When link goes down, disable the port's Rx path to drop the incoming
traffic closer to the wire, instead of accepting them in for further
Rx processing, only to eventually drop them at the port's RxQs. This
prevents unnecessary congestion in the Rx path. The port's Rx path
will be re-enabled once the link up event is received in the firmware
event queue.
Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
If there is no VXLAN among pattern items, the action will be turned down.
Signed-off-by: Ivan Malov <ivan.malov@oktetlabs.ru>
Reviewed-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Reviewed-by: Andy Moreton <amoreton@xilinx.com>
Provide necessary facilities for handling this action.
Signed-off-by: Ivan Malov <ivan.malov@oktetlabs.ru>
Reviewed-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Reviewed-by: Andy Moreton <amoreton@xilinx.com>
Current API signature makes it hard to add other entities that
belong to a flow specification. Pass the flow specification
so that additional members can be accessed through the spec.
Signed-off-by: Igor Romanov <igor.romanov@oktetlabs.ru>
Signed-off-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Reviewed-by: Andy Moreton <amoreton@xilinx.com>
Reviewed-by: Ivan Malov <ivan.malov@oktetlabs.ru>
In this commit we generalize the movement of user-specified
meta data between mbufs and FPGA AXIS tuser fields using
user-defined hook functions.
- Previous use of PMD dynfields are removed
- Remove emptied rte_pmd_ark.h
- Hook function added to ark_user_ext
- Add hook function calls in Rx and Tx paths
- Update guide with example of hook function use
Signed-off-by: Ed Czeck <ed.czeck@atomicrules.com>
- New PCIe IDs using net/ark driver
- Update Version IDs and structures specified by hardware
- New internal descriptor status for TX
- Adjust data placement in RX operations, headroom in retained for
segmented mbufs
Signed-off-by: Ed Czeck <ed.czeck@atomicrules.com>
Allocate mbufs for Rx path in bulk of at least 64 buffers
to improve performance. Allow recovery even without
a Rx operation to support lack of buffers in pool.
Fixes: be410a861598 ("net/ark: add recovery for lack of mbufs")
Cc: stable@dpdk.org
Signed-off-by: Ed Czeck <ed.czeck@atomicrules.com>
Fixes: b33ccdb17f55 ("net/ark: provide API for hardware modules MPU RQP and pktdir")
Cc: stable@dpdk.org
Signed-off-by: Ed Czeck <ed.czeck@atomicrules.com>
Rename the key opaque pointer from rte_tls_key to
rte_thread_key to avoid confusion with transport layer security.
Also rename and remove the "_tls" term from the following
functions to avoid redundancy:
rte_thread_tls_key_create
rte_thread_tls_key_delete
rte_thread_tls_value_set
rte_thread_tls_value_get
Suggested-by: Vladimir Medvedkin <vladimir.medvedkin@intel.com>
Suggested-by: Morten Brørup <mb@smartsharesystems.com>
Signed-off-by: Tal Shnaiderman <talshn@nvidia.com>
Acked-by: Morten Brørup <mb@smartsharesystems.com>
Improve portability (especially with musl libc)
by replacing the non-standard type 'uint' with 'size_t'.
Fixes: 746664d546fb ("net/igc: support flow API")
Cc: stable@dpdk.org
Suggested-by: David Marchand <david.marchand@redhat.com>
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Haiyue Wang <haiyue.wang@intel.com>
Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Acked-by: David Marchand <david.marchand@redhat.com>
Improve portability by replacing non-standard 'uint' with 'unsigned int'.
This solves the build error with musl libc:
In file included from ../drivers/net/cxgbe/cxgbe.h:9,
from ../drivers/net/cxgbe/cxgbe_ethdev.c:37:
../drivers/net/cxgbe/base/common.h:201:4: error: unknown type name 'uint'
201 | uint synmapen:1; /* SYN Map Enable */
| ^~~~
Fixes: bfcb257d3014 ("net/cxgbe: enable RSS for VF")
Cc: stable@dpdk.org
Signed-off-by: Natanael Copa <ncopa@alpinelinux.org>
Reviewed-by: Morten Brørup <mb@smartsharesystems.com>
Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Acked-by: David Marchand <david.marchand@redhat.com>
The page size is often retrieved from the macro PAGE_SIZE.
If PAGE_SIZE is not defined, it is either using hard coded default,
or getting the system value from the UNIX-only function sysconf().
Such definitions are replaced with the generic function
rte_mem_page_size() defined for each supported OS.
Removing PAGE_SIZE definitions will fix dlb drivers for musl libc,
because #ifdef checks were missing, causing redefinition errors.
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Andrew Boyer <aboyer@pensando.io>
Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Acked-by: David Marchand <david.marchand@redhat.com>
Acked-by: Timothy McDaniel <timothy.mcdaniel@intel.com>
The feature macro _GNU_SOURCE is defined globally,
but there was some remaining useless settings.
The internal definition in config/meson.build is kept,
all other internal definitions of _GNU_SOURCE are removed,
except in examples, which can be built as external applications.
Note: external applications do not inherit of _GNU_SOURCE.
Fixes: 5d7b673d5fd6 ("mk: build with _GNU_SOURCE defined by default")
Fixes: 28188cee2aa0 ("build: enable BSD features visibility for FreeBSD")
Fixes: e6cdc54cc0ef ("net/mlx5: add socket server for external tools")
Cc: stable@dpdk.org
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Acked-by: David Marchand <david.marchand@redhat.com>
The Rx metadata might use the metadata register C0 to keep the
values. The same register C0 might be used by kernel for source
vport value handling, kernel uses upper half of the register,
leaving the lower half for application usage.
In the extended metadata mode 1 (dv_xmeta_en devarg is
assigned with value 1) the metadata width is 16 bits only,
the Rx datapath code fetched the entire 32-bit value of the
metadata register and presented one to application. The patch
provides data masking depending on the chosen metadata mode.
Fixes: 6c55b622a956 ("net/mlx5: set dynamic flow metadata in Rx queues")
Cc: stable@dpdk.org
Signed-off-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
This patch adds support for the timestamp format settings for
the receive and send queues. If the firmware version x.30.1000
or above is installed and the NIC timestamps are configured
with the real-time format, the default zero values for newly
added fields cause the queue creation to fail.
The patch queries the timestamp formats supported by the hardware
and sets the configuration values in queue context accordingly.
Fixes: 86fc67fc9315 ("net/mlx5: create advanced RxQ object via DevX")
Fixes: ae18a1ae9692 ("net/mlx5: support Tx hairpin queues")
Fixes: 15c3807e86ab ("common/mlx5: support DevX QP operations")
Cc: stable@dpdk.org
Signed-off-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
Acked-by: Ori Kam <orika@nvidia.com>
Having to force an alignment does not impact the user,
so it should not be a warning.
The log level is reduced from warning to debug.
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Matan Azrad <matan@nvidia.com>
The macro DRV_LOG already includes a terminating line feed character
defined in PMD_DRV_LOG_.
The extra line feeds added in some messages are removed.
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Matan Azrad <matan@nvidia.com>
Most debug logs are using DRV_LOG(DEBUG,)
but some were using DEBUG().
The macro DEBUG is doing nothing if not compiled with
RTE_LIBRTE_MLX5_DEBUG.
As it is not used in the data path, the macro DEBUG
can be replaced with DRV_LOG.
Then all debug logs can be enabled at runtime with:
--log-level pmd.net.mlx5:debug
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Matan Azrad <matan@nvidia.com>
The macro DEBUG was doing nothing if not compiled with
RTE_LIBRTE_MLX4_DEBUG.
As it is not used in the data path, it can be always enabled at
compilation time. Then it can be enabled at runtime with:
--log-level pmd.net.mlx4:debug
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Matan Azrad <matan@nvidia.com>
To support more representor type, this patch introduces representor type
enum. The enum is subject to be extended to support new representor in
patches upcoming.
For each devarg structure, only one type supported.
Signed-off-by: Xueming Li <xuemingl@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
Acked-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Acked-by: Hyong Youb Kim <hyonkim@cisco.com>
Move function to get efx family from net driver into common driver.
Signed-off-by: Vijay Kumar Srivastava <vsrivast@xilinx.com>
Signed-off-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Driver would be probed only for the net device class.
Signed-off-by: Vijay Kumar Srivastava <vsrivast@xilinx.com>
Signed-off-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
timesync adjust and write_time APIs needed to account for the Rx and Tx
timestamp counters as well. Fix it since it was not done earlier.
Fixes: b11cceb83a34 ("net/bnxt: support timesync")
Cc: stable@dpdk.org
Reviewed-by: Lance Richardson <lance.richardson@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Somnath Kotur <somnath.kotur@broadcom.com>
Fix to return count in xstats get op in all cases.
Driver was returning 0 if the 'xstats' parameter being passed to
xstats_get_op was NULL. This won't work on some applications that
rely on a valid count being passed even in this case so that it can
allocate memory accordingly followed by a reissue of the xstats_get_op
to get the actual stats populated by the driver.
Fixes: 063e59ddd28e ("net/bnxt: fix crash in xstats get")
Cc: stable@dpdk.org
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Lance Richardson <lance.richardson@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Somnath Kotur <somnath.kotur@broadcom.com>
Address build issues with Clang and without glibc on ppc64le.
Vector can be a keyword and should not be used in code.
Renaming it to avoid conflict.
Reported-by: Piotr Kubaj <pkubaj@freebsd.org>
Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
In the init path, driver ignores few of the HWRM command failures.
There is no need to log the error message in those cases.
Fixes: 3fb93bc7c349 ("net/bnxt: initialize parent PF information")
Fixes: 4e3f887bec4b ("net/bnxt: support HWRM port PHY qcaps")
Cc: stable@dpdk.org
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Fix to return an error when the HWRM version that the driver
is compiled against is incompatible with the FW that is actually
running on the card. This is determined based on the req length
indicated by FW against the value supported in the HWRM.
Fixes: 804e746c7b73 ("net/bnxt: add hardware resource manager init code")
Cc: stable@dpdk.org
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
1. Renamed bnxt_hwrm_alloc_vf_info()/bnxt_hwrm_free_vf_info to
bnxt_alloc_vf_info()/bnxt_free_vf_info as it does not
issue any HWRM command to fw.
2. Fix missing unlock when memory allocation fails.
Fixes: b7778e8a1c00 ("net/bnxt: refactor to properly allocate resources for PF/VF")
Cc: stable@dpdk.org
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Fix HWRM_VER_GET command to handle DEV_NOT_RDY state.
Driver should fail probe if the device is not ready.
Conversely, the HWRM_VER_GET poll after reset can safely
retry until the existing timeout is exceeded.
Fixes: 804e746c7b73 ("net/bnxt: add hardware resource manager init code")
Cc: stable@dpdk.org
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Reviewed-by: Randy Schacher <stuart.schacher@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
When host SW issues a HWRM_RING_FREE for Tx/Rx/AGG ring in HW,
the FW flushes the BDs associated with the ring and performs other
cleanup in the HW. The host software should ideally check for an
indication from the FW indicating this step has been completed
successfully to avoid unexpected errors during cleanup.
The FW issues a HWRM_DONE response to the RING_FREE request on
the corresponding CQ ring. Poll the CQs during cleanup and
ensure the HWRM_FREE command is completed not just based on the
value of valid bit but also the HWRM_DONE response for the ring.
If the HWRM_DONE response is not seen, force the cleanup to
complete just based on the valid bit.
Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Remove code for setting consumed mbuf pointers to NULL from the
vector receive functions as a minor performance optimization.
Signed-off-by: Lance Richardson <lance.richardson@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Fix a number of issues in the bnxt receive descriptor status
function, including:
- Provide status of receive descriptor instead of completion
descriptor.
- Remove invalid comparison of raw ring index with masked ring
index.
- Correct misinterpretation of offset parameter as ring index.
- Correct misuse of completion ring index for mbuf ring (the
two rings have different sizes).
Fixes: 0fe613bb87b2 ("net/bnxt: support Rx descriptor status")
Cc: stable@dpdk.org
Signed-off-by: Lance Richardson <lance.richardson@broadcom.com>
Reviewed-by: Andy Gospodarek <gospo@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
On Thor, Rx timestamp is present in the Rx completion record.
Only 32 bits of the timestamp is present in the completion.
The driver needs to periodically poll the current 48 bit
free running timer using the HWRM_PORT_TS_QUERY command.
It can combine the upper 16 bits from the HWRM response
with the lower 32 bits in the Rx completion to produce
the 48 bit timestamp for the Rx packet.
This patch adds an alarm thread to periodically poll the current 48 bit
free running timer using the HWRM_PORT_TS_QUERY command.
This avoids issuing the hwrm command from the Rx handler.
This patch also handles the timer roll over condition.
Fixes: 6cbd89f9f3d8 ("net/bnxt: support PTP for Thor")
Cc: stable@dpdk.org
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Lance Richardson <lance.richardson@broadcom.com>
Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Moved fw readiness check to a new routine bnxt_check_fw_ready().
During error recovery, driver needs to wait for fw readiness.
For that, it uses bnxt_hwrm_ver_get() function now and that
function does parsing of the VER_GET response as well.
Added a new lightweight function bnxt_hwrm_poll_ver_get() for polling
the firmware readiness which issues VER_GET and checks for success
without processing the command response.
Fixes: df6cd7c1f73a ("net/bnxt: handle reset notify async event from FW")
Cc: stable@dpdk.org
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
During some fatal firmware error conditions, the PCI config space
register 0x2e which normally contains the subsystem ID will become
0xffff. This register will revert back to the normal value after
the chip has completed core reset. If we detect this condition,
we can poll this config register immediately for the value to revert.
Because we use config read cycles to poll this register, there is no
possibility of Master Abort if we happen to read it during core reset.
This speeds up recovery significantly as we don't have to wait for the
conservative min_time before polling to see if the firmware has come
out of reset. As soon as this register changes value we can proceed
to re-initialize the device.
Fixes: df6cd7c1f73a ("net/bnxt: handle reset notify async event from FW")
Cc: stable@dpdk.org
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
This is a new async message that the firmware can send to check if it
can communicate with the driver. This is an added error detection
scheme that firmware can use if it suspects errors in the PCIe
interface. When the driver receives this async message, it will reply
back echoing some data in the async message. If the firmware is not
getting the reply with the proper data after some retries, error
recovery will kick in.
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Lance Richardson <lance.richardson@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
1. Used port id in async event logs.
2. Added a debug log in bnxt_hwrm_func_driver_unregister().
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Update HWRM headers to version 1.10.2.15
which updates the backing store API for additional TQM rings.
Add support for 9th TQM ring using latest firmware interface.
Also make sure that we set only necessary bits in the enables
field in backing store request.
Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Brought in the latest hsi_struct_def_dpdk.h.
HWRM API is now updated to version 1.10.2.15.
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Update queues per VNIC in single queue mode.
bp->rx_num_qs_per_vnic is not initialized in the single queue mode.
As a result of this when an interface is reconfigured to single
queue mode from an existing multiqueue mode, bp->rx_num_qs_per_vnic
is not updated to the value of 1. Hence, the driver will try to
access more than one queue resulting in a crash.
This patch fixes it by initializing bp->rx_num_qs_per_vnic in the
single queue mode as well.
Fixes: 36024b2e7fe5 ("net/bnxt: allow dynamic creation of VNIC")
Cc: stable@dpdk.org
Signed-off-by: Venkat Duvvuru <venkatkumar.duvvuru@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Removed an unnecessary extra blank line.
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
PMD should not set any flags to receive RoCE traffic while
configuring the vnic. Since the PMD does not support RoCE
some of the flags and code is unused. Clean it up.
Fixes: b7778e8a1c00 ("net/bnxt: refactor to properly allocate resources for PF/VF")
Cc: stable@dpdk.org
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
This commit introduces support for preferred busy polling
to the AF_XDP PMD. This feature aims to improve single-core
performance for AF_XDP sockets under heavy load.
A new vdev arg is introduced called 'busy_budget' whose default
value is 64. busy_budget is the value supplied to the kernel
with the SO_BUSY_POLL_BUDGET socket option and represents the
busy-polling NAPI budget. To set the budget to a different value
eg. 256:
--vdev=net_af_xdp0,iface=eth0,busy_budget=256
Preferred busy polling is enabled by default provided a kernel with
version >= v5.11 is in use. To disable it, set the budget to zero.
The following settings are also strongly recommended to be used in
conjunction with this feature:
echo 2 | sudo tee /sys/class/net/eth0/napi_defer_hard_irqs
echo 200000 | sudo tee /sys/class/net/eth0/gro_flush_timeout
.. where eth0 is the interface being used by the PMD.
Signed-off-by: Ciara Loftus <ciara.loftus@intel.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
