dpdk.git
3 years agonet/hns3: remove meaningless packet buffer rollback
Huisong Li [Sat, 15 May 2021 00:52:35 +0000 (08:52 +0800)]
net/hns3: remove meaningless packet buffer rollback

Packet buffer allocation and hardware pause configuration fail normally
when a reset occurs. If the execution fails, rollback of the packet
buffer still fails. So this rollback is meaningless.

Fixes: 62e3ccc2b94c ("net/hns3: support flow control")
Cc: stable@dpdk.org
Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agonet/hns3: fix requested FC mode rollback
Huisong Li [Sat, 15 May 2021 00:52:34 +0000 (08:52 +0800)]
net/hns3: fix requested FC mode rollback

Currently, the "requested_fc_mode" lacks rollback when enabling link
FC or PFC fails.
For example, this may result an incorrect FC mode after a reset.

Fixes: d4fdb71a0e7b ("net/hns3: fix flow control mode")
Fixes: 62e3ccc2b94c ("net/hns3: support flow control")
Cc: stable@dpdk.org
Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agonet/hns3: fix Rx/Tx queue numbers check
Huisong Li [Sat, 15 May 2021 00:52:33 +0000 (08:52 +0800)]
net/hns3: fix Rx/Tx queue numbers check

The Rx/Tx queue numbers should be greater than TC number, this patch adds
this check for PF before updating the mapping between TC and queue.

Fixes: a951c1ed3ab5 ("net/hns3: support different numbers of Rx and Tx queues")
Fixes: 76d794566d43 ("net/hns3: maximize queue number")
Cc: stable@dpdk.org
Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agovdpa/mlx5: fix device unplug
Matan Azrad [Thu, 13 May 2021 18:40:20 +0000 (21:40 +0300)]
vdpa/mlx5: fix device unplug

The vDPA PCI device unplug process should release all the private
device resources and also to unregister the device.

The device unregistration was missed what remained the device data
invalid in the rte_vhost library.

Unregister the device in unplug process via the remove operation.

Fixes: 95276abaaf0a ("vdpa/mlx5: introduce Mellanox vDPA driver")
Cc: stable@dpdk.org
Reported-by: Eli Britstein <elibr@nvidia.com>
Signed-off-by: Matan Azrad <matan@nvidia.com>
Tested-by: Eli Britstein <elibr@nvidia.com>
Acked-by: Xueming Li <xuemingl@nvidia.com>
Reviewed-by: Chenbo Xia <chenbo.xia@intel.com>
3 years agonet/vhost: restore pseudo TSO support
David Marchand [Tue, 18 May 2021 07:07:27 +0000 (09:07 +0200)]
net/vhost: restore pseudo TSO support

The net/vhost PMD does not comply with the ethdev offload API as it does
not report Rx/Tx offload capabilities wrt TSO and checksum offloading.
On the other hand, the net/vhost PMD lets guest negotiates TSO and
checksum offloading.

Changing the behavior for Rx/Tx offload flags handling won't
improve/fix this situation and will break applications that might have
been relying on implicit support of TSO in this driver.

Revert this behavior change until we have a complete fix.

Fixes: ca7036b4af3a ("vhost: fix offload flags in Rx path")
Cc: stable@dpdk.org
Signed-off-by: David Marchand <david.marchand@redhat.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
3 years agovhost: restore IOTLB mempool allocation
David Marchand [Mon, 17 May 2021 08:59:51 +0000 (10:59 +0200)]
vhost: restore IOTLB mempool allocation

IOTLB messages will be sent when some queues are not enabled. If we
initialize IOTLB in vhost_user_set_vring_num, it could happen that IOTLB
update comes when IOTLB pool of disabled queues are not initialized.

Fixes: 968bbc7e2e50 ("vhost: avoid IOTLB mempool allocation while IOMMU disabled")

Signed-off-by: David Marchand <david.marchand@redhat.com>
Reviewed-by: Chenbo Xia <chenbo.xia@intel.com>
3 years agovhost: fix stored last used index
Balazs Nemeth [Fri, 14 May 2021 16:02:07 +0000 (18:02 +0200)]
vhost: fix stored last used index

The optimization introduced by
commit d18db8049c7c ("vhost: read last used index once")
didn't account for the fact that vhost_flush_enqueue_shadow_packed
increments the last_used_idx.
For this reason, store last_used_idx after the potential call to
vhost_flush_enqueue_shadow_packed.

Bugzilla ID: 699
Fixes: d18db8049c7c ("vhost: read last used index once")

Signed-off-by: Balazs Nemeth <bnemeth@redhat.com>
Reviewed-by: David Marchand <david.marchand@redhat.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Tested-by: Wei Ling <weix.ling@intel.com>
3 years agonet/mlx5: fix metering cleanup on stop
Li Zhang [Thu, 13 May 2021 08:35:44 +0000 (11:35 +0300)]
net/mlx5: fix metering cleanup on stop

A meter may handle Rx queue reference in his sub-policies.
In stop operation, all the Rx queues are released.

Wrongly, the meter reference was not released before
destroying the Rx queues what cause an error in stop.

Release the Rx queues meter references in stop operation.

Fixes: fc6ce56bba05 ("net/mlx5: prepare sub-policy for flow with meter")

Signed-off-by: Li Zhang <lizh@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
3 years agonet/mlx5: fix counter offset detection
Suanming Mou [Thu, 13 May 2021 08:05:15 +0000 (11:05 +0300)]
net/mlx5: fix counter offset detection

Currently, the counter offset support is discovered by creating the
rule with invalid offset counter and drop action in root table. If
the rule creation fails with EINVAL errno, that mean counter offset
is not supported in root table.

However, drop action may not be supported in some rdma-core version
in root table. In this case, the discover code will not work properly.

This commits changes flow attribute to egress. That removes all the
extra fate actions in the flow to avoid any unsupported fate actions
make the discover code fail time to time.

Fixes: 994829e695c0 ("net/mlx5: remove single counter container")
Cc: stable@dpdk.org
Signed-off-by: Suanming Mou <suanmingm@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
3 years agonet/mlx5: fix leak when configured repeatedly
Yunjian Wang [Thu, 13 May 2021 03:39:54 +0000 (11:39 +0800)]
net/mlx5: fix leak when configured repeatedly

Currently, configuring a mlx device, it will allocate its
own process private in mlx5_proc_priv_init() and only frees
it when closing the device. This will lead to a memory leak,
when a device is configured repeatedly.

For example:
for(...)
do
    rte_eth_dev_configure
    rte_eth_rx_queue_setup
    rte_eth_tx_queue_setup
    rte_eth_dev_start
    rte_eth_dev_stop
done

Fixes: 120dc4a7dcd3 ("net/mlx5: remove device register remap")
Cc: stable@dpdk.org
Signed-off-by: Yunjian Wang <wangyunjian@huawei.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
3 years agonet/mlx4: fix leak when configured repeatedly
Yunjian Wang [Thu, 13 May 2021 03:38:56 +0000 (11:38 +0800)]
net/mlx4: fix leak when configured repeatedly

Currently, configuring a mlx device, it will allocate its
own process private in mlx5_proc_priv_init() and only frees
it when closing the device. This will lead to a memory leak,
when a device is configured repeatedly.

For example:
for(...)
do
    rte_eth_dev_configure
    rte_eth_rx_queue_setup
    rte_eth_tx_queue_setup
    rte_eth_dev_start
    rte_eth_dev_stop
done

Fixes: 97d37d2c1f6b ("net/mlx4: remove device register remap")
Cc: stable@dpdk.org
Signed-off-by: Yunjian Wang <wangyunjian@huawei.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
3 years agonet/nfp: fix internal port addressing
Heinrich Kuhn [Fri, 14 May 2021 09:19:34 +0000 (11:19 +0200)]
net/nfp: fix internal port addressing

Depending on the breakout mode of the physical ports the internal NFP
port number might differ from the actual physical port number. Prior to
this patch the physical port number was used when making configuration
changes to the physical ports (enable, admin up etc). After this change
the internal port number is now correctly used for configuration
changes.

Fixes: 5e15e799d697 ("net/nfp: create separate entity for PF device")
Cc: stable@dpdk.org
Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
Signed-off-by: Simon Horman <simon.horman@netronome.com>
3 years agodoc: fix release notes for QAT compression PMD
Adam Dybkowski [Thu, 13 May 2021 12:05:42 +0000 (13:05 +0100)]
doc: fix release notes for QAT compression PMD

This patch extends the DPDK 21.05 release notes with the QAT GEN3
compression update.

Fixes: da573c0e4205 ("compress/qat: enable compression on GEN3")

Signed-off-by: Adam Dybkowski <adamx.dybkowski@intel.com>
Acked-by: Fan Zhang <roy.fan.zhang@intel.com>
3 years agocrypto/zuc: fix build with GCC 11
Kevin Traynor [Fri, 14 May 2021 15:08:34 +0000 (16:08 +0100)]
crypto/zuc: fix build with GCC 11

GCC 11 complains that some arrays may be uninitialized in
process_zuc_hash_op(). This is because their initialization
depends on num_ops being > 0.

This function is only called with num_ops > 0 because of
checks in process_zuc_hash_op().

To remove the warning initialize the arrays.

Fixes: 0b133c36ad7d ("crypto/zuc: support IPsec Multi-buffer lib v0.54")
Cc: stable@dpdk.org
Signed-off-by: Kevin Traynor <ktraynor@redhat.com>
3 years agotest/crypto: fix build with GCC 11
Kevin Traynor [Wed, 5 May 2021 08:53:14 +0000 (09:53 +0100)]
test/crypto: fix build with GCC 11

Fix the allocation for sessions, to prevent an array-bounds
warning with GCC 11. Set the not created session to NULL.

Fixes: 202d375c60bc ("app/test: add cryptodev unit and performance tests")
Cc: stable@dpdk.org
Signed-off-by: Kevin Traynor <ktraynor@redhat.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
3 years agodoc: remove obsolete future considerations in flow guide
Thomas Monjalon [Mon, 29 Mar 2021 16:52:14 +0000 (18:52 +0200)]
doc: remove obsolete future considerations in flow guide

After 4 years, rte_flow has evolved enough to not require
special notes about what could be added in future.
Part of the removed plans were obsolete anyway.

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Ori Kam <orika@nvidia.com>
3 years agodoc: update minimum supported Linux kernel
Stephen Hemminger [Mon, 19 Apr 2021 22:06:17 +0000 (15:06 -0700)]
doc: update minimum supported Linux kernel

The DPDK project is only committed to supporting upstream kernel
versions that are still in support. The 3.16 kernel has reached
End Of Life (in June 2020). The next LTS kernel is 4.4 and is
supported until February 2020.

This does not change the existing policy that distribution kernels
that are older are still supported.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Thomas Monjalon <thomas@monjalon.net>
3 years agodoc: fix description of some mlx5 features
Asaf Penso [Sun, 16 May 2021 10:56:20 +0000 (10:56 +0000)]
doc: fix description of some mlx5 features

The support of the new RTE_FLOW_ITEM_TYPE_INTEGRITY
was added in the release notes 21.02 by mistake.

The support of the Sub-Function representors was missing
in the release notes and the mlx5 guide.

Fixes: 79f8952783d0 ("net/mlx5: support integrity flow item")
Fixes: cb95feefdd03 ("net/mlx5: support sub-function representor")

Signed-off-by: Asaf Penso <asafp@nvidia.com>
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
3 years agodoc: add tested Intel platforms with Intel NICs
Yan Xia [Wed, 19 May 2021 08:55:18 +0000 (16:55 +0800)]
doc: add tested Intel platforms with Intel NICs

Add tested Intel platforms with Intel NICs to v21.05 release note.

Signed-off-by: Yan Xia <yanx.xia@intel.com>
3 years agodoc: add tested platforms with Mellanox NICs
Raslan Darawsheh [Mon, 17 May 2021 12:39:37 +0000 (15:39 +0300)]
doc: add tested platforms with Mellanox NICs

Add tested platforms with Mellanox NICs to the 21.05 release notes.

Signed-off-by: Raslan Darawsheh <rasland@nvidia.com>
3 years agobuild: check symbol maps in developer mode
David Marchand [Thu, 13 May 2021 08:34:15 +0000 (10:34 +0200)]
build: check symbol maps in developer mode

Hook check-symbol-maps.sh in the symbol check when in developer mode to
help developers catch issues before submitting their changes.

Signed-off-by: David Marchand <david.marchand@redhat.com>
Reviewed-by: Ray Kinsella <mdr@ashroe.eu>
3 years agodevtools: catch symbol duplicates in version map
David Marchand [Thu, 13 May 2021 08:34:14 +0000 (10:34 +0200)]
devtools: catch symbol duplicates in version map

Add a check on versioned symbol duplicates in map files.

Signed-off-by: David Marchand <david.marchand@redhat.com>
3 years agodevtools: fix orphan symbols check with busybox
David Marchand [Thu, 13 May 2021 08:34:13 +0000 (10:34 +0200)]
devtools: fix orphan symbols check with busybox

Avoid relying on GNU grep --exclude option.

Fixes: f8ad40dc998c ("devtools: check orphan symbols in map files")
Cc: stable@dpdk.org
Signed-off-by: David Marchand <david.marchand@redhat.com>
3 years agodevtools: add acronyms in dictionary for commit checks
Ferruh Yigit [Fri, 14 May 2021 09:03:08 +0000 (10:03 +0100)]
devtools: add acronyms in dictionary for commit checks

GCC     -> GNU Compiler Collection
ID      -> Identification/Identity/Identifier
IP      -> Internet Protocol
QinQ    -> IEEE 802.1Q in 802.1Q
SoC     -> System on a Chip
VEB     -> Virtual Ethernet Bridge
Windows ->

Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
3 years agoconfig/arm: add Qualcomm Centriq 2400
Thierry Herbelot [Fri, 7 May 2021 08:48:15 +0000 (10:48 +0200)]
config/arm: add Qualcomm Centriq 2400

From the documentation:
"The SoC configuration is a combination of implementer and CPU part number
 configuration and SoC-specific configuration."

Align Qualcomm SoC configuration with the configuration of other
server SoCs (eMAG, Kunpeng 9x0): add a soc configuration to the existing
implementer configuration, and a cross file.

Signed-off-by: Thierry Herbelot <thierry.herbelot@6wind.com>
Acked-by: Ruifeng Wang <ruifeng.wang@arm.com>
3 years agousertools: add Intel DLB device binding
Timothy McDaniel [Mon, 26 Apr 2021 15:34:50 +0000 (10:34 -0500)]
usertools: add Intel DLB device binding

Add DLB to usertools/dpdk-devbind.py so that it shows up
as an eventdev, and is identified as Intel DLB.

Signed-off-by: Timothy McDaniel <timothy.mcdaniel@intel.com>
3 years agotest/timer: remove license boilerplate
Stephen Hemminger [Tue, 16 Mar 2021 21:00:27 +0000 (14:00 -0700)]
test/timer: remove license boilerplate

This file has redundant BSD license text that is already
replaced by the use of SPDX license id.
Having both is unnecessary and potentially confusing.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
3 years agodoc: add SPDX tag to custom CSS file
Stephen Hemminger [Tue, 16 Mar 2021 21:00:26 +0000 (14:00 -0700)]
doc: add SPDX tag to custom CSS file

Trivial file, but add SPDX license tag per policy.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
3 years agodevtools: ignore JSON files for SPDX check
Stephen Hemminger [Tue, 16 Mar 2021 21:00:25 +0000 (14:00 -0700)]
devtools: ignore JSON files for SPDX check

The JSON file format does not support comments so there is no good
way to add SPDX license identifier.  This solves false positives
that arrive from the use of JSON in crypto dev tests.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
3 years agocommon/mlx5: add SPDX tag in Windows headers
Stephen Hemminger [Tue, 16 Mar 2021 21:00:24 +0000 (14:00 -0700)]
common/mlx5: add SPDX tag in Windows headers

These files were added without required SPDX headers.

Fixes: 7525ebd8ebb0 ("common/mlx5: add glue functions on Windows")
Cc: stable@dpdk.org
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Tal Shnaiderman <talshn@nvidia.com>
3 years agotest/crypto: fix result when PMD is not loaded
Ciara Power [Tue, 18 May 2021 14:19:55 +0000 (14:19 +0000)]
test/crypto: fix result when PMD is not loaded

The return value for a test when the required PMD is not loaded should
be TEST_SKIPPED, rather than TEST_FAILED.

Fixes: 8bfdd8a7f0f1 ("test/crypto: refactor to use sub test suites")

Signed-off-by: Ciara Power <ciara.power@intel.com>
Acked-by: Fan Zhang <roy.fan.zhang@intel.com>
3 years agotest: fix division by zero
Min Hu (Connor) [Thu, 13 May 2021 01:10:37 +0000 (09:10 +0800)]
test: fix division by zero

Variable i is used as a denominator which may be zero, and
this may result in segmentation fault.

This patch fixed it.

Fixes: 948bc3d6d095 ("test: add reciprocal based division")
Cc: stable@dpdk.org
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
3 years agoexamples/l3fwd-power: fix empty poll thresholds
Hongbo Zheng [Wed, 21 Apr 2021 03:53:39 +0000 (11:53 +0800)]
examples/l3fwd-power: fix empty poll thresholds

Fix assignment errors of ep_hgh_edpi in function parse_ep_config.

Fixes: a137d012a0dd ("examples/l3fwd-power: support traffic pattern aware control")
Cc: stable@dpdk.org
Signed-off-by: Hongbo Zheng <zhenghongbo3@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
Reviewed-by: Reshma Pattan <reshma.pattan@gmail.com>
Acked-by: David Hunt <david.hunt@intel.com>
3 years agodevtools: check flow API doc tables
Thomas Monjalon [Tue, 6 Apr 2021 22:28:12 +0000 (00:28 +0200)]
devtools: check flow API doc tables

The script check-doc-vs-code.sh may be used to add
some automatic checks of the doc.

If run without any argument, a complete check is done.
The optional argument is a git history reference point
to check faster only what has changed since this commit.

In this commit, the only check is for rte_flow tables,
achieved through the script parse-flow-support.sh.
If run without a .ini reference, it prints rte_flow tables.
Note: detected features are marked with the value Y,
while the real .ini file could have special values like I.
The script allow parsing exceptions (exclude or include),
like for bnxt code which lists unsupported items and actions.

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
---
v6 changes:
- fix redundant drivers
- ignore indirect action
- prefix misses with a category (item or action)

3 years agodoc: add flow API features tables
Thomas Monjalon [Wed, 7 Apr 2021 20:56:51 +0000 (22:56 +0200)]
doc: add flow API features tables

The NICs overview table lists all supported features per driver.
There was a single row for "Flow API",
although rte_flow is composed of many items and actions.

The row "Flow API" is replaced with two new tables for items and actions.

Also, since rte_flow is not implemented in all drivers,
it would be ugly to add empty sections in some files.
That's why the error message for missing INI section is removed.

The lists are sorted alphabetically.
The extra files for some VF and vectorized data paths are not filled.

Signed-off-by: Asaf Penso <asafp@nvidia.com>
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Kiran Kumar K <kirankumark@marvell.com>
---
v6 changes:
- rebase/update
- remove deprecated shared action

3 years agodoc: rename sfc features file
Thomas Monjalon [Wed, 7 Apr 2021 21:03:59 +0000 (23:03 +0200)]
doc: rename sfc features file

The driver directory is drivers/net/sfc
but the features file was doc/guides/nics/features/sfc_efx.ini.

sfc_efx.ini is renamed sfc.ini to match the driver directory name.
It will help automatic checks of this file.

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
3 years agotest/table: fix build with GCC 11
Ferruh Yigit [Mon, 17 May 2021 15:57:39 +0000 (16:57 +0100)]
test/table: fix build with GCC 11

Build error:
../app/test/test_table_tables.c: In function ‘test_table_stub’:
../app/test/test_table_tables.c:31:9:
warning: ‘memset’ offset [0, 31] is out of the bounds [0, 0]
[-Warray-bounds]
         memset((uint8_t *)mbuf + sizeof(struct rte_mbuf) + 32, 0, 32); \
         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
../app/test/test_table_tables.c:151:25:
note: in expansion of macro ‘PREPARE_PACKET’
  151 |                         PREPARE_PACKET(mbufs[i], 0xadadadad);
      |                         ^~~~~~~~~~~~~~

'key' points to mbuf header + 32 bytes, and memset clears next 32 bytes
of 'key', so overall there needs to be 64 bytes after mbuf header.
Adding a mbuf size check before memset.

The original code has an assumption that mbuf data buffer follows mbuf
header, this patch accepts same assumption.

Bugzilla ID: 677
Fixes: 5205954791cb ("app/test: packet framework unit tests")
Cc: stable@dpdk.org
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
3 years agoversion: 21.05-rc3
Thomas Monjalon [Wed, 12 May 2021 20:40:07 +0000 (22:40 +0200)]
version: 21.05-rc3

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
3 years agotest/power: fix turbo test
David Hunt [Wed, 12 May 2021 16:32:54 +0000 (17:32 +0100)]
test/power: fix turbo test

when turbo is enabled or disabled, the frequency is set to a low non-turbo
frequency, so we need to set to the frequency expected by the test before
checking.

Fixes: aeaeaf5f2d62 ("test/power: add cases for turbo feature")
Cc: stable@dpdk.org
Signed-off-by: David Hunt <david.hunt@intel.com>
3 years agotest/power: fix low frequency test when turbo enabled
David Hunt [Wed, 12 May 2021 16:32:53 +0000 (17:32 +0100)]
test/power: fix low frequency test when turbo enabled

With the intel_pstate driver and turbo enabled, indexing is slightly
different to normal, so to get the test to work properly, enable
turbo at the start.

Fixes: ed7c51a6a680 ("app/test: vm power management")
Cc: stable@dpdk.org
Signed-off-by: David Hunt <david.hunt@intel.com>
3 years agotest/power: add turbo mode to frequency check
David Hunt [Wed, 12 May 2021 16:32:52 +0000 (17:32 +0100)]
test/power: add turbo mode to frequency check

With the intel_pstate driver and turbo enabled, the top frequency in
the frequency array is the P1+1, i.e. 2300001, whereas the frequency
shown in scaling_cur_freq could be a lot higher.

This patch adds a flag to the check_cur_freq function so that we can
specify if a frequency is greater than expected (turbo mode), in which
case the check should be successful.

Fixes: aeaeaf5f2d62 ("test/power: add cases for turbo feature")
Cc: stable@dpdk.org
Signed-off-by: David Hunt <david.hunt@intel.com>
3 years agotest/power: fix CPU frequency check
David Hunt [Wed, 12 May 2021 16:32:51 +0000 (17:32 +0100)]
test/power: fix CPU frequency check

Different drivers present the current cpu core frequency in different
sysfs files. Some present it in cpuinfo_cur_freq, some in scaling_cur_freq,
and some actually present it in both.

This patch attempts to open one, if that fails, tries the other.

Fixes: d550a8cc31f3 ("app/test: enhance power manager unit tests")
Cc: stable@dpdk.org
Signed-off-by: David Hunt <david.hunt@intel.com>
3 years agotest: check flow classifier creation
Min Hu (Connor) [Thu, 22 Apr 2021 06:13:54 +0000 (14:13 +0800)]
test: check flow classifier creation

'cls->cls' will be NULL if flow classifier create has failed,
then segmentation fault will occur if the variable is used.

This patch fixed it.

Fixes: 9c9befea4f57 ("test: add flow classify unit tests")
Cc: stable@dpdk.org
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
Acked-by: Bernard Iremonger <bernard.iremonger@intel.com>
3 years agoexamples/skeleton: fix NUMA check of port and core
Min Hu (Connor) [Thu, 29 Apr 2021 00:50:48 +0000 (08:50 +0800)]
examples/skeleton: fix NUMA check of port and core

According to the comments and logging, the author just hope user to use
the core and device which are in the same numa node for optimal
performance. If not, A warning gives out.

This patch fixes the check for a device on the node 0.

Fixes: 7107e471a6c7 ("examples/skeleton: very simple code for packet forwarding")
Cc: stable@dpdk.org
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
3 years agoexamples/l2fwd-cat: fix NUMA check of port and core
Min Hu (Connor) [Thu, 29 Apr 2021 00:50:47 +0000 (08:50 +0800)]
examples/l2fwd-cat: fix NUMA check of port and core

According to the comments and logging, the author just hope user to use
the core and device which are in the same numa node for optimal
performance. If not, A warning gives out.

This patch fixes the check for a device on the node 0.

Fixes: f6baccbc2b3b ("examples/l2fwd-cat: add sample application for PQoS CAT and CDP")
Cc: stable@dpdk.org
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
3 years agoexamples/flow_classify: fix NUMA check of port and core
Min Hu (Connor) [Thu, 29 Apr 2021 00:50:46 +0000 (08:50 +0800)]
examples/flow_classify: fix NUMA check of port and core

According to the comments and logging, the author just hope user to use
the core and device which are in the same numa node for optimal
performance. If not, A warning gives out.

For example in flow_classify:
./build/flow_classify -a 0000:7d:00.1  -l 93
Here:
0000:7d:00.1 is on numa node 0.
core 93  is on numa node 3.

The two are not in same numa node, but no warning gives out in old codes
when device is on node 0.
This patch includes the node 0 in the check.

Fixes: bab16ddaf2c1 ("examples/flow_classify: add sample application")
Cc: stable@dpdk.org
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
Tested-by: Bernard Iremonger <bernard.iremonger@intel.com>
Acked-by: Bernard Iremonger <bernard.iremonger@intel.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
3 years agoexamples/rxtx_callbacks: fix port ID format specifier
Dmitry Kozlyuk [Wed, 5 May 2021 21:54:35 +0000 (00:54 +0300)]
examples/rxtx_callbacks: fix port ID format specifier

This fixes -Wformat warning with clang 10.0.0 on Windows.

Fixes: f8244c6399d9 ("ethdev: increase port id range")
Cc: stable@dpdk.org
Signed-off-by: Dmitry Kozlyuk <dmitry.kozliuk@gmail.com>
Acked-by: Tyler Retzlaff <roretzla@linux.microsoft.com>
Acked-by: Stephen Hemminger <stephen@networkplumber.org>
3 years agoraw/ioat: remove special case for no status reporting
Bruce Richardson [Wed, 12 May 2021 14:50:00 +0000 (14:50 +0000)]
raw/ioat: remove special case for no status reporting

The special fast-path for returning completed descriptors without
reporting status or user-handles returns the number of completed ring
slots used, rather than the number of actual user-submitted jobs. This
means that the counts returned are too high, as the batch descriptor
slots would be included in the total. Therefore remove this special
case, and use the normal status-processing path so that the returned
count is correct in all cases.

Fixes: 245efe544d8e ("raw/ioat: report status of completed jobs")

Reported-by: Jiayu Hu <jiayu.hu@intel.com>
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Kevin Laatz <kevin.laatz@intel.com>
3 years agoraw/ioat: fix ring space checks
Bruce Richardson [Wed, 12 May 2021 14:49:59 +0000 (14:49 +0000)]
raw/ioat: fix ring space checks

When enqueuing a descriptor, when checking that there is at least one
slot free for the current descriptor and a later batch descriptor, we
need to test for both two free and one free, in case the last write
was a batch descriptor which is allowed to use the "spare" slot.

Similarly, when computing the free space in the ring to return to the
user, we need to take account of the same condition, so that we do not
return a "-1" ring space value, by blindly subtracting "2".

Fixes: 245efe544d8e ("raw/ioat: report status of completed jobs")

Reported-by: Jiayu Hu <jiayu.hu@intel.com>
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Kevin Laatz <kevin.laatz@intel.com>
3 years agoapp/crypto-perf: check memory allocation
Min Hu (Connor) [Thu, 6 May 2021 06:13:59 +0000 (14:13 +0800)]
app/crypto-perf: check memory allocation

Return value of a function 'rte_zmalloc' is dereferenced without
checking, and it may call segmentation fault.

This patch fixed it.

Fixes: f8be1786b1b8 ("app/crypto-perf: introduce performance test application")
Cc: stable@dpdk.org
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
3 years agocrypto/qat: fix null authentication request
Adam Dybkowski [Mon, 10 May 2021 10:20:11 +0000 (11:20 +0100)]
crypto/qat: fix null authentication request

This patch fixes the NULL auth generation case where the request
shouldn't contain the authentication result address. Allows to run
ipsec_autotest with a QAT device.

Fixes: 65beb9abca6d ("crypto/qat: fix null auth when using VFIO")
Cc: stable@dpdk.org
Signed-off-by: Adam Dybkowski <adamx.dybkowski@intel.com>
Acked-by: Fan Zhang <roy.fan.zhang@intel.com>
3 years agotest/crypto: build block-cipher suite dynamically
Ciara Power [Wed, 12 May 2021 11:36:55 +0000 (11:36 +0000)]
test/crypto: build block-cipher suite dynamically

In the existing implementation, the blockcipher test cases are being run
and reported as one test case per type, even though multiple test cases
are hidden in each. For example, "test_AES_chain_all" runs 46 test cases.
Each blockcipher type should have a testsuite instead.

The blockcipher testsuite is dynamically built, depending on the
blockcipher type chosen. The testcase struct is modified to allow
running a testcase with data, which is used for data required when
running each blockcipher testcase.

The blockcipher testsuites are added dynamically to parent testsuites
as sub-testsuites where needed.

Signed-off-by: Ciara Power <ciara.power@intel.com>
Acked-by: Declan Doherty <declan.doherty@intel.com>
Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
Tested-by: Ruifeng Wang <ruifeng.wang@arm.com>
3 years agotest/crypto: fix return value of a skipped test
Ciara Power [Wed, 12 May 2021 11:36:54 +0000 (11:36 +0000)]
test/crypto: fix return value of a skipped test

The blockcipher testcase return value TEST_SUCCESS was incorrect for
one conditional check, it should have been TEST_SKIPPED similar to the
other condition checks in this function when the testcase is skipped.

Fixes: 4868f6591c6f ("test/crypto: add cases for raw datapath API")
Cc: stable@dpdk.org
Signed-off-by: Ciara Power <ciara.power@intel.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
Acked-by: Declan Doherty <declan.doherty@intel.com>
Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>
Tested-by: Ruifeng Wang <ruifeng.wang@arm.com>
3 years agotest/crypto: move test suite parameters to header file
Ciara Power [Wed, 12 May 2021 11:36:53 +0000 (11:36 +0000)]
test/crypto: move test suite parameters to header file

The testsuite params struct and ut functions are now in the cryptodev
test header file. This will allow them be used outside of the
cryptodev_test.c file. They will be used in a subsequent patch by the
blockcipher test.

As a result of this change, slight renaming changes were necessary
for ipsec and asym tests, to avoid a clash in names.

Signed-off-by: Ciara Power <ciara.power@intel.com>
Acked-by: Declan Doherty <declan.doherty@intel.com>
Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
Tested-by: Ruifeng Wang <ruifeng.wang@arm.com>
3 years agotest/crypto: replace unsupported with skipped
Ciara Power [Wed, 12 May 2021 11:36:52 +0000 (11:36 +0000)]
test/crypto: replace unsupported with skipped

Testcases were previously using -ENOTSUP and TEST_SKIPPED return
statuses interchangeably. Both resulted in the testcase not being run.

These return statuses are now standardised to TEST_SKIPPED.

Signed-off-by: Ciara Power <ciara.power@intel.com>
Acked-by: Declan Doherty <declan.doherty@intel.com>
Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
Tested-by: Ruifeng Wang <ruifeng.wang@arm.com>
3 years agotest/crypto: refactor to use sub test suites
Ciara Power [Wed, 12 May 2021 11:36:51 +0000 (11:36 +0000)]
test/crypto: refactor to use sub test suites

The existing implementation runs a giant cryptodev testsuite for most
autotests, which in turns runs one setup function regardless of device.

This is now broken down into multiple testsuites,
that are used as sub-testsuites. Each autotest runs a general crypto
parent test suite, to which the sub-testsuites are added.

For example, the AESNI_MB test runs "Cryptodev Unit Test Suite",
which has a setup function only to configure testsuite params.
Creation of vdevs in the setup function is no longer supported,
it is expected the user does this when running the app.
This autotest previously just ran the cryptodev_testsuite,
but now has the smaller sub-testsuites added to the parent suite instead.
The same test cases are being run as before.

The scheduler autotest creates its own parent testsuite with nested
sub-testsuites, rather than using the cryptodev testsuite mentioned above.
This is due to it being more complex in execution,
by requiring setting different modes before running tests.
The scheduler autotest no longer requires the extra test cases to
attach/set mode/detach when running the blockcipher test cases for
each mode. The attach/set mode/detach functionality is now tested in a
sub-testsuite. When running the sub-testsuites for each mode,
the attach/set mode/detach happens in the setup and teardown functions
for that sub-testsuite.

Signed-off-by: Ciara Power <ciara.power@intel.com>
Acked-by: Declan Doherty <declan.doherty@intel.com>
Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
Tested-by: Ruifeng Wang <ruifeng.wang@arm.com>
3 years agotest: introduce parent test suite format
Ciara Power [Wed, 12 May 2021 11:36:50 +0000 (11:36 +0000)]
test: introduce parent test suite format

The current structure for unit testing only allows for running a
test suite with nested test cases. This means all test cases for an
autotest must be in one suite, which is not ideal.
For example, in some cases we may want to run multiple lists of test
cases that each require different setup, so should be in separate suites.

The unit test suite struct is modified to hold a pointer to a list of
sub-testsuite pointers, along with the list of testcases as before.

Signed-off-by: Ciara Power <ciara.power@intel.com>
Acked-by: Declan Doherty <declan.doherty@intel.com>
Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
Reviewed-by: Aaron Conole <aconole@redhat.com>
Tested-by: Ruifeng Wang <ruifeng.wang@arm.com>
3 years agotest: refactor unit test suite runner
Ciara Power [Wed, 12 May 2021 11:36:49 +0000 (11:36 +0000)]
test: refactor unit test suite runner

Some small changes were made to the unit test suite runner for
readability and to enable reuse of some of the function in a later patch.

On test suite setup skip/fail, the loop to count testcases as
skipped/failed has been moved to another function.
This will allow for recursion in a later patch when nested sub-testsuites
are used.

The unit test suite runner accessed the list of testcases in the suite
structure every time the testcase was used. This is now replaced by a
testcase variable which improves readability.

A macro has been introduced for readability, instead of using open
coded loops.

Rather than keep local variable status counts for testcases,
these are added to the test suite structure.

The summary output now prints the suite name, this will be useful later
when multiple nested sub-testsuites are being run.

Signed-off-by: Ciara Power <ciara.power@intel.com>
Acked-by: Aaron Conole <aconole@redhat.com>
Acked-by: Declan Doherty <declan.doherty@intel.com>
Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
Tested-by: Ruifeng Wang <ruifeng.wang@arm.com>
3 years agonet/mlx5: fix metering with flow ID 0
Shun Hao [Wed, 12 May 2021 10:20:59 +0000 (13:20 +0300)]
net/mlx5: fix metering with flow ID 0

When getting meter flow_id bits, there's an issue that not handling
correctly if flow_id is 0.

This fix this issue that when flow_id is 0, treat it as 1 bit.

Fixes: 83306d6c465a ("net/mlx5: fix meter statistics")
Cc: stable@dpdk.org
Signed-off-by: Shun Hao <shunh@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
3 years agonet/mlx5: fix default context in flow age action
Jiawei Wang [Wed, 12 May 2021 12:09:51 +0000 (15:09 +0300)]
net/mlx5: fix default context in flow age action

One of the user parameters for the flow AGE action is the
action context. This context should be provided back to the
user when the action is aged-out.
While this context is NULL, a default value should be provided
by the PMD: the rte_flow pointer in case of rte_flow_create API
and the action pointer in case of the rte_flow_action_handle API.

The default for rte_flow_action_handle was set correctly,
while in case of rte_flow_create it wrongly remained NULL.

This patch set the default value for rte_flow_create case to be
the rte_flow pointer.

Fixes: f9bc5274a6f9 ("net/mlx5: allow age modes combination")
Cc: stable@dpdk.org
Signed-off-by: Jiawei Wang <jiaweiw@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
3 years agonet/mlx5: fix age action in transfer root group
Jiawei Wang [Wed, 12 May 2021 12:09:50 +0000 (15:09 +0300)]
net/mlx5: fix age action in transfer root group

Current the ASO age action was supported in the non-root table,
and the counter based age action was be used in the root table.

The FDB table skips group 0 on MLX5 PMD by adding implicit rule
that jump to non-root table, but PMD code use the original group
value for checking.

This patch adds the transfer checking for ASO age action.

Fixes: f9bc5274a6f9 ("net/mlx5: allow age modes combination")
Cc: stable@dpdk.org
Signed-off-by: Jiawei Wang <jiaweiw@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
3 years agonet/mlx5: fix RSS flow item expansion for NVGRE
Jiawei Wang [Wed, 12 May 2021 10:24:08 +0000 (13:24 +0300)]
net/mlx5: fix RSS flow item expansion for NVGRE

Currently RSS expansion only supports GRE and GRE KEY.
This patch adds RSS expansion for NVGRE item so PMD can expand flow item
correctly.

Fixes: ea81c1b816f7 ("net/mlx5: fix NVGRE matching")
Cc: stable@dpdk.org
Signed-off-by: Jiawei Wang <jiaweiw@nvidia.com>
Acked-by: Xiaoyu Min <jackmin@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
3 years agonet/mlx5: fix mirroring with metering action
Jiawei Wang [Wed, 12 May 2021 06:53:24 +0000 (09:53 +0300)]
net/mlx5: fix mirroring with metering action

While there's mirror action prior to the meter action in the E-Switch
flow, means that the packets should be duplicated into port firstly,
and then do meter and send to the original destination.
MLX5 PMD will split the above E-Switch flow into two sub flows,
similar as mirror with modify action before.

Fixes: 07627fbf1506 ("net/mlx5: support E-Switch mirroring with modify action")
Cc: stable@dpdk.org
Signed-off-by: Jiawei Wang <jiaweiw@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
3 years agonet/mlx5: fix LAG representor probing on PF1 PCI
Xueming Li [Mon, 10 May 2021 13:13:42 +0000 (16:13 +0300)]
net/mlx5: fix LAG representor probing on PF1 PCI

In case of bonding, orchestrator wants to use same devargs for LAG and
non-LAG scenario to probe representor on PF1 using PF1 PCI address
like "<DBDF_PF1>,representor=pf1vf[0-3]".

This patch changes PCI address check policy to allow PF1 PCI address for
representors on PF1.

Note: detaching PF0 device can't remove representors on PF1. It's
recommended to use primary(PF0) PCI address to probe representors on
both PFs.

Fixes: f926cce3fa94 ("net/mlx5: refactor bonding representor probing")
Cc: stable@dpdk.org
Signed-off-by: Xueming Li <xuemingl@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
3 years agonet/mlx5: fix secondary process initialization ordering
Chengwen Feng [Mon, 10 May 2021 12:06:03 +0000 (20:06 +0800)]
net/mlx5: fix secondary process initialization ordering

The memory barrier is used to ensure that the response is returned
only after the Tx/Rx function is set, it should place after the Rx/Tx
function is set.

Fixes: 2aac5b5d119f ("net/mlx5: sync stop/start with secondary process")
Cc: stable@dpdk.org
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
3 years agonet/mlx4: fix secondary process initialization ordering
Chengwen Feng [Mon, 10 May 2021 12:06:02 +0000 (20:06 +0800)]
net/mlx4: fix secondary process initialization ordering

The memory barrier is used to ensure that the response is returned
only after the Tx/Rx function is set, it should place after the Rx/Tx
function is set.

Fixes: 0203d33a1059 ("net/mlx4: support secondary process")
Cc: stable@dpdk.org
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
3 years agonet/tap: fix build with GCC 11
Ferruh Yigit [Tue, 11 May 2021 13:14:35 +0000 (14:14 +0100)]
net/tap: fix build with GCC 11

Reproduced with '--buildtype=debugoptimized' config,
compiler version: gcc (GCC) 12.0.0 20210509 (experimental)

There are multiple build errors, like:
In file included from ../drivers/net/tap/tap_flow.c:13:
In function ‘rte_jhash_2hashes’,
    inlined from ‘rte_jhash’ at ../lib/hash/rte_jhash.h:284:2,
    inlined from ‘tap_flow_set_handle’ at
../drivers/net/tap/tap_flow.c:1306:12,
    inlined from ‘rss_enable’ at ../drivers/net/tap/tap_flow.c:1909:3,
    inlined from ‘priv_flow_process’ at
../drivers/net/tap/tap_flow.c:1228:11:
../lib/hash/rte_jhash.h:238:9:
warning: ‘flow’ may be used uninitialized [-Wmaybe-uninitialized]
  238 |         __rte_jhash_2hashes(key, length, pc, pb, 1);
      |         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
../drivers/net/tap/tap_flow.c: In function ‘priv_flow_process’:
../lib/hash/rte_jhash.h:81:1: note: by argument 1 of type ‘const void *’
to ‘__rte_jhash_2hashes.constprop’ declared here
 81 | __rte_jhash_2hashes(const void *key, uint32_t length, uint32_t *pc,
    | ^~~~~~~~~~~~~~~~~~~
../drivers/net/tap/tap_flow.c:1028:1: note: ‘flow’ declared here
 1028 | priv_flow_process(struct pmd_internals *pmd,
      | ^~~~~~~~~~~~~~~~~

Fix strict aliasing rule by using union.

Bugzilla ID: 690
Fixes: de96fe68ae95 ("net/tap: add basic flow API patterns and actions")
Cc: stable@dpdk.org
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Acked-by: Kevin Traynor <ktraynor@redhat.com>
3 years agonet/ice/base: fix build with GCC 11
Ferruh Yigit [Tue, 11 May 2021 13:14:34 +0000 (14:14 +0100)]
net/ice/base: fix build with GCC 11

Reproduced with '--buildtype=debugoptimized' config,
compiler version: gcc (GCC) 12.0.0 20210509 (experimental)

There are multiple build errors, like:
../drivers/net/ice/base/ice_switch.c: In function ‘ice_add_marker_act’:
../drivers/net/ice/base/ice_switch.c:3727:15:
warning: array subscript ‘struct ice_aqc_sw_rules_elem[0]’
is partly outside array bounds of ‘unsigned char[52]’
[-Warray-bounds]
 3727 |         lg_act->type = CPU_TO_LE16(ICE_AQC_SW_RULES_T_LG_ACT);
      |               ^~
In file included from ../drivers/net/ice/base/ice_type.h:52,
                 from ../drivers/net/ice/base/ice_common.h:8,
                 from ../drivers/net/ice/base/ice_switch.h:8,
                 from ../drivers/net/ice/base/ice_switch.c:5:
../drivers/net/ice/base/ice_osdep.h:209:29:
note: referencing an object of size 52 allocated by ‘rte_zmalloc’
  209 | #define ice_malloc(h, s)    rte_zmalloc(NULL, s, 0)
      |                             ^~~~~~~~~~~~~~~~~~~~~~~
../drivers/net/ice/base/ice_switch.c:3720:50:
note: in expansion of macro ‘ice_malloc’
  lg_act = (struct ice_aqc_sw_rules_elem *)ice_malloc(hw, rules_size);

These errors are mainly because allocated memory is cast to
"struct ice_aqc_sw_rules_elem *" but allocated size is less than the size
of "struct ice_aqc_sw_rules_elem".

"struct ice_aqc_sw_rules_elem" has multiple other structs has unions,
based on which one is used allocated memory being less than the size of
"struct ice_aqc_sw_rules_elem" is logically correct but compiler is
complaining about it.

Since the allocation is done explicitly and both producer and consumer
are internal, safe to ignore the warnings. Also to prevent any side
affect disabling the compiler warning for now, until proper fix done.

Reducing the warning disable to gcc >= 11 version.

Bugzilla ID: 678
Fixes: c7dd15931183 ("net/ice/base: add virtual switch code")
Fixes: 02acdce2f553 ("net/ice/base: add MAC filter with marker and counter")
Fixes: f89aa3affa9e ("net/ice/base: support removing advanced rule")
Cc: stable@dpdk.org
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Acked-by: Haiyue Wang <haiyue.wang@intel.com>
3 years agonet/bnx2x: fix build with GCC 11
Ferruh Yigit [Tue, 11 May 2021 13:14:33 +0000 (14:14 +0100)]
net/bnx2x: fix build with GCC 11

Reproduced with '--buildtype=debugoptimized' config,
compiler version: gcc (GCC) 12.0.0 20210509 (experimental)

Build error:
In file included from ../drivers/net/bnx2x/bnx2x.c:16:
../drivers/net/bnx2x/bnx2x.c: In function ‘bnx2x_hc_ack_sb’:
../drivers/net/bnx2x/bnx2x.h:1528:35:
         warning: ‘igu_ack’ is used uninitialized [-Wuninitialized]
 #define REG_WR32(sc, offset, val) bnx2x_reg_write32(sc, (offset), val)
                                   ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
../drivers/net/bnx2x/bnx2x.h:1531:33:
note: in expansion of macro ‘REG_WR32’
 1531 | #define REG_WR(sc, offset, val) REG_WR32(sc, offset, val)
      |                                 ^~~~~~~~
../drivers/net/bnx2x/bnx2x.h:1916:9: note: in expansion of macro ‘REG_WR’
 1916 |         REG_WR(sc, hc_addr, *val);
      |         ^~~~~~
../drivers/net/bnx2x/bnx2x.h:1905:33: note: ‘igu_ack’ declared here
 1905 |         struct igu_ack_register igu_ack;
      |                                 ^~~~~~~

REG_WR32 requires 'uint32_t', use union instead of cast to 'uint32_t'.

Bugzilla ID: 692
Fixes: 38dff79ba736 ("net/bnx2x: update HSI")
Cc: stable@dpdk.org
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Acked-by: Kevin Traynor <ktraynor@redhat.com>
3 years agonet/bnx2x: fix build with GCC 11
Ferruh Yigit [Tue, 11 May 2021 13:14:32 +0000 (14:14 +0100)]
net/bnx2x: fix build with GCC 11

Reproduced with '--buildtype=debugoptimized' config,
compiler version: gcc (GCC) 12.0.0 20210509 (experimental)

Build error:
In file included from ../drivers/net/bnx2x/bnx2x_rxtx.c:8:
../drivers/net/bnx2x/bnx2x_rxtx.c: In function ‘bnx2x_upd_rx_prod_fast’:
../drivers/net/bnx2x/bnx2x.h:1528:35:
    warning: ‘rx_prods’ is used uninitialized [-Wuninitialized]
 #define REG_WR32(sc, offset, val) bnx2x_reg_write32(sc, (offset), val)
                                   ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
../drivers/net/bnx2x/bnx2x.h:1531:33:
note: in expansion of macro ‘REG_WR32’
 1531 | #define REG_WR(sc, offset, val) REG_WR32(sc, offset, val)
      |                                 ^~~~~~~~
../drivers/net/bnx2x/bnx2x_rxtx.c:331:9:
note: in expansion of macro ‘REG_WR’
  331 |         REG_WR(sc, fp->ustorm_rx_prods_offset, val[0]);
      |         ^~~~~~
../drivers/net/bnx2x/bnx2x_rxtx.c:324:40: note: ‘rx_prods’ declared here
  324 |         struct ustorm_eth_rx_producers rx_prods = { 0 };
      |                                        ^~~~~~~~

REG_WR32 requires 'uint32_t', use union instead of cast to 'uint32_t'.

Bugzilla ID: 692
Fixes: 38dff79ba736 ("net/bnx2x: update HSI")
Cc: stable@dpdk.org
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Acked-by: Kevin Traynor <ktraynor@redhat.com>
3 years agonet/igc: fix speed configuration
Alvin Zhang [Wed, 12 May 2021 08:28:26 +0000 (16:28 +0800)]
net/igc: fix speed configuration

Fixed speed mode is not supported currently, this patch
removes configurations for this mode and adds fault handling
for ETH_LINK_SPEED_FIXED.

Fixes: 4f09bc55ac3d ("net/igc: implement device base operations")
Cc: stable@dpdk.org
Signed-off-by: Alvin Zhang <alvinx.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
3 years agonet/i40e: fix VF RSS configuration
Alvin Zhang [Wed, 12 May 2021 09:23:11 +0000 (17:23 +0800)]
net/i40e: fix VF RSS configuration

The kernel driver supports VF RSS configuration message
"VIRTCHNL_OP_GET_RSS_HENA_CAPS and VIRTCHNL_OP_SET_RSS_HENA",
this patch adds PMD support for these messages.

Fixes: b81295c474b0 ("net/i40e: add user callback for VF to PF message")
Cc: stable@dpdk.org
Signed-off-by: Alvin Zhang <alvinx.zhang@intel.com>
Acked-by: Beilei Xing <beilei.xing@intel.com>
3 years agonet/ice: fix VSI array out of bounds access
Jie Wang [Wed, 12 May 2021 03:14:07 +0000 (03:14 +0000)]
net/ice: fix VSI array out of bounds access

In the loop, when the index of array "vsi->rss_key" is equal
to "vsi->rss_key_size", the array will be accessed out of bounds.

Fixes: 50370662b727 ("net/ice: support device and queue ops")
Cc: stable@dpdk.org
Signed-off-by: Jie Wang <jie1x.wang@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
3 years agonet/iavf: fix Tx context descriptor
Beilei Xing [Wed, 12 May 2021 08:10:14 +0000 (16:10 +0800)]
net/iavf: fix Tx context descriptor

The QW0 of Tx context descriptor should be reset to 0, otherwise the
previous hardware writeback value may pollute the next context descriptor
write.

Fixes: a2b29a7733ef ("net/avf: enable basic Rx Tx")
Cc: stable@dpdk.org
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Signed-off-by: Beilei Xing <beilei.xing@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
3 years agonet/i40e: fix offload flag checking in simple Tx
Alvin Zhang [Tue, 11 May 2021 08:45:31 +0000 (16:45 +0800)]
net/i40e: fix offload flag checking in simple Tx

Tx offload flags 'PKT_TX_IPV6, PKT_TX_IPV4, PKT_TX_OUTER_IPV6,
PKT_TX_OUTER_IPV4' are supported in simple datapath.

This patch removes these offload flags from packet checking in simple
Tx datapath and defines 2 macro I40E_TX_OFFLOAD_SIMPLE_SUP_MASK
and I40E_TX_OFFLOAD_SIMPLE_NOTSUP_MASK.

Fixes: 146ffa81d05e ("net/i40e: add Tx preparation for simple Tx datapath")
Cc: stable@dpdk.org
Signed-off-by: Alvin Zhang <alvinx.zhang@intel.com>
Acked-by: Leyi Rong <leyi.rong@intel.com>
3 years agonet/ena: disable ops not supported by secondary process
Stanislaw Kardach [Wed, 12 May 2021 10:13:46 +0000 (12:13 +0200)]
net/ena: disable ops not supported by secondary process

For dev_ops not supported by the secondary process, either return -EPERM
or return without doing anything. In both cases log a warning.

It's still application's responsibility to avoid calls like that and
those changes are for debugging/informational purposes.

Signed-off-by: Stanislaw Kardach <kda@semihalf.com>
Reviewed-by: Michal Krawczyk <mk@semihalf.com>
Reviewed-by: Igor Chauskin <igorch@amazon.com>
Reviewed-by: Shay Agroskin <shayagr@amazon.com>
3 years agonet/ena: make ethdev references multi-process safe
Stanislaw Kardach [Wed, 12 May 2021 10:13:45 +0000 (12:13 +0200)]
net/ena: make ethdev references multi-process safe

rte_pci_device and rte_eth_dev are process-local structures. Therefore
ena_adapter::pdev and ena_adapter::rte_dev cannot be used universally.

Both ena_timer_wd_callback and ena_interrupt_handler_rte needs access to
the rte_eth_dev, but as they are being setup and executed in the primary
process, it is safe to pass there the same pointer, which is used for
the device configuration.

In all other cases, except the eth_ena_dev_init(), the rte_eth_dev_data
is used instead.

Signed-off-by: Stanislaw Kardach <kda@semihalf.com>
Signed-off-by: Michal Krawczyk <mk@semihalf.com>
Reviewed-by: Igor Chauskin <igorch@amazon.com>
Reviewed-by: Shay Agroskin <shayagr@amazon.com>
3 years agonet/ena: indicate Rx RSS hash presence
Michal Krawczyk [Wed, 12 May 2021 10:13:44 +0000 (12:13 +0200)]
net/ena: indicate Rx RSS hash presence

To make it possible to the app to determine if the hash was calculated
for the packet or not, the PKT_RX_RSS_HASH should be set in the mbuf's
ol_flags.

As the PMD wasn't setting that, the application couldn't check if there
is a hash in a proper way.

The hash is valid only if it's UDP or TCP and the IP packet wasn't
fragmented.

Fixes: e5df9f33db00 ("net/ena: fix passing RSS hash to mbuf")
Cc: stable@dpdk.org
Signed-off-by: Michal Krawczyk <mk@semihalf.com>
Reviewed-by: Igor Chauskin <igorch@amazon.com>
Reviewed-by: Amit Bernstein <amitbern@amazon.com>
3 years agoapp/testpmd: support action meter color
Haifei Luo [Wed, 12 May 2021 10:49:33 +0000 (13:49 +0300)]
app/testpmd: support action meter color

Currently action RTE_FLOW_ACTION_TYPE_METER_COLOR is defined.
Add the CLI for this action:  color type (types)
There are three types: green, yellow and red.

Example for the new policy meter CLIs:
   add port meter policy 0 1 g_actions color type green / end y_actions
     color type yellow / end r_actions color type red / end

In the above command, the action type is
RTE_FLOW_ACTION_TYPE_METER_COLOR, the meter policy action list:
green -> green, yellow -> yellow, red -> red.

Signed-off-by: Haifei Luo <haifeil@nvidia.com>
Acked-by: Ori Kam <orika@nvidia.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
3 years agovhost: fix sign extension in async packed ring
Cheng Jiang [Wed, 12 May 2021 02:30:56 +0000 (02:30 +0000)]
vhost: fix sign extension in async packed ring

Change the variable type in store_dma_desc_info_packed() to fix
suspicious implicit sign extension.

Coverity issue: 370608, 370610, 370612
Fixes: 873e8dad6f49 ("vhost: support packed ring in async datapath")

Signed-off-by: Cheng Jiang <cheng1.jiang@intel.com>
3 years agovhost: fix sign extension in async split ring
Cheng Jiang [Wed, 12 May 2021 02:30:41 +0000 (02:30 +0000)]
vhost: fix sign extension in async split ring

Change the variable type in store_dma_desc_info_split() to fix
suspicious implicit sign extension.

Coverity issue: 370604, 370607, 370609
Fixes: 3d6cb86b0de5 ("vhost: refactor async split ring functions")

Signed-off-by: Cheng Jiang <cheng1.jiang@intel.com>
3 years agonet/ark: fix control thread argument
Ed Czeck [Tue, 11 May 2021 20:58:54 +0000 (16:58 -0400)]
net/ark: fix control thread argument

Thread argument changed to wrong value during thread name addition,
fixing that bug.

Fixes: fdefe038eb9b ("net/ark: set generator delay thread name")

Signed-off-by: Ed Czeck <ed.czeck@atomicrules.com>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
3 years agonet/mlx5: fix tunnel offload private items location
Gregory Etelson [Thu, 6 May 2021 09:57:51 +0000 (12:57 +0300)]
net/mlx5: fix tunnel offload private items location

Tunnel offload API requires application to query PMD for specific flow
items and actions. Application uses these PMD specific elements to
build flow rules according to the tunnel offload model.
The model does not restrict private elements location in a flow rule,
but the current MLX5 PMD implementation expects that tunnel offload
rule will begin with PMD specific elements.
The patch removes that placement limitation.

Fixes: 4ec6360de37d ("net/mlx5: implement tunnel offload")
Cc: stable@dpdk.org
Signed-off-by: Gregory Etelson <getelson@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
3 years agonet/enic: enable GENEVE offload via VNIC configuration
John Daley [Tue, 11 May 2021 19:25:26 +0000 (12:25 -0700)]
net/enic: enable GENEVE offload via VNIC configuration

The admin-configured vNIC settings (i.e. via CIMC or UCSM) now include
Geneve offload. Use that setting to decide whether to enable or
disable Geneve offload and remove the devarg 'geneve-opt'.

Also, the firmware now allows the driver to change the Geneve port
number. So extend udp_tunnel_port_{add,del} to accept Geneve port, in
addition to VXLAN.

Fixes: 93fb21fdbe23 ("net/enic: enable overlay offload for VXLAN and GENEVE")
Cc: stable@dpdk.org
Signed-off-by: John Daley <johndale@cisco.com>
Reviewed-by: Hyong Youb Kim <hyonkim@cisco.com>
3 years agonet/ice: fix leak on thread termination
David Marchand [Tue, 11 May 2021 11:33:58 +0000 (13:33 +0200)]
net/ice: fix leak on thread termination

A terminated pthread should be joined or detached so that its associated
resources are released.

The "ice-reset-<vf_id>" threads are used to service some reset task in
the background, but they are never joined by the thread that created
them.
The easiest solution is to detach new threads.

The Windows EAL did not provide a pthread_detach wrapper but there is no
resource to release for Windows threads, so add an empty wrapper.

Fixes: 3b3757bda3c3 ("net/ice: get VF hardware index in DCF")
Cc: stable@dpdk.org
Signed-off-by: David Marchand <david.marchand@redhat.com>
Acked-by: Haiyue Wang <haiyue.wang@intel.com>
Acked-by: Dmitry Kozlyuk <dmitry.kozliuk@gmail.com>
3 years agonet/ark: fix leak on thread termination
David Marchand [Tue, 11 May 2021 11:33:57 +0000 (13:33 +0200)]
net/ark: fix leak on thread termination

A terminated pthread should be joined or detached so that its associated
resources are released.

The "ark-delay-pg" thread is just used to delay some task but it is never
joined by the thread that created it.
The easiest solution is to detach the new thread.

Fixes: 727b3fe292bc ("net/ark: integrate PMD")
Cc: stable@dpdk.org
Signed-off-by: David Marchand <david.marchand@redhat.com>
Acked-by: Ed Czeck <ed.czeck@atomicrules.com>
3 years agoapp/testpmd: fix tunnel offload flows cleanup
Gregory Etelson [Tue, 11 May 2021 08:03:31 +0000 (11:03 +0300)]
app/testpmd: fix tunnel offload flows cleanup

Tunnel offload model requires application to obtain PMD related flow
items or actions to construct a flow rule. These elements acquire
internal PMD flow resources that must be explicitly released.

The patch destroys tunnel offload PMD resources after flow creation
failure.

Fixes: 1b9f274623b8 ("app/testpmd: add commands for tunnel offload")
Cc: stable@dpdk.org
Signed-off-by: Gregory Etelson <getelson@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
3 years agonet/hns3: fail setting FEC if one bit mode is not supported
Chengwen Feng [Mon, 10 May 2021 13:38:13 +0000 (21:38 +0800)]
net/hns3: fail setting FEC if one bit mode is not supported

If the FEC mode was not supported, it should return error code.

This patch also adds a space when log error info.

Fixes: 9bf2ea8dbc65 ("net/hns3: support FEC")
Cc: stable@dpdk.org
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agonet/hns3: fix ordering in secondary process initialization
Chengwen Feng [Mon, 10 May 2021 13:38:12 +0000 (21:38 +0800)]
net/hns3: fix ordering in secondary process initialization

The memory barrier is used to ensure that the response is returned
only after the Tx/Rx function is set, it should place after the Rx/Tx
function is set.

Fixes: 23d4b61fee5d ("net/hns3: support multiple process")
Cc: stable@dpdk.org
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agonet/hns3: fix secondary process request start/stop Rx/Tx
Chengwen Feng [Mon, 10 May 2021 13:38:11 +0000 (21:38 +0800)]
net/hns3: fix secondary process request start/stop Rx/Tx

This secondary process should not send request to start/stop Rx/Tx,
this patch fixes it.

Fixes: 23d4b61fee5d ("net/hns3: support multiple process")
Cc: stable@dpdk.org
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agonet/hns3: fix mailbox message ID in log
Chengwen Feng [Mon, 10 May 2021 13:38:10 +0000 (21:38 +0800)]
net/hns3: fix mailbox message ID in log

The mailbox message id is uint8_t, but the unsupported mailbox message
id was logged by uint16.

Fixes: 463e748964f5 ("net/hns3: support mailbox")
Cc: stable@dpdk.org
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agonet/hns3: fix TM QCN error event report by MSI-X
Chengwen Feng [Sat, 8 May 2021 07:40:57 +0000 (15:40 +0800)]
net/hns3: fix TM QCN error event report by MSI-X

The TM QCN error event should report by RAS other than MSIX.

Also this patch adds fifo int enable configuration before the TM QCN
error event is enabled.

Fixes: f53a793bb7c2 ("net/hns3: add more hardware error types")
Fixes: 3903c05382c5 ("net/hns3: remove read when enabling TM QCN error event")
Cc: stable@dpdk.org
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agonet/txgbe: fix QinQ strip
Jiawen Wu [Sat, 8 May 2021 06:35:26 +0000 (14:35 +0800)]
net/txgbe: fix QinQ strip

Support to enable and disable QINQ hardware strip, when configure VLAN
offload with QINQ strip mask. If there are packets have QINQ tag to RSS,
users should enable QINQ strip before configuring the RSS.

Fixes: 220b0e49bc47 ("net/txgbe: support VLAN")
Cc: stable@dpdk.org
Signed-off-by: Jiawen Wu <jiawenwu@trustnetic.com>
3 years agonet/hns3: fix querying flow director counter for out param
Chengwen Feng [Fri, 7 May 2021 09:08:18 +0000 (17:08 +0800)]
net/hns3: fix querying flow director counter for out param

The hardware doesn't support counting the number of bytes that through
the fdir rule. Therefore, the corresponding out parameters (e.g.
bytes_set/bytes) is set to zero.

Fixes: fcba820d9b9e ("net/hns3: support flow director")
Cc: stable@dpdk.org
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agonet/hns3: fix VF alive notification after config restore
Hongbo Zheng [Fri, 7 May 2021 09:08:17 +0000 (17:08 +0800)]
net/hns3: fix VF alive notification after config restore

Currently in the VF reset scenario, the VF performs the set
alive operation before restoring the configuration completed,
which may cause the hardware to work in an abnormal state.

This patch fix this problem by set VF alive after restoring
the configuration is completed.

Fixes: a5475d61fa34 ("net/hns3: support VF")
Cc: stable@dpdk.org
Signed-off-by: Hongbo Zheng <zhenghongbo3@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agonet/hns3: clear hash map on flow director clear
Chengwen Feng [Fri, 7 May 2021 09:08:16 +0000 (17:08 +0800)]
net/hns3: clear hash map on flow director clear

The fdir hash map hold the pointers of fdir rule elements, it needs to
be set to NULL when clear all fdir rules.

Fixes: fcba820d9b9e ("net/hns3: support flow director")
Cc: stable@dpdk.org
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agonet/hns3: fix log on flow director clear
Chengwen Feng [Fri, 7 May 2021 09:08:15 +0000 (17:08 +0800)]
net/hns3: fix log on flow director clear

If clear FDIR rules fail, the error code was logged, but the error code
was useless because it was the sum of all fail code.

This patch fixes it by log the success cnt and fail cnt.

Fixes: fcba820d9b9e ("net/hns3: support flow director")
Fixes: 8eed8acc812e ("net/hns3: add error code to some logs")
Cc: stable@dpdk.org
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agonet/hns3: return error on PCI config write failure
Chengwen Feng [Fri, 7 May 2021 09:08:14 +0000 (17:08 +0800)]
net/hns3: return error on PCI config write failure

This patch returns error code when calling rte_pci_write_config() API.

Fixes: 6dd32ded17d8 ("net/hns3: check PCI config space write")
Cc: stable@dpdk.org
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agodoc: add more description in hns3 guide
Min Hu (Connor) [Thu, 6 May 2021 13:30:12 +0000 (21:30 +0800)]
doc: add more description in hns3 guide

This patch adds more description for hns3 PMD document, that is:
'Sample Application Notes', 'Statistics','Performance tuning'.

Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agonet/af_xdp: fix build with Linux < 5.4
Ciara Loftus [Mon, 10 May 2021 08:47:54 +0000 (09:47 +0100)]
net/af_xdp: fix build with Linux < 5.4

Prior to this change, two implementations of rx_syscall_handler
existed although only one was needed (for the zero copy path which
is only available from kernel 5.4 and onwards). Remove the second
definition from compat.h and move the first definition back to where
it is called in the Rx function. Doing this removes a build warning
on kernels before 5.4 which complained about the second function
being defined but not used.

Fixes: 2aa51cdd559e ("net/af_xdp: fix trigger for syscall on Tx")

Signed-off-by: Ciara Loftus <ciara.loftus@intel.com>