If RSS action contains non zero hash key length and NULL
key buffer pointer the default hash key should be used.
The check for the NULL pointer this was missing in the mlx4
PMD causing crash, for example, in testpmd with command:
flow validate 0 ingress group 0
pattern eth / ipv4 / end
actions rss queues 0 end key_len 40 / end
Fixes: ac8d22de2394 ("ethdev: flatten RSS configuration in flow API")
Cc: stable@dpdk.org
Signed-off-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
When RSS expand, if there is no expansion happened but completion
happened because user only input next protocol field instead of item
i.e, ether type == 0x8100 instead of VLAN, an extra flow is created with
missing item in order to filter traffic strictly.
However, after [1] and [2] the rte_flow_item_eth itself is enough to
filter out VLAN traffic, the VLAN item is not needed.
[1]: commit 09315fc83861 ("ethdev: add VLAN attributes to ethernet and VLAN items")
[2]: commit 86b59a1af671 ("net/mlx5: support VLAN matching fields")
This redundant flow will cause failure in some scenarios on group 0 due
to they are the same FTE.
Fixes: fc2dd8dd492f ("ethdev: fix expand RSS flows")
Cc: stable@dpdk.org
Signed-off-by: Xiaoyu Min <jackmin@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
Extend the range of immediate value used in the MODIFY_FIELD action
from 32 to 64 bits to conform to the rte_flow_action_modify_data spec.
Apply appropriate big endian conversion to the immediate value
according to a destination field bit width.
Fixes: 641dbe4fb053 ("net/mlx5: support modify field flow action")
Cc: stable@dpdk.org
Signed-off-by: Alexander Kozyrev <akozyrev@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
Converting modify_field action masks to the big endian format is wrong
for small (less than 4 bytes) fields. Use the BE conversions appropriate
for a field size, not rte_cpu_to_be_32 for everything.
Fixes: 144127ba5660 ("net/mlx5: adjust modify field action endianness")
Cc: stable@dpdk.org
Signed-off-by: Alexander Kozyrev <akozyrev@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
Mellanox hardware can only modify any packet field in 32-bit chunks,
which means 4 such chunks are needed to modify an IPv6 address.
The modification order of these chunks starts from the most significant
bits for the IPv6 address. That leads to confusing results when trying
to modify either source or destination address via the MODIFY_FIELD
action. Fix the order of 32-bit chunks for IPv6 addresses modification
by starting from the least significant bits.
Fixes: 641dbe4fb053 ("net/mlx5: support modify field flow action")
Cc: stable@dpdk.org
Signed-off-by: Alexander Kozyrev <akozyrev@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
In Windows DevX returns the rate of the current link speed
in bit/s, this rate was converted to Mibit/s instead of the Mbit/s
rate expected by DPDK resulting in wrong link speed reporting.
Fixes: 6fbd73709ee4 ("net/mlx5: support link update on Windows")
Cc: stable@dpdk.org
Signed-off-by: Tal Shnaiderman <talshn@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
The MODIFY_FIELD action requires the extended metadata support
in order to manipulate on METADATA register as well as on MARK register.
Check if it is supported and reject the MODIFY_FIELD action if it is not
just like it was done before for the MARK register modifications.
Fixes: 0588d64ffde3 ("net/mlx5: check extended metadata for mark modification")
Cc: stable@dpdk.org
Signed-off-by: Alexander Kozyrev <akozyrev@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
New FDIR parsing are added to handle the fragmented IPv4/IPv6 packet.
Signed-off-by: Ting Xu <ting.xu@intel.com>
Signed-off-by: Jeff Guo <jia.guo@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
Add new virtchnl protocol header type and fields for IP fragment packets
to support RSS hash and FDIR.
Signed-off-by: Ting Xu <ting.xu@intel.com>
Signed-off-by: Jeff Guo <jia.guo@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
When removing the mac filter in ice_remove_all_mac_vlan_filters(),
TAILQ_FOREACH_SAFE should be used instead of TAILQ_FOREACH,
Otherwise, it will result in a illegal pointer access.
Fixes: e0dcf94a0d7f ("net/ice: support VLAN ops")
Cc: stable@dpdk.org
Signed-off-by: Wenwu Ma <wenwux.ma@intel.com>
Tested-by: Zhihong Peng <zhihongx.peng@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
According to E1000_ETH_OVERHEAD definition, max_rx_pkt_len contains
one VLAN tag size. Therefore when config RLPML register, if dual VLAN
not enabled there is no need to add VLAN tag size to max_rx_pkt_len,
otherwise only one another VLAN tag size should be added to.
Fixes: e51abef39382 ("igb: fix max RX packet size and support dual VLAN")
Cc: stable@dpdk.org
Signed-off-by: Alvin Zhang <alvinx.zhang@intel.com>
Tested-by: Lingli Chen <linglix.chen@intel.com>
Acked-by: Haiyue Wang <haiyue.wang@intel.com>
When create a rule with following invalid RSS type combinations,
it should fail.
Invalid RSS combinations list:
- ETH_RSS_IPV4 | ETH_RSS_NONFRAG_IPV4_TCP
- ETH_RSS_IPV6 | ETH_RSS_NONFRAG_IPV6_TCP
This patch adds these combinations in 'invalid_rss_comb' array to
do valid check, if the combination check failed, the rule will be
created failed.
Signed-off-by: Lingyu Liu <lingyu.liu@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
The completion queue index could be implicitly extended past its
uint16_t size when multiplied by the size of the descriptor. While
this should not be a problem, coverity flags it. Do the extension
explicitly by casting the index to uintptr_t.
Coverity issue: 161317
Fixes: 8b428cb5a92e ("net/enic: use 64B completion queue entries if available")
Cc: stable@dpdk.org
Signed-off-by: John Daley <johndale@cisco.com>
Reviewed-by: Hyong Youb Kim <hyonkim@cisco.com>
The indication fields in the MCDI response are individual
bits, but the current code mistakenly compares the larger
dword with 1. This breaks encap. type discovery. Fix that.
Fixes: 891408c45a63 ("common/sfc_efx/base: indicate MAE support for encapsulation")
Cc: stable@dpdk.org
Signed-off-by: Ivan Malov <ivan.malov@oktetlabs.ru>
Reviewed-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Reviewed-by: Andy Moreton <amoreton@xilinx.com>
Currently, after the reset is complete, the PMD restores the FEC
according to the FEC configuration reserved in the driver. If there is a
concurrency between the FEC setup operation and the restore operation
after a reset, the FEC status of the last hardware may be unknown.
This patch adds the step of obtaining the lock when setting the FEC to
avoid concurrency between restore operation and setting operation.
Fixes: 9bf2ea8dbc65 ("net/hns3: support FEC")
Cc: stable@dpdk.org
Signed-off-by: Chengchang Tang <tangchengchang@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
At the end of the reset, the state of queues need to be restored
according to the states saved in the driver. If the start and stop
operations of the queues are concurrent at this time, it may cause the
final status to be uncertain.
This patch requires queues to acquire the hw lock before starting and
stopping. If the device is being restored due to reset at this time, it
will block until the reset is completed.
Fixes: fa29fe45a7b4 ("net/hns3: support queue start and stop")
Cc: stable@dpdk.org
Signed-off-by: Chengchang Tang <tangchengchang@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
During the task queue pairs reset, the getimeofday is used to obtain the
timestamp to determine whether the command execution times out. But
gettimeofday is not monotonous, it can be modified by system
administrators, so the timing may not be accurate or even cause the loop
to wait consistently.
And actually, in this scenario, it is not necessary to obtain the
timestamp.
This patch removes the operation of obtaining the timestamp from the task
queue pairs reset function.
Fixes: bba636698316 ("net/hns3: support Rx/Tx and related operations")
Cc: stable@dpdk.org
Signed-off-by: Chengchang Tang <tangchengchang@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
Currently, the packet type calculated by
vlan/ovlan/l3id/l4id/ol3id/ol4id fields have the following problems:
1) Identify error when exist VLAN strip which will lead to the data
buffer has non VLAN header but mbuf's ptype have L2_ETHER_VLAN flag.
2) Some packet identifies error, eg: hardware report it's RARP or
unknown packet, but ptype will marked with L2_ETHER .
So driver will calculate packet type only by l3id/l4id/ol3id/ol4id
fields.
Fixes: 0e98d5e6d9c3 ("net/hns3: fix packet type report in Rx")
Fixes: bba636698316 ("net/hns3: support Rx/Tx and related operations")
Cc: stable@dpdk.org
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
Currently, if RAS interrupt and FLR occurred at the same time, FLR will
be detected and corresponding schedule state will be set during RAS
interrupt processing. However, the schedule state value will be
overridden in subsequent RAS processing, resulting in FLR processing
failure. This patch solves this problem.
Fixes: 2790c6464725 ("net/hns3: support device reset")
Cc: stable@dpdk.org
Signed-off-by: Hongbo Zheng <zhenghongbo3@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
This patch adds rollback processing when updating imissed
stats failed in PF init.
Fixes: 3e9f3042d7c8 ("net/hns3: add imissed packet stats")
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
Currently, there are four control variables (rx_simple_allowed,
rx_vec_allowed, tx_simple_allowed and tx_vec_allowed) which are used
to impact the selection of Rx/Tx burst function.
The purpose of the design is to provide a way to control the selection
of Rx/Tx burst function by modifying it's values, but these variables
have no entry to modify unless make intrusive modifications.
Now we already support runtime config to select Rx/Tx function, these
variables could be removed.
Fixes: a124f9e9591b ("net/hns3: add runtime config to select IO burst function")
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
This patch adds debug info for Rx/Tx burst function which was choosing.
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
Currently, PF driver will report lsc when it detects the link status
change, it's not a generic implementation.
We refactor PF lsc event report by following scheme:
1. PF driver marks RTE_PCI_DRV_INTR_LSC in rte_pci_driver by default.
2. In the init stage, PF driver will detect whether firmware supports
lsc interrupt or not, driver will clear RTE_ETH_DEV_INTR_LSC flag if
firmware doesn't support lsc interrupt.
3. PF driver will report lsc event only when dev_conf.intr_conf.lsc is
set.
Note: If the firmware supports lsc interrupt, we also keep periodic
polling to deal with the interrupt loss.
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
Currently, VF driver periodically obtains link status from PF kernel
driver, and reports lsc event when detects link status change. Because
the period is 1 second, it's probably too late to report especially
in such as bonding scenario.
To solve this problem we use the following scheme:
1. PF kernel driver support immediate push link status to all VFs when
it detects the link status changes.
2. VF driver will detect PF kernel driver whether support push link
status in device init stage by sending request link info mailbox
message to PF, PF then tell VF the push capability by extend
HNS3_MBX_LINK_STAT_CHANGE mailbox message.
3. VF driver marks RTE_PCI_DRV_INTR_LSC in rte_pci_driver by default,
when it detects PF doesn't support push link status then it will clear
RTE_ETH_DEV_INTR_LSC flag.
So if PF kernel driver supports push link status to VF, then VF driver
will have RTE_ETH_DEV_INTR_LSC capability.
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
This patch changes the experimental raw data path dequeue burst API.
Originally the API enforces the user to provide callback function
to get maximum dequeue count. This change gives the user one more
option to pass directly the expected dequeue count.
Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
This commit fixes problem with to small offset when both offsets
(auth, cipher) are non zero in digest encrypt case,
when using out-of-place and sgl.
Fixes: 40002f6c2a24 ("crypto/qat: extend support for digest-encrypted auth-cipher")
Cc: stable@dpdk.org
Signed-off-by: Arek Kusztal <arkadiuszx.kusztal@intel.com>
Acked-by: Declan Doherty <declan.doherty@intel.com>
This patch implements Single-Pass AES-GMAC possible on QAT GEN3
which improves the performance. On GEN1 and GEN2 the previous
chained method is used.
Signed-off-by: Adam Dybkowski <adamx.dybkowski@intel.com>
Acked-by: Declan Doherty <declan.doherty@intel.com>
Added support for DIGEST_ENCRYPTED mode for octeontx
and octeontx2 platforms.
Signed-off-by: Tejasree Kondoj <ktejasree@marvell.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
A temporary session is created for sessionless crypto operations.
rte_cryptodev_sym_session_create() should be used for creating the
temporary session as it initializes the session structure in the
correct way.
Fixes: caeba5062c39 ("crypto/octeontx: improve symmetric session-less path")
Cc: stable@dpdk.org
Signed-off-by: Ankur Dwivedi <adwivedi@marvell.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
Using explicit enum instead of ambiguous integer value
Signed-off-by: Nicolas Chautru <nicolas.chautru@intel.com>
Reviewed-by: Tom Rix <trix@redhat.com>
Add support for enabling or disabling drivers for Arm cross build. Do
not implement any enable/disable lists yet.
Enabling drivers is useful when building for an SoC where we only want
to build a few drivers. That way the list won't be too long.
Similarly, disabling drivers is useful when we want to disable only a
few drivers.
Both of these are advantageous mainly in aarch64 -> aarch64 (or arch ->
same arch) builds, where the build machine may have the required driver
dependencies, yet we don't want to build drivers for a specific SoC.
If enable_drivers is a non-empty list, build only those drivers,
otherwise build all drivers and add them to enable_drivers. If
disable_drivers is non-empty list, build all drivers specified in
enable_drivers except those in disable_drivers.
There are two drivers, bus/pci and bus/vdev, which break the build if
not enabled. Address this by always enabling these if the user disables
them or doesn't specify in their allowlist.
Also remove the old Makefile arm configuration options which don't do
anything in Meson.
Signed-off-by: Juraj Linkeš <juraj.linkes@pantheon.tech>
Reviewed-by: Honnappa Nagarahalli <honnappa.nagarahalli@arm.com>
Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
The PCI device address is being used for sending mailbox which was
introduced in previous commit which replaced the macros so that
multiple DPI blocks in the hardware can be supported.
This patch fixes a NULL pointer access by assigning the PCI device
structure to dpivf.
Fixes: 4495bd887d38 ("raw/octeontx2_dma: support multiple DPI blocks")
Cc: stable@dpdk.org
Signed-off-by: Radha Mohan Chintakuntla <radhac@marvell.com>
Users of <rte_ip.h> relied on it to provide IP-related defines,
like IPPROTO_* constants, but still had to include POSIX headers
for inet_pton() and other standard IP-related facilities.
Extend <rte_ip.h> so that it is a single header to gain access
to IP-related facilities on any OS. Use it to replace POSIX includes
in components enabled on Windows. Move missing constants from Windows
networking shim to OS shim header and include it where needed.
Remove Windows networking shim that is no longer needed.
Signed-off-by: Dmitry Kozlyuk <dmitry.kozliuk@gmail.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Ranjit Menon <ranjit.menon@intel.com>
DPDK code often relies on functions and macros that are not standard C,
but are found on all platforms, even if by slightly different names.
Windows <rte_os.h> provided macros or inline definitions for such symbols.
However, when placed in public header, these symbols were unnecessarily
exposed, breaking consumer POSIX compatibility code.
Move most of the shims to <rte_os_shim.h>, a header to be used instead
of <rte_os.h> by internal code. Include it in libraries and PMDs that
previously imported shims from <rte_os.h>. Directly replace shims that
were only used inside EAL:
* index -> strchr, rindex -> strrchr
* sleep -> rte_delay_us_sleep
* strerror_r -> strerror_s
Signed-off-by: Dmitry Kozlyuk <dmitry.kozliuk@gmail.com>
Acked-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Ranjit Menon <ranjit.menon@intel.com>
In current design, legacy parser rte_devargs_parse() saved scratch
buffer to devargs.args while new parser rte_devargs_layers_parse() saved
to devargs.data. Code using devargs had to know the difference and
cleaned up memory accordingly - error prone.
This patch unifies scratch buffer to data field, introduces
rte_devargs_reset() function to wrap the memory clean up logic.
Signed-off-by: Xueming Li <xuemingl@nvidia.com>
Acked-by: Ray Kinsella <mdr@ashroe.eu>
Reviewed-by: Gaetan Rivet <grive@u256.net>
When device is re-configured, memory allocated for work slot is freed
and new memory is allocated. Due to this we may loose some important
configurations/mappings done with initial work slot memory.
For example, whenever rte_event_eth_tx_adapter_queue_add is called
some important meta i.e. txq handle is stored in work slot structure.
If device gets reconfigured after this tx adaptor add, txq to work
slot mapping will be lost resulting in seg fault during packet
processing, as txq handle could not be retrieved from work slot.
Fixes: 67b5f4686459 ("event/octeontx2: add port config functions")
Cc: stable@dpdk.org
Signed-off-by: Harman Kalra <hkalra@marvell.com>
Use virtual counter for estimating current bucket as PMU cannot be
reliably used to estimate time.
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Reduce amount of memory used by chunk pool when the mempool used
is OCTEONTX2 NPA.
Previously, the number of chunks configured when NPA is used is
equal to the number of timers requested plus the number of buckets
and if the max timeout is long enough w.r.t. resolution requested
there will a large number of buckets which would cause high memory
usage.
Reduce the number of chunks when NPA is used to the number of timers
requested as buckets that are processed chunk lists are automatically
freed.
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Simplify timer bucket estimation we need not align buckets to
power of 2 instead use reciprocal division to compute mod.
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
When XAQ pool is being re-configured, and if the same memzone
is used for fc_mem when freeing the old mempool the fc_mem
will be incorrectly updated with the free count.
Fixes: ffa4ec0b6063 ("event/octeontx2: allow adapters to resize inflight buffers")
Cc: stable@dpdk.org
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Remove event/dlb driver from DPDK code base.
Updated release note's removal section to reflect the same.
Also updated doc/guides/rel_notes/release_20_11.rst to fix the
the missing link issue due to removal of doc/guides/eventdevs/dlb.rst
Signed-off-by: Timothy McDaniel <timothy.mcdaniel@intel.com>
Today it is difficult to know if the SW Eventdev PMD is making
forward progress when it runs an iteration of its service. This
commit adds two xstats to give better visibility to the application.
The new xstats provide an application with which Eventdev ports
received work in the last iteration of scheduling, as well if
forward progress was made by the scheduler.
This patch implements an xstat for the SW PMD that exposes a
bitmask of ports that were scheduled to. In the unlikely case
that the SW PMD instance has 64 or more ports, return UINT64_MAX.
Signed-off-by: Harry van Haaren <harry.van.haaren@intel.com>
Several functions introduced in the addition of the Windows support
to mlx5 were missing the __rte_internal tag.
This miss is better revealed when symbols became exported on Linux too,
and it is caught by the symbol checker with --buildtype=debug.
Fixes: 1552fb287166 ("common/mlx5: add alloc/dealloc PD on Windows")
Fixes: 1969ee424405 ("common/mlx5: add UMEM reg/dereg functions on Windows")
Fixes: ba420719823c ("common/mlx5: add reg/dereg MR on Windows")
Fixes: 56ea803e878e ("build: remove Windows export symbol list")
Cc: stable@dpdk.org
Signed-off-by: Tal Shnaiderman <talshn@nvidia.com>
