->hw;
device_t dev = adapter->dev;
u32 err;
err = hw->phy.ops.identify_sfp(hw);
if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
device_printf(dev,
"Unsupported SFP+ module type was detected.\n");
return;
}
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
err = hw->mac.ops.setup_sfp(hw);
if (err == IXGBE_ERR_SFP_NOT_SUPPORTED) {
device_printf(dev,
"Setup failure - unsupported SFP+ module type.\n");
return;
}
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
taskqueue_enqueue(adapter->tq, &adapter->msf_task);
return;
}
/*
** Tasklet for handling MSF (multispeed fiber) interrupts
*/
static void
ixgbe_handle_msf(void *context, int pending)
{
struct adapter *adapter = context;
struct ixgbe_hw *hw = &adapter->hw;
u32 autoneg;
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
bool negotiate;
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
int err;
err = hw->phy.ops.identify_sfp(hw);
if (!err) {
ixgbe_setup_optics(adapter);
INIT_DEBUGOUT1("ixgbe_sfp_probe: flags: %X\n", adapter->optics);
}
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
autoneg = hw->phy.autoneg_advertised;
if ((!autoneg) && (hw->mac.ops.get_link_capabilities))
hw->mac.ops.get_link_capabilities(hw, &autoneg, &negotiate);
if (hw->mac.ops.setup_link)
First, sync to internal shared code, and then Fixes: - flow control - don't override user value on re-init - fix to make 1G optics work correctly - change to interrupt enabling - some bits were incorrect for certain hardware. - certain stats fixes, remove a duplicate increment of ierror, thanks to Scott Long for pointing these out. - shared code link interface changed, requiring some core code changes to accomodate this. - add an m_adj() to ETHER_ALIGN on the recieve side, this was requested by Mike Karels, thanks Mike. - Multicast code corrections also thanks to Mike Karels.
2013-03-04 23:07:40 +00:00
hw->mac.ops.setup_link(hw, autoneg, TRUE);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
ifmedia_removeall(&adapter->media);
ixgbe_add_media_types(adapter);
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
return;
}
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
/*
** Tasklet for handling interrupts from an external PHY
*/
static void
ixgbe_handle_phy(void *context, int pending)
{
struct adapter *adapter = context;
struct ixgbe_hw *hw = &adapter->hw;
int error;
error = hw->phy.ops.handle_lasi(hw);
if (error == IXGBE_ERR_OVERTEMP)
device_printf(adapter->dev,
"CRITICAL: EXTERNAL PHY OVER TEMP!! "
" PHY will downshift to lower power state!\n");
else if (error)
device_printf(adapter->dev,
"Error handling LASI interrupt: %d\n",
error);
return;
}
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
#ifdef IXGBE_FDIR
/*
** Tasklet for reinitializing the Flow Director filter table
*/
static void
ixgbe_reinit_fdir(void *context, int pending)
{
struct adapter *adapter = context;
Remove extraneous semicolons, no functional changes. Submitted by: Marc Balmer <marc@msys.ch> MFC after: 1 week
2010-01-07 21:01:37 +00:00
struct ifnet *ifp = adapter->ifp;
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
if (adapter->fdir_reinit != 1) /* Shouldn't happen */
return;
ixgbe_reinit_fdir_tables_82599(&adapter->hw);
adapter->fdir_reinit = 0;
Update to the ixgbe driver: - Add a couple of new devices - Flow control changes in shared and core code - Bug fix to Flow Director for 82598 - Shared code sync to internal with required core change Thanks to those helping in the testing and improvements to this driver! MFC after:5 days
2012-07-05 20:51:44 +00:00
/* re-enable flow director interrupts */
IXGBE_WRITE_REG(&adapter->hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
/* Restart the interface */
ifp->if_drv_flags |= IFF_DRV_RUNNING;
Update the ixgbe driver to 1.7.4, this includes support for the new 82599 adapter family, adds header split, and many small fixes. The driver should now be added to the GENERIC kernel. MFC after: 2 weeks
2009-04-10 00:22:48 +00:00
return;
}
Update driver to Intel version 2.0.7: This adds new feature support for the 82599, a hardware assist to LRO, doing this required a large revamp to the RX cleanup code because the descriptor ring may not be processed out of order, this necessitated the elimination of global pointers. Additionally, the RX routine now does not refresh mbufs on every descriptor, rather it will do a range, and then update the hardware pointer at that time. These are performance oriented changes. The TX side now has a cleaner simpler watchdog algorithm as well, in TX cleanup a read of ticks is stored, that can then be compared in local_timer to determine if there is a hang. Various other cleanups along the way, thanks to all who have provided input and testing.
2009-12-07 21:30:54 +00:00
#endif
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
/*********************************************************************
*
* Configure DMA Coalescing
*
**********************************************************************/
static void
ixgbe_config_dmac(struct adapter *adapter)
{
struct ixgbe_hw *hw = &adapter->hw;
struct ixgbe_dmac_config *dcfg = &hw->mac.dmac_config;
if (hw->mac.type < ixgbe_mac_X550 ||
!hw->mac.ops.dmac_config)
return;
if (dcfg->watchdog_timer ^ adapter->dmac ||
dcfg->link_speed ^ adapter->link_speed) {
dcfg->watchdog_timer = adapter->dmac;
dcfg->fcoe_en = false;
dcfg->link_speed = adapter->link_speed;
dcfg->num_tcs = 1;
INIT_DEBUGOUT2("dmac settings: watchdog %d, link speed %d\n",
dcfg->watchdog_timer, dcfg->link_speed);
hw->mac.ops.dmac_config(hw);
}
}
/*
* Checks whether the adapter supports Energy Efficient Ethernet
* or not, based on device ID.
*/
static void
ixgbe_check_eee_support(struct adapter *adapter)
{
struct ixgbe_hw *hw = &adapter->hw;
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
adapter->eee_support = adapter->eee_enabled =
(hw->device_id == IXGBE_DEV_ID_X550T ||
hw->device_id == IXGBE_DEV_ID_X550EM_X_KR);
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
}
/*
* Checks whether the adapter's ports are capable of
* Wake On LAN by reading the adapter's NVM.
*
* Sets each port's hw->wol_enabled value depending
* on the value read here.
*/
static void
ixgbe_check_wol_support(struct adapter *adapter)
{
struct ixgbe_hw *hw = &adapter->hw;
u16 dev_caps = 0;
/* Find out WoL support for port */
adapter->wol_support = hw->wol_enabled = 0;
ixgbe_get_device_caps(hw, &dev_caps);
if ((dev_caps & IXGBE_DEVICE_CAPS_WOL_PORT0_1) ||
((dev_caps & IXGBE_DEVICE_CAPS_WOL_PORT0) &&
hw->bus.func == 0))
adapter->wol_support = hw->wol_enabled = 1;
/* Save initial wake up filter configuration */
adapter->wufc = IXGBE_READ_REG(hw, IXGBE_WUFC);
return;
}
/*
* Prepare the adapter/port for LPLU and/or WoL
*/
static int
ixgbe_setup_low_power_mode(struct adapter *adapter)
{
struct ixgbe_hw *hw = &adapter->hw;
device_t dev = adapter->dev;
s32 error = 0;
mtx_assert(&adapter->core_mtx, MA_OWNED);
/* Limit power management flow to X550EM baseT */
if (hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T
&& hw->phy.ops.enter_lplu) {
/* Turn off support for APM wakeup. (Using ACPI instead) */
IXGBE_WRITE_REG(hw, IXGBE_GRC,
IXGBE_READ_REG(hw, IXGBE_GRC) & ~(u32)2);
/*
* Clear Wake Up Status register to prevent any previous wakeup
* events from waking us up immediately after we suspend.
*/
IXGBE_WRITE_REG(hw, IXGBE_WUS, 0xffffffff);
/*
* Program the Wakeup Filter Control register with user filter
* settings
*/
IXGBE_WRITE_REG(hw, IXGBE_WUFC, adapter->wufc);
/* Enable wakeups and power management in Wakeup Control */
IXGBE_WRITE_REG(hw, IXGBE_WUC,
IXGBE_WUC_WKEN | IXGBE_WUC_PME_EN);
/* X550EM baseT adapters need a special LPLU flow */
hw->phy.reset_disable = true;
ixgbe_stop(adapter);
error = hw->phy.ops.enter_lplu(hw);
if (error)
device_printf(dev,
"Error entering LPLU: %d\n", error);
hw->phy.reset_disable = false;
} else {
/* Just stop for other adapters */
ixgbe_stop(adapter);
}
return error;
}
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/**********************************************************************
*
* Update the board statistics counters.
*
**********************************************************************/
static void
ixgbe_update_stats_counters(struct adapter *adapter)
{
struct ixgbe_hw *hw = &adapter->hw;
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
u32 missed_rx = 0, bprc, lxon, lxoff, total;
u64 total_missed_rx = 0;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
adapter->stats.pf.crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
adapter->stats.pf.illerrc += IXGBE_READ_REG(hw, IXGBE_ILLERRC);
adapter->stats.pf.errbc += IXGBE_READ_REG(hw, IXGBE_ERRBC);
adapter->stats.pf.mspdc += IXGBE_READ_REG(hw, IXGBE_MSPDC);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
for (int i = 0; i < 16; i++) {
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
adapter->stats.pf.qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
adapter->stats.pf.qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
adapter->stats.pf.qprdc[i] += IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
}
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
adapter->stats.pf.mlfc += IXGBE_READ_REG(hw, IXGBE_MLFC);
adapter->stats.pf.mrfc += IXGBE_READ_REG(hw, IXGBE_MRFC);
adapter->stats.pf.rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
/* Hardware workaround, gprc counts missed packets */
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
adapter->stats.pf.gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
adapter->stats.pf.gprc -= missed_rx;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
- Add the RX refresh changes from igb to ixgbe - Also a couple minor tweaks to the TX code from the same source. - Add the INET ioctl code which has been missing from this driver, and which caused IP aliases to reset the interface. - Last, some minor logic changes that just reflect upcoming hardware support, but have no other functional effect now. MFC after a week
2011-04-25 23:34:21 +00:00
if (hw->mac.type != ixgbe_mac_82598EB) {
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
adapter->stats.pf.gorc += IXGBE_READ_REG(hw, IXGBE_GORCL) +
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
((u64)IXGBE_READ_REG(hw, IXGBE_GORCH) << 32);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
adapter->stats.pf.gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL) +
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
((u64)IXGBE_READ_REG(hw, IXGBE_GOTCH) << 32);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
adapter->stats.pf.tor += IXGBE_READ_REG(hw, IXGBE_TORL) +
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
((u64)IXGBE_READ_REG(hw, IXGBE_TORH) << 32);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
adapter->stats.pf.lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
adapter->stats.pf.lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
A couple of changes: first, make header split on by default and remove from its dependency on LRO, my tests have shown that its always beneficial, even when doing bridging. Second, fix up a few problems in the statistics code, the adapter dependencies had gotten lost so some code that should only run on 82599 was always running, this resulted in bogus flow control numbers on 82598.
2009-08-24 20:41:51 +00:00
} else {
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
adapter->stats.pf.lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
adapter->stats.pf.lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
A couple of changes: first, make header split on by default and remove from its dependency on LRO, my tests have shown that its always beneficial, even when doing bridging. Second, fix up a few problems in the statistics code, the adapter dependencies had gotten lost so some code that should only run on 82599 was always running, this resulted in bogus flow control numbers on 82598.
2009-08-24 20:41:51 +00:00
/* 82598 only has a counter in the high register */
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
adapter->stats.pf.gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
adapter->stats.pf.gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
adapter->stats.pf.tor += IXGBE_READ_REG(hw, IXGBE_TORH);
A couple of changes: first, make header split on by default and remove from its dependency on LRO, my tests have shown that its always beneficial, even when doing bridging. Second, fix up a few problems in the statistics code, the adapter dependencies had gotten lost so some code that should only run on 82599 was always running, this resulted in bogus flow control numbers on 82598.
2009-08-24 20:41:51 +00:00
}
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
/*
* Workaround: mprc hardware is incorrectly counting
* broadcasts, so for now we subtract those.
*/
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
adapter->stats.pf.bprc += bprc;
adapter->stats.pf.mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
Update driver to version 2.3.8: CRITICAL FIX - with stats changes the older 82598 will panic and trash the stack on driver load, FCOE registers ONLY exist in 82599 and must not be read otherwise. kern/153951 - to correct incorrect media type on adapters with pluggable modules I have eliminated the old static table in favor of a new dynamic shared code routine. This also has the benefit of detecting changes when a different module is inserted. Performance/enhancement to the Flow Director code from my linux coworker (the developer of the code). Fixes from Michael Tuexen - a data corruption problem on the 82599 (CRITICAL), fix so the buf size correctly adjusts as the cluster changes, and max descriptors are set properly. Also added 16K clusters for those REALLY big jumbos :) In the RX path, the RX LOCK was not being released, and this causes LOR problems. Add the code that igb already has. Sync with in house shared code, this was necessary for the Flow Director fix. MFC in 2 days
2011-01-19 19:36:27 +00:00
if (hw->mac.type == ixgbe_mac_82598EB)
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
adapter->stats.pf.mprc -= bprc;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
adapter->stats.pf.prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
adapter->stats.pf.prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
adapter->stats.pf.prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
adapter->stats.pf.prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
adapter->stats.pf.prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
adapter->stats.pf.prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
adapter->stats.pf.lxontxc += lxon;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
adapter->stats.pf.lxofftxc += lxoff;
This is driver version 1.4.4 of the Intel ixgbe driver. -It has new hardware support -It uses a new method of TX cleanup called Head Write Back -It includes the provisional generic TCP LRO feature contributed by Myricom and made general purpose by me. This should move into the stack upon approval but for this driver drop its in here. -Also bug fixes and etc... MFC in a week if no serious issues arise.
2008-05-16 18:46:30 +00:00
total = lxon + lxoff;
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
adapter->stats.pf.gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
adapter->stats.pf.mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
adapter->stats.pf.ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
adapter->stats.pf.gptc -= total;
adapter->stats.pf.mptc -= total;
adapter->stats.pf.ptc64 -= total;
adapter->stats.pf.gotc -= total * ETHER_MIN_LEN;
adapter->stats.pf.ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
adapter->stats.pf.rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
adapter->stats.pf.roc += IXGBE_READ_REG(hw, IXGBE_ROC);
adapter->stats.pf.rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
adapter->stats.pf.mngprc += IXGBE_READ_REG(hw, IXGBE_MNGPRC);
adapter->stats.pf.mngpdc += IXGBE_READ_REG(hw, IXGBE_MNGPDC);
adapter->stats.pf.mngptc += IXGBE_READ_REG(hw, IXGBE_MNGPTC);
adapter->stats.pf.tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
adapter->stats.pf.tpt += IXGBE_READ_REG(hw, IXGBE_TPT);
adapter->stats.pf.ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
adapter->stats.pf.ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
adapter->stats.pf.ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
adapter->stats.pf.ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
adapter->stats.pf.ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
adapter->stats.pf.bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
adapter->stats.pf.xec += IXGBE_READ_REG(hw, IXGBE_XEC);
adapter->stats.pf.fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
adapter->stats.pf.fclast += IXGBE_READ_REG(hw, IXGBE_FCLAST);
Update driver to version 2.3.8: CRITICAL FIX - with stats changes the older 82598 will panic and trash the stack on driver load, FCOE registers ONLY exist in 82599 and must not be read otherwise. kern/153951 - to correct incorrect media type on adapters with pluggable modules I have eliminated the old static table in favor of a new dynamic shared code routine. This also has the benefit of detecting changes when a different module is inserted. Performance/enhancement to the Flow Director code from my linux coworker (the developer of the code). Fixes from Michael Tuexen - a data corruption problem on the 82599 (CRITICAL), fix so the buf size correctly adjusts as the cluster changes, and max descriptors are set properly. Also added 16K clusters for those REALLY big jumbos :) In the RX path, the RX LOCK was not being released, and this causes LOR problems. Add the code that igb already has. Sync with in house shared code, this was necessary for the Flow Director fix. MFC in 2 days
2011-01-19 19:36:27 +00:00
/* Only read FCOE on 82599 */
- Add the RX refresh changes from igb to ixgbe - Also a couple minor tweaks to the TX code from the same source. - Add the INET ioctl code which has been missing from this driver, and which caused IP aliases to reset the interface. - Last, some minor logic changes that just reflect upcoming hardware support, but have no other functional effect now. MFC after a week
2011-04-25 23:34:21 +00:00
if (hw->mac.type != ixgbe_mac_82598EB) {
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
adapter->stats.pf.fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
adapter->stats.pf.fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
adapter->stats.pf.fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
adapter->stats.pf.fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
adapter->stats.pf.fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
}
/* Fill out the OS statistics structure */
IXGBE_SET_IPACKETS(adapter, adapter->stats.pf.gprc);
IXGBE_SET_OPACKETS(adapter, adapter->stats.pf.gptc);
IXGBE_SET_IBYTES(adapter, adapter->stats.pf.gorc);
IXGBE_SET_OBYTES(adapter, adapter->stats.pf.gotc);
IXGBE_SET_IMCASTS(adapter, adapter->stats.pf.mprc);
IXGBE_SET_OMCASTS(adapter, adapter->stats.pf.mptc);
IXGBE_SET_COLLISIONS(adapter, 0);
IXGBE_SET_IQDROPS(adapter, total_missed_rx);
IXGBE_SET_IERRORS(adapter, adapter->stats.pf.crcerrs
+ adapter->stats.pf.rlec);
Convert to if_get_counter().
2014-09-28 07:29:45 +00:00
}
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
#if __FreeBSD_version >= 1100036
Convert to if_get_counter().
2014-09-28 07:29:45 +00:00
static uint64_t
ixgbe_get_counter(struct ifnet *ifp, ift_counter cnt)
{
struct adapter *adapter;
Various fixes to the stats in igb(4), ixgbe(4), and ixl(4). - Use hardware counters for ifnet stats in igb(4) when possible. This ensures these stats include packets that bypass the regular stack via netmap. - Don't derefence values off the end of the igb(4) VF stats structure. Instead, add a dedicated if_get_counter method for igb(4) VF interfaces. - Report missed packets on igb(4) as input queue drops rather than an input error. - Report bug_ring drop counts as output queue drops for igb(4) and ixgbe(4). - Export the buf_ring drop stats for individual rings via sysctl on ixgbe(4). - Fix a typo that in ixl(4) that caused output queue drops to be reported as input queue drops and input queue drops to be unreported. Differential Revision: https://reviews.freebsd.org/D2402 Reviewed by: jfv, rstone (6) Sponsored by: Norse Corp, Inc.
2015-04-30 18:23:38 +00:00
struct tx_ring *txr;
uint64_t rv;
Convert to if_get_counter().
2014-09-28 07:29:45 +00:00
adapter = if_getsoftc(ifp);
switch (cnt) {
case IFCOUNTER_IPACKETS:
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
return (adapter->ipackets);
Convert to if_get_counter().
2014-09-28 07:29:45 +00:00
case IFCOUNTER_OPACKETS:
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
return (adapter->opackets);
Convert to if_get_counter().
2014-09-28 07:29:45 +00:00
case IFCOUNTER_IBYTES:
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
return (adapter->ibytes);
Convert to if_get_counter().
2014-09-28 07:29:45 +00:00
case IFCOUNTER_OBYTES:
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
return (adapter->obytes);
Convert to if_get_counter().
2014-09-28 07:29:45 +00:00
case IFCOUNTER_IMCASTS:
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
return (adapter->imcasts);
Convert to if_get_counter().
2014-09-28 07:29:45 +00:00
case IFCOUNTER_OMCASTS:
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
return (adapter->omcasts);
Convert to if_get_counter().
2014-09-28 07:29:45 +00:00
case IFCOUNTER_COLLISIONS:
return (0);
case IFCOUNTER_IQDROPS:
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
return (adapter->iqdrops);
Various fixes to the stats in igb(4), ixgbe(4), and ixl(4). - Use hardware counters for ifnet stats in igb(4) when possible. This ensures these stats include packets that bypass the regular stack via netmap. - Don't derefence values off the end of the igb(4) VF stats structure. Instead, add a dedicated if_get_counter method for igb(4) VF interfaces. - Report missed packets on igb(4) as input queue drops rather than an input error. - Report bug_ring drop counts as output queue drops for igb(4) and ixgbe(4). - Export the buf_ring drop stats for individual rings via sysctl on ixgbe(4). - Fix a typo that in ixl(4) that caused output queue drops to be reported as input queue drops and input queue drops to be unreported. Differential Revision: https://reviews.freebsd.org/D2402 Reviewed by: jfv, rstone (6) Sponsored by: Norse Corp, Inc.
2015-04-30 18:23:38 +00:00
case IFCOUNTER_OQDROPS:
rv = 0;
txr = adapter->tx_rings;
for (int i = 0; i < adapter->num_queues; i++, txr++)
rv += txr->br->br_drops;
return (rv);
Convert to if_get_counter().
2014-09-28 07:29:45 +00:00
case IFCOUNTER_IERRORS:
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
return (adapter->ierrors);
Convert to if_get_counter().
2014-09-28 07:29:45 +00:00
default:
return (if_get_counter_default(ifp, cnt));
}
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
#endif
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
/** ixgbe_sysctl_tdh_handler - Handler function
* Retrieves the TDH value from the hardware
*/
static int
ixgbe_sysctl_tdh_handler(SYSCTL_HANDLER_ARGS)
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
{
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
int error;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
struct tx_ring *txr = ((struct tx_ring *)oidp->oid_arg1);
if (!txr) return 0;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
unsigned val = IXGBE_READ_REG(&txr->adapter->hw, IXGBE_TDH(txr->me));
error = sysctl_handle_int(oidp, &val, 0, req);
if (error || !req->newptr)
return error;
return 0;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
/** ixgbe_sysctl_tdt_handler - Handler function
* Retrieves the TDT value from the hardware
*/
static int
ixgbe_sysctl_tdt_handler(SYSCTL_HANDLER_ARGS)
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
{
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
int error;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
struct tx_ring *txr = ((struct tx_ring *)oidp->oid_arg1);
if (!txr) return 0;
unsigned val = IXGBE_READ_REG(&txr->adapter->hw, IXGBE_TDT(txr->me));
error = sysctl_handle_int(oidp, &val, 0, req);
if (error || !req->newptr)
return error;
return 0;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
/** ixgbe_sysctl_rdh_handler - Handler function
* Retrieves the RDH value from the hardware
*/
static int
ixgbe_sysctl_rdh_handler(SYSCTL_HANDLER_ARGS)
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
{
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
int error;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
struct rx_ring *rxr = ((struct rx_ring *)oidp->oid_arg1);
if (!rxr) return 0;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
unsigned val = IXGBE_READ_REG(&rxr->adapter->hw, IXGBE_RDH(rxr->me));
error = sysctl_handle_int(oidp, &val, 0, req);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
if (error || !req->newptr)
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
return error;
return 0;
}
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
/** ixgbe_sysctl_rdt_handler - Handler function
* Retrieves the RDT value from the hardware
*/
static int
ixgbe_sysctl_rdt_handler(SYSCTL_HANDLER_ARGS)
{
int error;
struct rx_ring *rxr = ((struct rx_ring *)oidp->oid_arg1);
if (!rxr) return 0;
unsigned val = IXGBE_READ_REG(&rxr->adapter->hw, IXGBE_RDT(rxr->me));
error = sysctl_handle_int(oidp, &val, 0, req);
if (error || !req->newptr)
return error;
return 0;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
static int
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
ixgbe_sysctl_interrupt_rate_handler(SYSCTL_HANDLER_ARGS)
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
{
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
int error;
struct ix_queue *que = ((struct ix_queue *)oidp->oid_arg1);
unsigned int reg, usec, rate;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
reg = IXGBE_READ_REG(&que->adapter->hw, IXGBE_EITR(que->msix));
usec = ((reg & 0x0FF8) >> 3);
if (usec > 0)
ixgbe changes: - remove experimental code for disabling CRC - use the correct constant for conversion between interrupt rate and EITR values (the previous values were off by a factor of 2) - make dev.ix.N.queueM.interrupt_rate a RW sysctl variable. Changing individual values affects the queue immediately, and propagates to all interfaces at the next reinit. - add dev.ix.N.queueM.irqs rdonly sysctl, to export the actual interrupt counts Netmap-related changes for ixgbe: - use the "new" format for TX descriptors in netmap mode. - pass interrupt mitigation delays to the user process doing poll() on a netmap file descriptor. On the RX side this means we will not check the ring more than once per interrupt. This gives the process a chance to sleep and process packets in larger batches, thus reducing CPU usage. On the TX side we take this even further: completed transmissions are reclaimed every half ring even if the NIC interrupts more often. This saves even more CPU without any additional tx delays. Generic Netmap-related changes: - align the netmap_kring to cache lines so that there is no false sharing (possibly useful for multiqueue NICs and MSIX interrupts, which are handled by different cores). It's a minor improvement but it does not cost anything. Reviewed by: Jack Vogel Approved by: Jack Vogel
2012-01-26 09:55:16 +00:00
rate = 500000 / usec;
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
else
rate = 0;
error = sysctl_handle_int(oidp, &rate, 0, req);
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
if (error || !req->newptr)
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
return error;
ixgbe changes: - remove experimental code for disabling CRC - use the correct constant for conversion between interrupt rate and EITR values (the previous values were off by a factor of 2) - make dev.ix.N.queueM.interrupt_rate a RW sysctl variable. Changing individual values affects the queue immediately, and propagates to all interfaces at the next reinit. - add dev.ix.N.queueM.irqs rdonly sysctl, to export the actual interrupt counts Netmap-related changes for ixgbe: - use the "new" format for TX descriptors in netmap mode. - pass interrupt mitigation delays to the user process doing poll() on a netmap file descriptor. On the RX side this means we will not check the ring more than once per interrupt. This gives the process a chance to sleep and process packets in larger batches, thus reducing CPU usage. On the TX side we take this even further: completed transmissions are reclaimed every half ring even if the NIC interrupts more often. This saves even more CPU without any additional tx delays. Generic Netmap-related changes: - align the netmap_kring to cache lines so that there is no false sharing (possibly useful for multiqueue NICs and MSIX interrupts, which are handled by different cores). It's a minor improvement but it does not cost anything. Reviewed by: Jack Vogel Approved by: Jack Vogel
2012-01-26 09:55:16 +00:00
reg &= ~0xfff; /* default, no limitation */
ixgbe_max_interrupt_rate = 0;
if (rate > 0 && rate < 500000) {
if (rate < 1000)
rate = 1000;
ixgbe_max_interrupt_rate = rate;
reg |= ((4000000/rate) & 0xff8 );
}
IXGBE_WRITE_REG(&que->adapter->hw, IXGBE_EITR(que->msix), reg);
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
return 0;
}
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
static void
ixgbe_add_device_sysctls(struct adapter *adapter)
{
device_t dev = adapter->dev;
struct ixgbe_hw *hw = &adapter->hw;
struct sysctl_oid_list *child;
struct sysctl_ctx_list *ctx;
ctx = device_get_sysctl_ctx(dev);
child = SYSCTL_CHILDREN(device_get_sysctl_tree(dev));
/* Sysctls for all devices */
SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "fc",
CTLTYPE_INT | CTLFLAG_RW, adapter, 0,
ixgbe_set_flowcntl, "I", IXGBE_SYSCTL_DESC_SET_FC);
SYSCTL_ADD_INT(ctx, child, OID_AUTO, "enable_aim",
CTLFLAG_RW,
&ixgbe_enable_aim, 1, "Interrupt Moderation");
SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "advertise_speed",
CTLTYPE_INT | CTLFLAG_RW, adapter, 0,
ixgbe_set_advertise, "I", IXGBE_SYSCTL_DESC_ADV_SPEED);
SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "thermal_test",
CTLTYPE_INT | CTLFLAG_RW, adapter, 0,
ixgbe_sysctl_thermal_test, "I", "Thermal Test");
/* for X550 devices */
if (hw->mac.type >= ixgbe_mac_X550)
SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "dmac",
CTLTYPE_INT | CTLFLAG_RW, adapter, 0,
ixgbe_sysctl_dmac, "I", "DMA Coalesce");
/* for X550T and X550EM backplane devices */
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
if (hw->device_id == IXGBE_DEV_ID_X550T ||
hw->device_id == IXGBE_DEV_ID_X550EM_X_KR) {
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
struct sysctl_oid *eee_node;
struct sysctl_oid_list *eee_list;
eee_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "eee",
CTLFLAG_RD, NULL,
"Energy Efficient Ethernet sysctls");
eee_list = SYSCTL_CHILDREN(eee_node);
SYSCTL_ADD_PROC(ctx, eee_list, OID_AUTO, "enable",
CTLTYPE_INT | CTLFLAG_RW, adapter, 0,
ixgbe_sysctl_eee_enable, "I",
"Enable or Disable EEE");
SYSCTL_ADD_PROC(ctx, eee_list, OID_AUTO, "negotiated",
CTLTYPE_INT | CTLFLAG_RD, adapter, 0,
ixgbe_sysctl_eee_negotiated, "I",
"EEE negotiated on link");
SYSCTL_ADD_PROC(ctx, eee_list, OID_AUTO, "tx_lpi_status",
CTLTYPE_INT | CTLFLAG_RD, adapter, 0,
ixgbe_sysctl_eee_tx_lpi_status, "I",
"Whether or not TX link is in LPI state");
SYSCTL_ADD_PROC(ctx, eee_list, OID_AUTO, "rx_lpi_status",
CTLTYPE_INT | CTLFLAG_RD, adapter, 0,
ixgbe_sysctl_eee_rx_lpi_status, "I",
"Whether or not RX link is in LPI state");
}
/* for certain 10GBaseT devices */
if (hw->device_id == IXGBE_DEV_ID_X550T ||
hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T) {
SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "wol_enable",
CTLTYPE_INT | CTLFLAG_RW, adapter, 0,
ixgbe_sysctl_wol_enable, "I",
"Enable/Disable Wake on LAN");
SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "wufc",
CTLTYPE_INT | CTLFLAG_RW, adapter, 0,
ixgbe_sysctl_wufc, "I",
"Enable/Disable Wake Up Filters");
}
/* for X550EM 10GBaseT devices */
if (hw->device_id == IXGBE_DEV_ID_X550EM_X_10G_T) {
struct sysctl_oid *phy_node;
struct sysctl_oid_list *phy_list;
phy_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "phy",
CTLFLAG_RD, NULL,
"External PHY sysctls");
phy_list = SYSCTL_CHILDREN(phy_node);
SYSCTL_ADD_PROC(ctx, phy_list, OID_AUTO, "temp",
CTLTYPE_INT | CTLFLAG_RD, adapter, 0,
ixgbe_sysctl_phy_temp, "I",
"Current External PHY Temperature (Celsius)");
SYSCTL_ADD_PROC(ctx, phy_list, OID_AUTO, "overtemp_occurred",
CTLTYPE_INT | CTLFLAG_RD, adapter, 0,
ixgbe_sysctl_phy_overtemp_occurred, "I",
"External PHY High Temperature Event Occurred");
}
}
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
/*
* Add sysctl variables, one per statistic, to the system.
*/
static void
ixgbe_add_hw_stats(struct adapter *adapter)
{
device_t dev = adapter->dev;
struct tx_ring *txr = adapter->tx_rings;
struct rx_ring *rxr = adapter->rx_rings;
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
struct sysctl_oid *tree = device_get_sysctl_tree(dev);
struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
struct ixgbe_hw_stats *stats = &adapter->stats.pf;
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
struct sysctl_oid *stat_node, *queue_node;
struct sysctl_oid_list *stat_list, *queue_list;
#define QUEUE_NAME_LEN 32
char namebuf[QUEUE_NAME_LEN];
/* Driver Statistics */
SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
CTLFLAG_RD, &adapter->dropped_pkts,
"Driver dropped packets");
SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "mbuf_defrag_failed",
CTLFLAG_RD, &adapter->mbuf_defrag_failed,
"m_defrag() failed");
SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_events",
CTLFLAG_RD, &adapter->watchdog_events,
"Watchdog timeouts");
SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq",
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
CTLFLAG_RD, &adapter->link_irq,
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
"Link MSIX IRQ Handled");
for (int i = 0; i < adapter->num_queues; i++, txr++) {
snprintf(namebuf, QUEUE_NAME_LEN, "queue%d", i);
queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
CTLFLAG_RD, NULL, "Queue Name");
queue_list = SYSCTL_CHILDREN(queue_node);
SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "interrupt_rate",
ixgbe changes: - remove experimental code for disabling CRC - use the correct constant for conversion between interrupt rate and EITR values (the previous values were off by a factor of 2) - make dev.ix.N.queueM.interrupt_rate a RW sysctl variable. Changing individual values affects the queue immediately, and propagates to all interfaces at the next reinit. - add dev.ix.N.queueM.irqs rdonly sysctl, to export the actual interrupt counts Netmap-related changes for ixgbe: - use the "new" format for TX descriptors in netmap mode. - pass interrupt mitigation delays to the user process doing poll() on a netmap file descriptor. On the RX side this means we will not check the ring more than once per interrupt. This gives the process a chance to sleep and process packets in larger batches, thus reducing CPU usage. On the TX side we take this even further: completed transmissions are reclaimed every half ring even if the NIC interrupts more often. This saves even more CPU without any additional tx delays. Generic Netmap-related changes: - align the netmap_kring to cache lines so that there is no false sharing (possibly useful for multiqueue NICs and MSIX interrupts, which are handled by different cores). It's a minor improvement but it does not cost anything. Reviewed by: Jack Vogel Approved by: Jack Vogel
2012-01-26 09:55:16 +00:00
CTLTYPE_UINT | CTLFLAG_RW, &adapter->queues[i],
Specify a CTLTYPE_FOO so that a future sysctl(8) change does not need to rely on the format string.
2011-01-18 21:14:23 +00:00
sizeof(&adapter->queues[i]),
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
ixgbe_sysctl_interrupt_rate_handler, "IU",
"Interrupt Rate");
ixgbe changes: - remove experimental code for disabling CRC - use the correct constant for conversion between interrupt rate and EITR values (the previous values were off by a factor of 2) - make dev.ix.N.queueM.interrupt_rate a RW sysctl variable. Changing individual values affects the queue immediately, and propagates to all interfaces at the next reinit. - add dev.ix.N.queueM.irqs rdonly sysctl, to export the actual interrupt counts Netmap-related changes for ixgbe: - use the "new" format for TX descriptors in netmap mode. - pass interrupt mitigation delays to the user process doing poll() on a netmap file descriptor. On the RX side this means we will not check the ring more than once per interrupt. This gives the process a chance to sleep and process packets in larger batches, thus reducing CPU usage. On the TX side we take this even further: completed transmissions are reclaimed every half ring even if the NIC interrupts more often. This saves even more CPU without any additional tx delays. Generic Netmap-related changes: - align the netmap_kring to cache lines so that there is no false sharing (possibly useful for multiqueue NICs and MSIX interrupts, which are handled by different cores). It's a minor improvement but it does not cost anything. Reviewed by: Jack Vogel Approved by: Jack Vogel
2012-01-26 09:55:16 +00:00
SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "irqs",
CTLFLAG_RD, &(adapter->queues[i].irqs),
"irqs on this queue");
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head",
Specify a CTLTYPE_FOO so that a future sysctl(8) change does not need to rely on the format string.
2011-01-18 21:14:23 +00:00
CTLTYPE_UINT | CTLFLAG_RD, txr, sizeof(txr),
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
ixgbe_sysctl_tdh_handler, "IU",
"Transmit Descriptor Head");
SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail",
Specify a CTLTYPE_FOO so that a future sysctl(8) change does not need to rely on the format string.
2011-01-18 21:14:23 +00:00
CTLTYPE_UINT | CTLFLAG_RD, txr, sizeof(txr),
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
ixgbe_sysctl_tdt_handler, "IU",
"Transmit Descriptor Tail");
Patch #8 Performance changes - this one improves locality, moving some counters and data to the ring struct from the adapter struct, also compressing some data in the move.
2012-11-30 23:45:55 +00:00
SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "tso_tx",
CTLFLAG_RD, &txr->tso_tx,
"TSO");
SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "no_tx_dma_setup",
CTLFLAG_RD, &txr->no_tx_dma_setup,
"Driver tx dma failure in xmit");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "no_desc_avail",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &txr->no_desc_avail,
"Queue No Descriptor Available");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "tx_packets",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &txr->total_packets,
"Queue Packets Transmitted");
Various fixes to the stats in igb(4), ixgbe(4), and ixl(4). - Use hardware counters for ifnet stats in igb(4) when possible. This ensures these stats include packets that bypass the regular stack via netmap. - Don't derefence values off the end of the igb(4) VF stats structure. Instead, add a dedicated if_get_counter method for igb(4) VF interfaces. - Report missed packets on igb(4) as input queue drops rather than an input error. - Report bug_ring drop counts as output queue drops for igb(4) and ixgbe(4). - Export the buf_ring drop stats for individual rings via sysctl on ixgbe(4). - Fix a typo that in ixl(4) that caused output queue drops to be reported as input queue drops and input queue drops to be unreported. Differential Revision: https://reviews.freebsd.org/D2402 Reviewed by: jfv, rstone (6) Sponsored by: Norse Corp, Inc.
2015-04-30 18:23:38 +00:00
SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "br_drops",
CTLFLAG_RD, &txr->br->br_drops,
"Packets dropped in buf_ring");
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
for (int i = 0; i < adapter->num_queues; i++, rxr++) {
snprintf(namebuf, QUEUE_NAME_LEN, "queue%d", i);
queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
CTLFLAG_RD, NULL, "Queue Name");
queue_list = SYSCTL_CHILDREN(queue_node);
struct lro_ctrl *lro = &rxr->lro;
snprintf(namebuf, QUEUE_NAME_LEN, "queue%d", i);
queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
CTLFLAG_RD, NULL, "Queue Name");
queue_list = SYSCTL_CHILDREN(queue_node);
SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head",
Specify a CTLTYPE_FOO so that a future sysctl(8) change does not need to rely on the format string.
2011-01-18 21:14:23 +00:00
CTLTYPE_UINT | CTLFLAG_RD, rxr, sizeof(rxr),
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
ixgbe_sysctl_rdh_handler, "IU",
"Receive Descriptor Head");
SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail",
Specify a CTLTYPE_FOO so that a future sysctl(8) change does not need to rely on the format string.
2011-01-18 21:14:23 +00:00
CTLTYPE_UINT | CTLFLAG_RD, rxr, sizeof(rxr),
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
ixgbe_sysctl_rdt_handler, "IU",
"Receive Descriptor Tail");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "rx_packets",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &rxr->rx_packets,
"Queue Packets Received");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "rx_bytes",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &rxr->rx_bytes,
"Queue Bytes Received");
Heavily optimize the case of small RX packets of 160 bytes or less. For this case, allocate a plain mbuf and copy the frame into it, then send the copy up the stack, leaving the original mbuf+cluster in place in the receive ring for immediate re-use. This saves a trip through 2 of the 3 zones of the compound mbuf allocator, a trip through busdma, and a trip through the 1 of the 3 mbuf destructors. For our load at Netflix, this can lower CPU consumption by as much as 20%. The copy algorithm is based on investigative work from Luigi Rizzo earlier in the year. Reviewed by: jfv Obtained from: Netflix
2012-08-31 10:07:38 +00:00
SYSCTL_ADD_UQUAD(ctx, queue_list, OID_AUTO, "rx_copies",
CTLFLAG_RD, &rxr->rx_copies,
"Copied RX Frames");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_INT(ctx, queue_list, OID_AUTO, "lro_queued",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &lro->lro_queued, 0,
"LRO Queued");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_INT(ctx, queue_list, OID_AUTO, "lro_flushed",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &lro->lro_flushed, 0,
"LRO Flushed");
}
/* MAC stats get the own sub node */
stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
CTLFLAG_RD, NULL, "MAC Statistics");
stat_list = SYSCTL_CHILDREN(stat_node);
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->crcerrs,
"CRC Errors");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "ill_errs",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->illerrc,
"Illegal Byte Errors");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "byte_errs",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->errbc,
"Byte Errors");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "short_discards",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->mspdc,
"MAC Short Packets Discarded");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "local_faults",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->mlfc,
"MAC Local Faults");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "remote_faults",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->mrfc,
"MAC Remote Faults");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rec_len_errs",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->rlec,
"Receive Length Errors");
Patch #4 - this does two things, it removes a number of statistics, these are FCOE stats (fiber channel over ethernet), something that FreeBSD does not yet have, they were mistaken for flow control by the implementor I believe. Secondly, the real flow control stats are oddly named with a 'link' tag on the front, it was requested by my validation engineer to make these stats have the same name as the igb driver for clarity and that seemed reasonable to me.
2012-11-30 22:54:14 +00:00
/* Flow Control stats */
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->lxontxc,
"Link XON Transmitted");
Patch #4 - this does two things, it removes a number of statistics, these are FCOE stats (fiber channel over ethernet), something that FreeBSD does not yet have, they were mistaken for flow control by the implementor I believe. Secondly, the real flow control stats are oddly named with a 'link' tag on the front, it was requested by my validation engineer to make these stats have the same name as the igb driver for clarity and that seemed reasonable to me.
2012-11-30 22:54:14 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
kern/153772 fix variable names. Thank you Andrew Boyer for catching these MFC in 3 days
2011-01-07 22:34:56 +00:00
CTLFLAG_RD, &stats->lxonrxc,
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
"Link XON Received");
Patch #4 - this does two things, it removes a number of statistics, these are FCOE stats (fiber channel over ethernet), something that FreeBSD does not yet have, they were mistaken for flow control by the implementor I believe. Secondly, the real flow control stats are oddly named with a 'link' tag on the front, it was requested by my validation engineer to make these stats have the same name as the igb driver for clarity and that seemed reasonable to me.
2012-11-30 22:54:14 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->lxofftxc,
"Link XOFF Transmitted");
Patch #4 - this does two things, it removes a number of statistics, these are FCOE stats (fiber channel over ethernet), something that FreeBSD does not yet have, they were mistaken for flow control by the implementor I believe. Secondly, the real flow control stats are oddly named with a 'link' tag on the front, it was requested by my validation engineer to make these stats have the same name as the igb driver for clarity and that seemed reasonable to me.
2012-11-30 22:54:14 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
kern/153772 fix variable names. Thank you Andrew Boyer for catching these MFC in 3 days
2011-01-07 22:34:56 +00:00
CTLFLAG_RD, &stats->lxoffrxc,
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
"Link XOFF Received");
/* Packet Reception Stats */
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_octets_rcvd",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->tor,
"Total Octets Received");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_rcvd",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->gorc,
"Good Octets Received");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_rcvd",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->tpr,
"Total Packets Received");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_rcvd",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->gprc,
"Good Packets Received");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_rcvd",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->mprc,
"Multicast Packets Received");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_rcvd",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->bprc,
"Broadcast Packets Received");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->prc64,
"64 byte frames received ");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->prc127,
"65-127 byte frames received");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->prc255,
"128-255 byte frames received");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->prc511,
"256-511 byte frames received");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->prc1023,
"512-1023 byte frames received");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->prc1522,
"1023-1522 byte frames received");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersized",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->ruc,
"Receive Undersized");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->rfc,
"Fragmented Packets Received ");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversized",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->roc,
"Oversized Packets Received");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabberd",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->rjc,
"Received Jabber");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "management_pkts_rcvd",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->mngprc,
"Management Packets Received");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "management_pkts_drpd",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->mngptc,
"Management Packets Dropped");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "checksum_errs",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->xec,
"Checksum Errors");
/* Packet Transmission Stats */
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->gotc,
"Good Octets Transmitted");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->tpt,
"Total Packets Transmitted");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->gptc,
"Good Packets Transmitted");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->bptc,
"Broadcast Packets Transmitted");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->mptc,
"Multicast Packets Transmitted");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "management_pkts_txd",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->mngptc,
"Management Packets Transmitted");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->ptc64,
"64 byte frames transmitted ");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->ptc127,
"65-127 byte frames transmitted");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->ptc255,
"128-255 byte frames transmitted");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->ptc511,
"256-511 byte frames transmitted");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->ptc1023,
"512-1023 byte frames transmitted");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
Interrupt handler, and stats changes from Michael Tuexen, thanks Michael!
2010-11-27 01:34:09 +00:00
CTLFLAG_RD, &stats->ptc1522,
"1024-1522 byte frames transmitted");
Add the actual source too :) Approved by: re
2007-07-11 23:03:16 +00:00
}
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
/*
** Set flow control using sysctl:
** Flow control values:
** 0 - off
** 1 - rx pause
** 2 - tx pause
** 3 - full
*/
static int
ixgbe_set_flowcntl(SYSCTL_HANDLER_ARGS)
{
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
int error, last;
struct adapter *adapter = (struct adapter *) arg1;
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
last = adapter->fc;
error = sysctl_handle_int(oidp, &adapter->fc, 0, req);
if ((error) || (req->newptr == NULL))
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
return (error);
Don't bother to run the flowcontrol code if there is no change. Thanks to Andrew for the tweak.
2011-01-22 00:19:15 +00:00
/* Don't bother if it's not changed */
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
if (adapter->fc == last)
Don't bother to run the flowcontrol code if there is no change. Thanks to Andrew for the tweak.
2011-01-22 00:19:15 +00:00
return (0);
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
switch (adapter->fc) {
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
case ixgbe_fc_rx_pause:
case ixgbe_fc_tx_pause:
case ixgbe_fc_full:
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
adapter->hw.fc.requested_mode = adapter->fc;
A few important fixes: - Testing TSO6 has led me to discover that HW RSC is a problematic feature, it is ONLY designed to work with IPv4 in the first place, and if IP forwarding is done it can't be disabled as LRO in the stack, also initial testing we've done at Intel shows an equal performance using TSO[46] on the TX and LRO on RX, if you ran older code on 82599 or later hardware you actually could have detrimental performance for this reason. So I am disabling the feature by default and all our adapters will now use LRO instead. - If you have flow control off and multiple queues it was possible when the buffer of one queue becomes full that all RX movement is stalled, to eliminate this problem a feature bit is now set that will allow packets to be dropped when full rather than stall. Note, the default is to have flow control on, and this keeps this from happening. - Because of the recent fixes in the stack, LRO is now auto-disabled when problematic, so I have decided to enable it by default in the capabilities in the driver. - There are some 1G modules used by some customers, a couple small tweaks to properly support those in the media code. - A note: we have now done some testing of TSO6 and using LRO with IPv6 and it all works great!! Seeing line rate in both directions in best cases. Thanks bz for your excellent work!!
2012-10-31 23:50:36 +00:00
if (adapter->num_queues > 1)
ixgbe_disable_rx_drop(adapter);
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
break;
case ixgbe_fc_none:
Updated ixgbe driver - version 1.6.2 -This version has header split, and as a result a number of aspects of the code have been improved/simplified. - Interrupt handling refined for performance - Many small bugs fixed along the way MFC after: ASAP - in time for 7.1
2008-11-26 23:41:18 +00:00
adapter->hw.fc.requested_mode = ixgbe_fc_none;
A few important fixes: - Testing TSO6 has led me to discover that HW RSC is a problematic feature, it is ONLY designed to work with IPv4 in the first place, and if IP forwarding is done it can't be disabled as LRO in the stack, also initial testing we've done at Intel shows an equal performance using TSO[46] on the TX and LRO on RX, if you ran older code on 82599 or later hardware you actually could have detrimental performance for this reason. So I am disabling the feature by default and all our adapters will now use LRO instead. - If you have flow control off and multiple queues it was possible when the buffer of one queue becomes full that all RX movement is stalled, to eliminate this problem a feature bit is now set that will allow packets to be dropped when full rather than stall. Note, the default is to have flow control on, and this keeps this from happening. - Because of the recent fixes in the stack, LRO is now auto-disabled when problematic, so I have decided to enable it by default in the capabilities in the driver. - There are some 1G modules used by some customers, a couple small tweaks to properly support those in the media code. - A note: we have now done some testing of TSO6 and using LRO with IPv6 and it all works great!! Seeing line rate in both directions in best cases. Thanks bz for your excellent work!!
2012-10-31 23:50:36 +00:00
if (adapter->num_queues > 1)
ixgbe_enable_rx_drop(adapter);
First, sync to internal shared code, and then Fixes: - flow control - don't override user value on re-init - fix to make 1G optics work correctly - change to interrupt enabling - some bits were incorrect for certain hardware. - certain stats fixes, remove a duplicate increment of ierror, thanks to Scott Long for pointing these out. - shared code link interface changed, requiring some core code changes to accomodate this. - add an m_adj() to ETHER_ALIGN on the recieve side, this was requested by Mike Karels, thanks Mike. - Multicast code corrections also thanks to Mike Karels.
2013-03-04 23:07:40 +00:00
break;
default:
adapter->fc = last;
return (EINVAL);
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
}
Update to the ixgbe driver: - Add a couple of new devices - Flow control changes in shared and core code - Bug fix to Flow Director for 82598 - Shared code sync to internal with required core change Thanks to those helping in the testing and improvements to this driver! MFC after:5 days
2012-07-05 20:51:44 +00:00
/* Don't autoneg if forcing a value */
adapter->hw.fc.disable_fc_autoneg = TRUE;
ixgbe_fc_enable(&adapter->hw);
This is an update to the new Intel 10G 82598 driver. The first drop was Beta, this code is expected to be the release version. Note that this driver code will build in either 6.2 or 7. If you use the code in 6.2 you will not get TSO or MSI/X support but it will function in a legacy mode. Approved by: re
2007-09-04 02:31:35 +00:00
return error;
}
BAH, I apologize, the wrong version of the code got fat fingered in place, this is the correct version that actually works... <sheepish grin> MFC: in a week
2010-06-30 01:10:08 +00:00
/*
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
** Control advertised link speed:
** Flags:
** 0x1 - advertise 100 Mb
** 0x2 - advertise 1G
** 0x4 - advertise 10G
BAH, I apologize, the wrong version of the code got fat fingered in place, this is the correct version that actually works... <sheepish grin> MFC: in a week
2010-06-30 01:10:08 +00:00
*/
static int
ixgbe_set_advertise(SYSCTL_HANDLER_ARGS)
{
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
int error = 0, requested;
BAH, I apologize, the wrong version of the code got fat fingered in place, this is the correct version that actually works... <sheepish grin> MFC: in a week
2010-06-30 01:10:08 +00:00
struct adapter *adapter;
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
device_t dev;
BAH, I apologize, the wrong version of the code got fat fingered in place, this is the correct version that actually works... <sheepish grin> MFC: in a week
2010-06-30 01:10:08 +00:00
struct ixgbe_hw *hw;
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
ixgbe_link_speed speed = 0;
BAH, I apologize, the wrong version of the code got fat fingered in place, this is the correct version that actually works... <sheepish grin> MFC: in a week
2010-06-30 01:10:08 +00:00
adapter = (struct adapter *) arg1;
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
dev = adapter->dev;
BAH, I apologize, the wrong version of the code got fat fingered in place, this is the correct version that actually works... <sheepish grin> MFC: in a week
2010-06-30 01:10:08 +00:00
hw = &adapter->hw;
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
requested = adapter->advertise;
error = sysctl_handle_int(oidp, &requested, 0, req);
First, sync to internal shared code, and then Fixes: - flow control - don't override user value on re-init - fix to make 1G optics work correctly - change to interrupt enabling - some bits were incorrect for certain hardware. - certain stats fixes, remove a duplicate increment of ierror, thanks to Scott Long for pointing these out. - shared code link interface changed, requiring some core code changes to accomodate this. - add an m_adj() to ETHER_ALIGN on the recieve side, this was requested by Mike Karels, thanks Mike. - Multicast code corrections also thanks to Mike Karels.
2013-03-04 23:07:40 +00:00
if ((error) || (req->newptr == NULL))
BAH, I apologize, the wrong version of the code got fat fingered in place, this is the correct version that actually works... <sheepish grin> MFC: in a week
2010-06-30 01:10:08 +00:00
return (error);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
/* Checks to validate new value */
if (adapter->advertise == requested) /* no change */
Update to the ixgbe driver: - Add a couple of new devices - Flow control changes in shared and core code - Bug fix to Flow Director for 82598 - Shared code sync to internal with required core change Thanks to those helping in the testing and improvements to this driver! MFC after:5 days
2012-07-05 20:51:44 +00:00
return (0);
BAH, I apologize, the wrong version of the code got fat fingered in place, this is the correct version that actually works... <sheepish grin> MFC: in a week
2010-06-30 01:10:08 +00:00
if (!((hw->phy.media_type == ixgbe_media_type_copper) ||
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
(hw->phy.multispeed_fiber))) {
device_printf(dev,
"Advertised speed can only be set on copper or "
"multispeed fiber media types.\n");
First, sync to internal shared code, and then Fixes: - flow control - don't override user value on re-init - fix to make 1G optics work correctly - change to interrupt enabling - some bits were incorrect for certain hardware. - certain stats fixes, remove a duplicate increment of ierror, thanks to Scott Long for pointing these out. - shared code link interface changed, requiring some core code changes to accomodate this. - add an m_adj() to ETHER_ALIGN on the recieve side, this was requested by Mike Karels, thanks Mike. - Multicast code corrections also thanks to Mike Karels.
2013-03-04 23:07:40 +00:00
return (EINVAL);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
}
BAH, I apologize, the wrong version of the code got fat fingered in place, this is the correct version that actually works... <sheepish grin> MFC: in a week
2010-06-30 01:10:08 +00:00
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
if (requested < 0x1 || requested > 0x7) {
device_printf(dev,
"Invalid advertised speed; valid modes are 0x1 through 0x7\n");
First, sync to internal shared code, and then Fixes: - flow control - don't override user value on re-init - fix to make 1G optics work correctly - change to interrupt enabling - some bits were incorrect for certain hardware. - certain stats fixes, remove a duplicate increment of ierror, thanks to Scott Long for pointing these out. - shared code link interface changed, requiring some core code changes to accomodate this. - add an m_adj() to ETHER_ALIGN on the recieve side, this was requested by Mike Karels, thanks Mike. - Multicast code corrections also thanks to Mike Karels.
2013-03-04 23:07:40 +00:00
return (EINVAL);
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
}
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
if ((requested & 0x1)
&& (hw->mac.type != ixgbe_mac_X540)
&& (hw->mac.type != ixgbe_mac_X550)) {
device_printf(dev, "Set Advertise: 100Mb on X540/X550 only\n");
First, sync to internal shared code, and then Fixes: - flow control - don't override user value on re-init - fix to make 1G optics work correctly - change to interrupt enabling - some bits were incorrect for certain hardware. - certain stats fixes, remove a duplicate increment of ierror, thanks to Scott Long for pointing these out. - shared code link interface changed, requiring some core code changes to accomodate this. - add an m_adj() to ETHER_ALIGN on the recieve side, this was requested by Mike Karels, thanks Mike. - Multicast code corrections also thanks to Mike Karels.
2013-03-04 23:07:40 +00:00
return (EINVAL);
}
BAH, I apologize, the wrong version of the code got fat fingered in place, this is the correct version that actually works... <sheepish grin> MFC: in a week
2010-06-30 01:10:08 +00:00
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
/* Set new value and report new advertised mode */
if (requested & 0x1)
speed |= IXGBE_LINK_SPEED_100_FULL;
if (requested & 0x2)
speed |= IXGBE_LINK_SPEED_1GB_FULL;
if (requested & 0x4)
speed |= IXGBE_LINK_SPEED_10GB_FULL;
BAH, I apologize, the wrong version of the code got fat fingered in place, this is the correct version that actually works... <sheepish grin> MFC: in a week
2010-06-30 01:10:08 +00:00
hw->mac.autotry_restart = TRUE;
First, sync to internal shared code, and then Fixes: - flow control - don't override user value on re-init - fix to make 1G optics work correctly - change to interrupt enabling - some bits were incorrect for certain hardware. - certain stats fixes, remove a duplicate increment of ierror, thanks to Scott Long for pointing these out. - shared code link interface changed, requiring some core code changes to accomodate this. - add an m_adj() to ETHER_ALIGN on the recieve side, this was requested by Mike Karels, thanks Mike. - Multicast code corrections also thanks to Mike Karels.
2013-03-04 23:07:40 +00:00
hw->mac.ops.setup_link(hw, speed, TRUE);
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
adapter->advertise = requested;
BAH, I apologize, the wrong version of the code got fat fingered in place, this is the correct version that actually works... <sheepish grin> MFC: in a week
2010-06-30 01:10:08 +00:00
return (error);
}
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
/*
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
* The following two sysctls are for X550 BaseT devices;
* they deal with the external PHY used in them.
*/
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
static int
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
ixgbe_sysctl_phy_temp(SYSCTL_HANDLER_ARGS)
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
{
struct adapter *adapter = (struct adapter *) arg1;
struct ixgbe_hw *hw = &adapter->hw;
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
u16 reg;
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
if (hw->device_id != IXGBE_DEV_ID_X550EM_X_10G_T) {
device_printf(adapter->dev,
"Device has no supported external thermal sensor.\n");
return (ENODEV);
}
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
if (hw->phy.ops.read_reg(hw, IXGBE_PHY_CURRENT_TEMP,
IXGBE_MDIO_VENDOR_SPECIFIC_1_DEV_TYPE,
&reg)) {
device_printf(adapter->dev,
"Error reading from PHY's current temperature register\n");
return (EAGAIN);
}
/* Shift temp for output */
reg = reg >> 8;
return (sysctl_handle_int(oidp, NULL, reg, req));
}
/*
* Reports whether the current PHY temperature is over
* the overtemp threshold.
* - This is reported directly from the PHY
*/
static int
ixgbe_sysctl_phy_overtemp_occurred(SYSCTL_HANDLER_ARGS)
{
struct adapter *adapter = (struct adapter *) arg1;
struct ixgbe_hw *hw = &adapter->hw;
u16 reg;
if (hw->device_id != IXGBE_DEV_ID_X550EM_X_10G_T) {
device_printf(adapter->dev,
"Device has no supported external thermal sensor.\n");
return (ENODEV);
}
if (hw->phy.ops.read_reg(hw, IXGBE_PHY_OVERTEMP_STATUS,
IXGBE_MDIO_VENDOR_SPECIFIC_1_DEV_TYPE,
&reg)) {
device_printf(adapter->dev,
"Error reading from PHY's temperature status register\n");
return (EAGAIN);
}
/* Get occurrence bit */
reg = !!(reg & 0x4000);
return (sysctl_handle_int(oidp, 0, reg, req));
}
/*
** Thermal Shutdown Trigger (internal MAC)
** - Set this to 1 to cause an overtemp event to occur
*/
static int
ixgbe_sysctl_thermal_test(SYSCTL_HANDLER_ARGS)
{
struct adapter *adapter = (struct adapter *) arg1;
struct ixgbe_hw *hw = &adapter->hw;
int error, fire = 0;
New hardware support: Intel X540 adapter support added. Some shared code reorganization along with the new adapter. Sync changes to OACTIVE in igb into this driver. Misc small fixes.
2012-01-30 16:42:02 +00:00
error = sysctl_handle_int(oidp, &fire, 0, req);
if ((error) || (req->newptr == NULL))
return (error);
if (fire) {
u32 reg = IXGBE_READ_REG(hw, IXGBE_EICS);
reg |= IXGBE_EICR_TS;
IXGBE_WRITE_REG(hw, IXGBE_EICS, reg);
}
return (0);
}
A few important fixes: - Testing TSO6 has led me to discover that HW RSC is a problematic feature, it is ONLY designed to work with IPv4 in the first place, and if IP forwarding is done it can't be disabled as LRO in the stack, also initial testing we've done at Intel shows an equal performance using TSO[46] on the TX and LRO on RX, if you ran older code on 82599 or later hardware you actually could have detrimental performance for this reason. So I am disabling the feature by default and all our adapters will now use LRO instead. - If you have flow control off and multiple queues it was possible when the buffer of one queue becomes full that all RX movement is stalled, to eliminate this problem a feature bit is now set that will allow packets to be dropped when full rather than stall. Note, the default is to have flow control on, and this keeps this from happening. - Because of the recent fixes in the stack, LRO is now auto-disabled when problematic, so I have decided to enable it by default in the capabilities in the driver. - There are some 1G modules used by some customers, a couple small tweaks to properly support those in the media code. - A note: we have now done some testing of TSO6 and using LRO with IPv6 and it all works great!! Seeing line rate in both directions in best cases. Thanks bz for your excellent work!!
2012-10-31 23:50:36 +00:00
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
/*
** Manage DMA Coalescing.
** Control values:
** 0/1 - off / on (use default value of 1000)
**
** Legal timer values are:
** 50,100,250,500,1000,2000,5000,10000
**
** Turning off interrupt moderation will also turn this off.
*/
static int
ixgbe_sysctl_dmac(SYSCTL_HANDLER_ARGS)
{
struct adapter *adapter = (struct adapter *) arg1;
struct ixgbe_hw *hw = &adapter->hw;
struct ifnet *ifp = adapter->ifp;
int error;
u16 oldval;
oldval = adapter->dmac;
error = sysctl_handle_int(oidp, &adapter->dmac, 0, req);
if ((error) || (req->newptr == NULL))
return (error);
switch (hw->mac.type) {
case ixgbe_mac_X550:
case ixgbe_mac_X550EM_x:
break;
default:
device_printf(adapter->dev,
"DMA Coalescing is only supported on X550 devices\n");
return (ENODEV);
}
switch (adapter->dmac) {
case 0:
/* Disabled */
break;
case 1: /* Enable and use default */
adapter->dmac = 1000;
break;
case 50:
case 100:
case 250:
case 500:
case 1000:
case 2000:
case 5000:
case 10000:
/* Legal values - allow */
break;
default:
/* Do nothing, illegal value */
adapter->dmac = oldval;
return (EINVAL);
}
/* Re-initialize hardware if it's already running */
if (ifp->if_drv_flags & IFF_DRV_RUNNING)
ixgbe_init(adapter);
return (0);
}
/*
* Sysctl to enable/disable the WoL capability, if supported by the adapter.
* Values:
* 0 - disabled
* 1 - enabled
*/
static int
ixgbe_sysctl_wol_enable(SYSCTL_HANDLER_ARGS)
{
struct adapter *adapter = (struct adapter *) arg1;
struct ixgbe_hw *hw = &adapter->hw;
int new_wol_enabled;
int error = 0;
new_wol_enabled = hw->wol_enabled;
error = sysctl_handle_int(oidp, &new_wol_enabled, 0, req);
if ((error) || (req->newptr == NULL))
return (error);
if (new_wol_enabled == hw->wol_enabled)
return (0);
if (new_wol_enabled > 0 && !adapter->wol_support)
return (ENODEV);
else
hw->wol_enabled = !!(new_wol_enabled);
return (0);
}
/*
* Sysctl to enable/disable the Energy Efficient Ethernet capability,
* if supported by the adapter.
* Values:
* 0 - disabled
* 1 - enabled
*/
static int
ixgbe_sysctl_eee_enable(SYSCTL_HANDLER_ARGS)
{
struct adapter *adapter = (struct adapter *) arg1;
struct ifnet *ifp = adapter->ifp;
int new_eee_enabled, error = 0;
new_eee_enabled = adapter->eee_enabled;
error = sysctl_handle_int(oidp, &new_eee_enabled, 0, req);
if ((error) || (req->newptr == NULL))
return (error);
if (new_eee_enabled == adapter->eee_enabled)
return (0);
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
if (new_eee_enabled > 0 && !adapter->eee_support)
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
return (ENODEV);
else
adapter->eee_enabled = !!(new_eee_enabled);
/* Re-initialize hardware if it's already running */
if (ifp->if_drv_flags & IFF_DRV_RUNNING)
ixgbe_init(adapter);
return (0);
}
/*
* Read-only sysctl indicating whether EEE support was negotiated
* on the link.
*/
static int
ixgbe_sysctl_eee_negotiated(SYSCTL_HANDLER_ARGS)
{
struct adapter *adapter = (struct adapter *) arg1;
struct ixgbe_hw *hw = &adapter->hw;
bool status;
status = !!(IXGBE_READ_REG(hw, IXGBE_EEE_STAT) & IXGBE_EEE_STAT_NEG);
return (sysctl_handle_int(oidp, 0, status, req));
}
/*
* Read-only sysctl indicating whether RX Link is in LPI state.
*/
static int
ixgbe_sysctl_eee_rx_lpi_status(SYSCTL_HANDLER_ARGS)
{
struct adapter *adapter = (struct adapter *) arg1;
struct ixgbe_hw *hw = &adapter->hw;
bool status;
status = !!(IXGBE_READ_REG(hw, IXGBE_EEE_STAT) &
IXGBE_EEE_RX_LPI_STATUS);
return (sysctl_handle_int(oidp, 0, status, req));
}
/*
* Read-only sysctl indicating whether TX Link is in LPI state.
*/
static int
ixgbe_sysctl_eee_tx_lpi_status(SYSCTL_HANDLER_ARGS)
{
struct adapter *adapter = (struct adapter *) arg1;
struct ixgbe_hw *hw = &adapter->hw;
bool status;
status = !!(IXGBE_READ_REG(hw, IXGBE_EEE_STAT) &
IXGBE_EEE_TX_LPI_STATUS);
return (sysctl_handle_int(oidp, 0, status, req));
}
/*
* Sysctl to enable/disable the types of packets that the
* adapter will wake up on upon receipt.
* WUFC - Wake Up Filter Control
* Flags:
* 0x1 - Link Status Change
* 0x2 - Magic Packet
* 0x4 - Direct Exact
* 0x8 - Directed Multicast
* 0x10 - Broadcast
* 0x20 - ARP/IPv4 Request Packet
* 0x40 - Direct IPv4 Packet
* 0x80 - Direct IPv6 Packet
*
* Setting another flag will cause the sysctl to return an
* error.
*/
static int
ixgbe_sysctl_wufc(SYSCTL_HANDLER_ARGS)
{
struct adapter *adapter = (struct adapter *) arg1;
int error = 0;
u32 new_wufc;
new_wufc = adapter->wufc;
error = sysctl_handle_int(oidp, &new_wufc, 0, req);
if ((error) || (req->newptr == NULL))
return (error);
if (new_wufc == adapter->wufc)
return (0);
if (new_wufc & 0xffffff00)
return (EINVAL);
else {
new_wufc &= 0xff;
new_wufc |= (0xffffff & adapter->wufc);
adapter->wufc = new_wufc;
}
return (0);
}
A few important fixes: - Testing TSO6 has led me to discover that HW RSC is a problematic feature, it is ONLY designed to work with IPv4 in the first place, and if IP forwarding is done it can't be disabled as LRO in the stack, also initial testing we've done at Intel shows an equal performance using TSO[46] on the TX and LRO on RX, if you ran older code on 82599 or later hardware you actually could have detrimental performance for this reason. So I am disabling the feature by default and all our adapters will now use LRO instead. - If you have flow control off and multiple queues it was possible when the buffer of one queue becomes full that all RX movement is stalled, to eliminate this problem a feature bit is now set that will allow packets to be dropped when full rather than stall. Note, the default is to have flow control on, and this keeps this from happening. - Because of the recent fixes in the stack, LRO is now auto-disabled when problematic, so I have decided to enable it by default in the capabilities in the driver. - There are some 1G modules used by some customers, a couple small tweaks to properly support those in the media code. - A note: we have now done some testing of TSO6 and using LRO with IPv6 and it all works great!! Seeing line rate in both directions in best cases. Thanks bz for your excellent work!!
2012-10-31 23:50:36 +00:00
/*
** Enable the hardware to drop packets when the buffer is
** full. This is useful when multiqueue,so that no single
** queue being full stalls the entire RX engine. We only
** enable this when Multiqueue AND when Flow Control is
** disabled.
*/
static void
ixgbe_enable_rx_drop(struct adapter *adapter)
{
struct ixgbe_hw *hw = &adapter->hw;
for (int i = 0; i < adapter->num_queues; i++) {
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
A few important fixes: - Testing TSO6 has led me to discover that HW RSC is a problematic feature, it is ONLY designed to work with IPv4 in the first place, and if IP forwarding is done it can't be disabled as LRO in the stack, also initial testing we've done at Intel shows an equal performance using TSO[46] on the TX and LRO on RX, if you ran older code on 82599 or later hardware you actually could have detrimental performance for this reason. So I am disabling the feature by default and all our adapters will now use LRO instead. - If you have flow control off and multiple queues it was possible when the buffer of one queue becomes full that all RX movement is stalled, to eliminate this problem a feature bit is now set that will allow packets to be dropped when full rather than stall. Note, the default is to have flow control on, and this keeps this from happening. - Because of the recent fixes in the stack, LRO is now auto-disabled when problematic, so I have decided to enable it by default in the capabilities in the driver. - There are some 1G modules used by some customers, a couple small tweaks to properly support those in the media code. - A note: we have now done some testing of TSO6 and using LRO with IPv6 and it all works great!! Seeing line rate in both directions in best cases. Thanks bz for your excellent work!!
2012-10-31 23:50:36 +00:00
srrctl |= IXGBE_SRRCTL_DROP_EN;
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(i), srrctl);
A few important fixes: - Testing TSO6 has led me to discover that HW RSC is a problematic feature, it is ONLY designed to work with IPv4 in the first place, and if IP forwarding is done it can't be disabled as LRO in the stack, also initial testing we've done at Intel shows an equal performance using TSO[46] on the TX and LRO on RX, if you ran older code on 82599 or later hardware you actually could have detrimental performance for this reason. So I am disabling the feature by default and all our adapters will now use LRO instead. - If you have flow control off and multiple queues it was possible when the buffer of one queue becomes full that all RX movement is stalled, to eliminate this problem a feature bit is now set that will allow packets to be dropped when full rather than stall. Note, the default is to have flow control on, and this keeps this from happening. - Because of the recent fixes in the stack, LRO is now auto-disabled when problematic, so I have decided to enable it by default in the capabilities in the driver. - There are some 1G modules used by some customers, a couple small tweaks to properly support those in the media code. - A note: we have now done some testing of TSO6 and using LRO with IPv6 and it all works great!! Seeing line rate in both directions in best cases. Thanks bz for your excellent work!!
2012-10-31 23:50:36 +00:00
}
}
static void
ixgbe_disable_rx_drop(struct adapter *adapter)
{
struct ixgbe_hw *hw = &adapter->hw;
for (int i = 0; i < adapter->num_queues; i++) {
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
u32 srrctl = IXGBE_READ_REG(hw, IXGBE_SRRCTL(i));
A few important fixes: - Testing TSO6 has led me to discover that HW RSC is a problematic feature, it is ONLY designed to work with IPv4 in the first place, and if IP forwarding is done it can't be disabled as LRO in the stack, also initial testing we've done at Intel shows an equal performance using TSO[46] on the TX and LRO on RX, if you ran older code on 82599 or later hardware you actually could have detrimental performance for this reason. So I am disabling the feature by default and all our adapters will now use LRO instead. - If you have flow control off and multiple queues it was possible when the buffer of one queue becomes full that all RX movement is stalled, to eliminate this problem a feature bit is now set that will allow packets to be dropped when full rather than stall. Note, the default is to have flow control on, and this keeps this from happening. - Because of the recent fixes in the stack, LRO is now auto-disabled when problematic, so I have decided to enable it by default in the capabilities in the driver. - There are some 1G modules used by some customers, a couple small tweaks to properly support those in the media code. - A note: we have now done some testing of TSO6 and using LRO with IPv6 and it all works great!! Seeing line rate in both directions in best cases. Thanks bz for your excellent work!!
2012-10-31 23:50:36 +00:00
srrctl &= ~IXGBE_SRRCTL_DROP_EN;
Revert last commit, to remove added skeleton tree.
2015-06-01 17:35:29 +00:00
IXGBE_WRITE_REG(hw, IXGBE_SRRCTL(i), srrctl);
A few important fixes: - Testing TSO6 has led me to discover that HW RSC is a problematic feature, it is ONLY designed to work with IPv4 in the first place, and if IP forwarding is done it can't be disabled as LRO in the stack, also initial testing we've done at Intel shows an equal performance using TSO[46] on the TX and LRO on RX, if you ran older code on 82599 or later hardware you actually could have detrimental performance for this reason. So I am disabling the feature by default and all our adapters will now use LRO instead. - If you have flow control off and multiple queues it was possible when the buffer of one queue becomes full that all RX movement is stalled, to eliminate this problem a feature bit is now set that will allow packets to be dropped when full rather than stall. Note, the default is to have flow control on, and this keeps this from happening. - Because of the recent fixes in the stack, LRO is now auto-disabled when problematic, so I have decided to enable it by default in the capabilities in the driver. - There are some 1G modules used by some customers, a couple small tweaks to properly support those in the media code. - A note: we have now done some testing of TSO6 and using LRO with IPv6 and it all works great!! Seeing line rate in both directions in best cases. Thanks bz for your excellent work!!
2012-10-31 23:50:36 +00:00
}
}
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
static void
ixgbe_rearm_queues(struct adapter *adapter, u64 queues)
{
u32 mask;
switch (adapter->hw.mac.type) {
case ixgbe_mac_82598EB:
mask = (IXGBE_EIMS_RTX_QUEUE & queues);
IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS, mask);
break;
case ixgbe_mac_82599EB:
case ixgbe_mac_X540:
case ixgbe_mac_X550:
Add support for certain Intel X550 devices. These include standalone X550 adapters, X552 10GbE backplane, and X552/X557-AT 10GBASE-T; with the latter two being integrated into Xeon D SoCs. As well, this bumps the ixgbe version number to 2.8.3, and includes updates to shared code for support for the new devices. Differential Revision: D2414 Reviewed by: gnn, adrian Approved by: jfv (mentor), gnn (mentor)
2015-04-30 22:53:27 +00:00
case ixgbe_mac_X550EM_x:
Update to the Intel ixgbe driver: - Split the driver into independent pf and vf loadables. This is in preparation for SRIOV support which will be following shortly. This also allows us to keep a seperate revision control over the two parts, making for easier sustaining. - Make the TX/RX code a shared/seperated file, in the old code base the ixv code would miss fixes that went into ixgbe, this model will eliminate that problem. - The driver loadables will now match the device names, something that has been requested for some time. - Rather than a modules/ixgbe there is now modules/ix and modules/ixv - It will also be possible to make your static kernel with only one or the other for streamlined installs, or both. Enjoy! Submitted by: jfv and erj
2015-03-17 18:32:28 +00:00
mask = (queues & 0xFFFFFFFF);
IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(0), mask);
mask = (queues >> 32);
IXGBE_WRITE_REG(&adapter->hw, IXGBE_EICS_EX(1), mask);
break;
default:
break;
}
}
Reference in New Issue Copy Permalink