Andrew Boyer [Mon, 18 Jan 2021 20:34:58 +0000 (12:34 -0800)]
net/ionic: observe endianness in Rx filter
The IONIC firmware is little-endian.
Add a new field to struct ionic_rx_filter to store the CPU-endian
match type.
Use a local variable for the VLAN when searching the hash table.
Matan Azrad [Wed, 6 Jan 2021 06:43:29 +0000 (06:43 +0000)]
vdpa/mlx5: fix configuration mutex cleanup
When the vDPA device is closed, the driver polling thread is canceled.
The polling thread locks the configuration mutex while it polls the CQs.
When the cancellation happens, it may terminate the thread inside the
critical section what remains the configuration mutex locked.
After device close, the driver may be configured again, in this case,
for example, when the first queue state is updated, the driver tries to
lock the mutex again and deadlock appears.
Initialize the mutex after the polling thread cancellation.
Joyce Kong [Tue, 26 Jan 2021 09:57:43 +0000 (17:57 +0800)]
net/virtio: fix Arm build with old compiler
In file included from ../drivers/net/virtio/virtio_rxtx_packed.c:22:0:
../drivers/net/virtio/virtio_rxtx_packed_neon.h:
In function ‘virtqueue_enqueue_batch_packed_vec’:
../drivers/net/virtio/virtio_rxtx_packed_neon.h:74:2: warning:
implicit declaration of function ‘vreinterpretq_p128_u32’
[-Wimplicit-function-declaration]
poly128_t cmp1 = vreinterpretq_p128_u32(~vceqq_u32(ref_seg, def_ref_seg));
The message shows ‘vreinterpretq_p128_u32’ intrinsic is not supported
because an old version gcc (gcc 4.8.5) is used. So fix the issue with
implementing the logic with other intrinsics.
Bugzilla ID: 621 Fixes: 530887469350 ("net/virtio: add vectorized packed ring NEON Tx") Fixes: 5971ce5e2a59 ("net/virtio: add vectorized packed ring NEON Rx") Reported-by: Subhi Masri <subhim@nvidia.com> Signed-off-by: Joyce Kong <joyce.kong@arm.com> Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com> Tested-by: Subhi Masri <subhim@nvidia.com>
Cheng Jiang [Tue, 12 Jan 2021 04:38:57 +0000 (04:38 +0000)]
examples/vhost: refactor vhost data path
Change the vm2vm data path to batch enqueue for better performance.
Support latest async vhost API, refactor vhost async data path,
replace rte_atomicNN_xxx to __atomic_XXX and clean some codes.
Signed-off-by: Cheng Jiang <cheng1.jiang@intel.com> Reviewed-by: Jiayu Hu <jiayu.hu@intel.com> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Cheng Jiang [Tue, 12 Jan 2021 04:38:56 +0000 (04:38 +0000)]
examples/vhost: add ioat ring space count and check
Add ioat ring space count and check, if ioat ring space is not enough
for the next async vhost packet enqueue, then just return to prevent
enqueue failure. Add rte_ioat_completed_ops() fail handler.
Signed-off-by: Cheng Jiang <cheng1.jiang@intel.com> Reviewed-by: Jiayu Hu <jiayu.hu@intel.com> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
This patch fixes virtio_user_dev_setup() error path,
by cleaning all resources it allocates. It introduces
virtio_user_dev_uninit_notify() that cleans all open
FDs. It implies assigning all FDs to -1 at init time.
With these changes done, virtio_user_dev_init_notify()
can be simplified.
Suggested-by: Adrian Moreno <amorenoz@redhat.com> Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com> Reviewed-by: Chenbo Xia <chenbo.xia@intel.com>
Maxime Coquelin [Tue, 26 Jan 2021 10:16:32 +0000 (11:16 +0100)]
net/virtio: make server mode blocking
This patch makes the Vhost-user backend server mode
blocking at init, waiting for the client connection.
The goal is to make the driver more reliable, as without
waiting for client connection, the Virtio driver has to
assume the Vhost-user backend will support all the
features it has advertized, which could lead to undefined
behaviour.
For example, without this patch, if the user enables packed
ring Virtio feature but the backend does not support it,
the ring initialized by the driver will not be compatible
with the backend.
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com> Reviewed-by: Chenbo Xia <chenbo.xia@intel.com>
Maxime Coquelin [Tue, 26 Jan 2021 10:16:30 +0000 (11:16 +0100)]
net/virtio: improve virtio-user errors handling
This patch adds error logs and propagates errors reported
by the backend. It also adds the device path in error logs
to make it easier to distinguish which device is facing the
issue.
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com> Reviewed-by: Chenbo Xia <chenbo.xia@intel.com>
Maxime Coquelin [Tue, 26 Jan 2021 10:16:21 +0000 (11:16 +0100)]
net/virtio: add virtio-user ops to set owner
This patch implements a dedicated callback for
sending owner request. All the requests will be
converted that way so that backends other than
Vhost-user don't have to work around being it.
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com> Reviewed-by: Chenbo Xia <chenbo.xia@intel.com>
Maxime Coquelin [Tue, 26 Jan 2021 10:16:20 +0000 (11:16 +0100)]
net/virtio: make vhost-user request sender consistent
This patch makes vhost_user_write() consistent with
vhost_user_read(), by passing a Vhost-user message
pointer instead of a buffer pointer and its length, which
is now calculated in the function.
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com> Reviewed-by: Chenbo Xia <chenbo.xia@intel.com>
Maxime Coquelin [Tue, 26 Jan 2021 10:16:11 +0000 (11:16 +0100)]
net/virtio: pack virtio HW struct
This patch improves the virtio_hw struct packing,
going from 88 down to 80 bytes with a 6 bytes hole in
the end of the first cacheline. Fields only used in the
slow path are placed in the end, so that hot path only
uses the first cacheline.
The patch also changes booleans fields to uint8_t type, and
fix inconsistencies in their assignments.
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com> Reviewed-by: Chenbo Xia <chenbo.xia@intel.com>
Maxime Coquelin [Tue, 26 Jan 2021 10:16:05 +0000 (11:16 +0100)]
net/virtio: force IOVA as VA mode for virtio-user
At least Vhost-user backend of Virtio-user PMD requires
IOVA as VA mode. Until now, it was implemented as a hack
by forcing to use mbuf's buf_addr field instead of buf_iova.
This patch removes all this logic and just fails probing
if IOVA as VA mode is not selected. It simplifies the
code overall, and removes some bus-specific logic from
generic virtio_ethdev.c.
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com> Reviewed-by: Chenbo Xia <chenbo.xia@intel.com>
Maxime Coquelin [Tue, 26 Jan 2021 10:16:02 +0000 (11:16 +0100)]
net/virtio: move PCI device init in dedicated file
This patch moves the PCI Ethernet device registration bits
in a dedicated patch. In following patches, more code will
be moved there, with the goal of making virtio_ethdev.c
truly bus-agnostic.
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com> Reviewed-by: Chenbo Xia <chenbo.xia@intel.com>
Maxime Coquelin [Tue, 26 Jan 2021 10:15:58 +0000 (11:15 +0100)]
net/virtio: fix getting old status on reconnect
This patch fixes getting reset status from the restarted
vhost-user backend in case of reconnection, instead of the
status at the time of the disconnection.
This issue was not spotted earlier because Vhost-user
protocol status feature was disabled in server mode.
Fixes: 47235f16505f ("net/virtio-user: set status on socket reconnect") Cc: stable@dpdk.org Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com> Reviewed-by: Chenbo Xia <chenbo.xia@intel.com>
Maxime Coquelin [Tue, 26 Jan 2021 10:15:57 +0000 (11:15 +0100)]
bus/vdev: add driver IOVA VA mode requirement
This patch adds driver flag in vdev bus driver so that
vdev drivers can require VA IOVA mode to be used, which
for example the case of Virtio-user PMD.
The patch implements the .get_iommu_class() callback, that
is called before devices probing to determine the IOVA mode
to be used, and adds a check right before the device is
probed to ensure compatible IOVA mode has been selected.
It also adds a ABI exception rule to accommodate with an
update on the driver registration API
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com> Acked-by: David Marchand <david.marchand@redhat.com> Reviewed-by: Chenbo Xia <chenbo.xia@intel.com>
Jiawei Wang [Tue, 26 Jan 2021 12:53:33 +0000 (14:53 +0200)]
net/mlx5: use global default miss for E-Switch sampling
In E-Switch steering domain there was dedicated default miss action
created for every sampling flow.
The patch replaces this one with the global default miss action.
Signed-off-by: Jiawei Wang <jiaweiw@nvidia.com> Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
This patch adds support of the mbuf fast free offload to the
transmit datapath. This offload allows freeing the mbufs on
transmit completion in the most efficient way. It requires
the all mbufs were allocated from the same pool, have
the reference counter value as 1, and have no any externally
attached buffers.
The mlx5 PMD supports packet data inlining by pushing data
to the transmit descriptor. If packet is short enough and all
data are inline, the mbuf is not needed for data send anymore
and can be freed.
The mbuf free was performed in the most inner loop building
the transmit descriptors. This patch postpones the mbuf free
transaction to the tx_burst routine exit, optimizing the loop
and allowing the bulk freeing for the multiple mbufs in single
pool API call.
ethdev: add IPv6 DSCP option for modify field action
IPv6 DSCP field ID is missing from the original list of Field IDs
for MODIFY_FIELD action. Add it to support IPv6 header fully.
Add ipv6_dscp option for the corresponding header field in testpmd.
Thomas Monjalon [Fri, 22 Jan 2021 17:58:04 +0000 (18:58 +0100)]
ethdev: fix close failure handling
If a failure happens when closing a port,
it was unnecessarily failing again in the function eth_err(),
because of a check against HW removal cause.
Indeed there is a big chance the port is released at this point.
Given the port is in the middle (or at the end) of a close process,
checking the error cause by accessing the port is a non-sense.
The error check is replaced by a simple return in the close function.
Bugzilla ID: 624 Fixes: 8a5a0aad5d3e ("ethdev: allow close function to return an error") Cc: stable@dpdk.org Reported-by: Anatoly Burakov <anatoly.burakov@intel.com> Signed-off-by: Thomas Monjalon <thomas@monjalon.net> Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru> Tested-by: Anatoly Burakov <anatoly.burakov@intel.com>
Dapeng Yu [Tue, 26 Jan 2021 03:03:08 +0000 (11:03 +0800)]
net/ixgbe: disable NFS filtering
Disable NFS header filtering whether NFS packets coalescing are
required or not, in order to make RSS can work on NFS packets.
The code without the patch does follow datasheet, but not consistent
with the ixgbe kernel driver. It causes NFS packets to be filtered
and make them flow into queue 0, before RSS can work on them.
Fixes: b826efba6de4 ("net/ixgbe: align register setting when RSC is disabled") Fixes: 8eecb3295aed ("ixgbe: add LRO support") Cc: stable@dpdk.org Signed-off-by: Dapeng Yu <dapengx.yu@intel.com> Acked-by: Jeff Guo <jia.guo@intel.com>
Haiyue Wang [Mon, 25 Jan 2021 11:44:21 +0000 (19:44 +0800)]
net/iavf: adjust VLAN initialize failure handling
Instead of returning error on VLAN init failure, just log the error, so
that the VF device can continue to be configured.
Fixes: 1c301e8c3cff ("net/iavf: support new VLAN capabilities") Signed-off-by: Haiyue Wang <haiyue.wang@intel.com> Acked-by: Qi Zhang <qi.z.zhang@intel.com>
Haiyue Wang [Mon, 25 Jan 2021 04:30:54 +0000 (12:30 +0800)]
net/iavf: fix VLAN offload requests to PF
If the underlying PF doesn't support a specific ethertype or the ability
to toggle VLAN insertion and/or stripping, then prevent the VF sending
an invalid message to the PF.
Fixes: 1c301e8c3cff ("net/iavf: support new VLAN capabilities") Signed-off-by: Haiyue Wang <haiyue.wang@intel.com> Acked-by: Qi Zhang <qi.z.zhang@intel.com>
Junfeng Guo [Fri, 22 Jan 2021 13:35:41 +0000 (13:35 +0000)]
net/iavf: fix VLAN strip configuration
For AVF with single VLAN mode (SVM), port VLAN stripping config
has already been disabled by PF. In this scenario, the error of
-ENOTSUP can be ignored.
Fixes: 1c301e8c3cff ("net/iavf: support new VLAN capabilities") Signed-off-by: Junfeng Guo <junfeng.guo@intel.com> Acked-by: Haiyue Wang <haiyue.wang@intel.com>
Suanming Mou [Sun, 24 Jan 2021 11:02:06 +0000 (19:02 +0800)]
net/mlx4: fix port attach in secondary process
Currently, the secondary process port UAR register mapping used by Tx
queue is done during port initializing.
Unluckily, in port hot-plug case, the secondary process will be
requested to initialize the port when primary process probe the port.
At that time, the port Tx queue number is still not configured, the
secondary process get Tx queue number as 0. This causes the UAR register
not be mapped as secondary process get Tx queue number 0.
This commit adds the check of Tx queue number in secondary process when
port starts is requested. Once the Tx queue number is not matching, do
UAR mapping with the latest Tx queue number.
Suanming Mou [Sun, 24 Jan 2021 11:02:05 +0000 (19:02 +0800)]
net/mlx5: fix port attach in secondary process
Currently, the secondary process port UAR register mapping used by Tx
queue is done during port initializing.
Unluckily, in port hot-plug case, the secondary process was requested
to initialize the port when primary process did not complete the
device configuration and the port Tx queue number is not configured
yet. Hence, the secondary process gets the zero Tx queue number during
probing, causing the UAR registers not be mapped in the correct
fashion.
This commit checks the configured number of Tx queues in secondary
process when the port start is requested. In case the Tx queue
number mismatch found the UAR mapping is reinitialized accordingly.
Suanming Mou [Sun, 24 Jan 2021 11:02:04 +0000 (19:02 +0800)]
net/mlx5: fix crash on secondary process port close
When secondary process starts, in rte_eth_dev_attach_secondary()
function, the secondary process port device data in struct rte_eth_dev
will be initialized to be shared with primary process port.
When failsafe sub-port hot-plug happens, both primary and secondary
process will release the sub-port, and primary process will clear the
sub-port device data in fs_dev_remove() deactivate stage first before
request secondary process to release the sub-port. In this case, the
secondary process will not be able to get the priv memory pointer from
the shared device data memory anymore, since the device data memory
has been cleared.
Since what secondary process needs in port detach is the UAR table size
to unmap the UAR addresses. It used Tx queue number as size of UAR table
in priv. In fact the uar_table_sz in struct mlx5_proc_priv means the
size of UAR register table - the number of UAR records. However, the
code set this field incorrectly to the size of mlx5_proc_priv structure.
This commit fixes UAR table size to match with relevant Tx queue number,
uses the UAR table size directly to avoid the secondary process to
access the priv pointer in the shared device data memory when unmapping
the UAR address.
Suanming Mou [Sun, 24 Jan 2021 11:02:03 +0000 (19:02 +0800)]
net/mlx5: fix multi-process port ID
The device port_id is used for inter-process communication and must
be the same both for primary and secondary process
This IPC port_id was configured with the invalid temporary value in
port spawn routine. This temporary value was used by the function
rte_eth_dev_get_port_by_name() to check whether the port exists.
This commit corrects the mp port_id with rte_eth_dev port_id.
Packets can get a wrong Flow Tag on x86 architecture with the Flow Tag
compression format (rxq_cqe_comp_en=2) enabled inside the SSE Rx burst.
The shuffle mask that extracts a Flow Tag from the pair of compressed
CQEs is reversed. This leads to the wrong Flow Tag assignment.
Correct the shuffle mask to get proper bytes for a Flow Tag from
miniCQEs.
Fixes: 54c2d46b160f ("net/mlx5: support flow tag and packet header miniCQEs") Cc: stable@dpdk.org Signed-off-by: Alexander Kozyrev <akozyrev@nvidia.com> Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
Gregory Etelson [Wed, 20 Jan 2021 19:17:10 +0000 (21:17 +0200)]
net/mlx5: fix drop action in tunnel offload mode
Tunnel offload mode allows application to restore partially offloaded
tunneled packets to its original state.
The mode was designed to optimize packet recovery. It must not
block flow actions that are allowed by MLX5 PMD.
The patch allows tunnel offload match rules to use drop flow action.
Gregory Etelson [Wed, 20 Jan 2021 19:17:09 +0000 (21:17 +0200)]
net/mlx5: fix mark action in active tunnel offload
Tunnel offload mode allows application to restore partially offloaded
tunneled packets to its original state.
MLX5 PMD stores internal data required to restore partially offloaded
packet in packet mark section. Therefore MLX5 PMD will not allow
applications to use mark action if tunnel offload mode was activated.
The restriction is applied both to regular and tunnel offload rules.
The patch rejects application rules with mark action while tunnel
offload is active.
Alvin Zhang [Thu, 21 Jan 2021 09:41:54 +0000 (17:41 +0800)]
app/testpmd: fix key for RSS flow rule
Since the patch '1848b117' has initialized the variable 'key' in
'struct rte_flow_action_rss' with 'NULL', the PMD cannot get the
RSS key now. Details as bellow:
This patch sets offset and size of the 'key' variable as the first
parameter of the token 'key'. Later, the address of the RSS key will
be copied to 'key' variable.
Fixes: 1848b117cca1 ("app/testpmd: fix RSS key for flow API RSS rule") Cc: stable@dpdk.org Signed-off-by: Alvin Zhang <alvinx.zhang@intel.com> Tested-by: Jun W Zhou <junx.w.zhou@intel.com> Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
Haiyue Wang [Thu, 21 Jan 2021 17:31:37 +0000 (01:31 +0800)]
net/ice: drain out DCF AdminQ command queue
The virtchnl message is handled one by one by checking opcode to match
the response for the request.
The DCF AdminQ command with buffer needs two virtchnl commands, one is
to handle the AdminQ descriptor, the other is to the handle AdminQ
buffer. If both of them are sent to PF successfully, it needs to wait
two responses from PF, even if the AdminQ descriptor command gets the
failure response. Since PF will handle them one by one, and send back
the response for each.
If not wait for the buffer message response until timeout to drain out
the virtchnl command queue, it will cause the next AdminQ command with
buffer to get the stall buffer response from previous.
Fixes: daa714d55c72 ("net/ice: handle AdminQ command by DCF") Cc: stable@dpdk.org Signed-off-by: Haiyue Wang <haiyue.wang@intel.com> Acked-by: Qi Zhang <qi.z.zhang@intel.com>
Somnath Kotur [Thu, 14 Jan 2021 02:19:14 +0000 (18:19 -0800)]
net/bnxt: refactor init/uninit
Move all the individual driver fields allocation routines to one
routine - bnxt_drv_init(). This houses all such routines where
memory needs to be allocated once during the driver's lifetime
and does not need to be torn down during error recovery.
Rename some function names in accordance with their functionality.
bnxt_init_board() is doing nothing more than mapping the PCI bars,
so rename it as such.
Given that there is a bnxt_shutdown_nic that is called in dev_stop_op,
rename it's counterpart - bnxt_init_chip() that is called in
dev_start_op, to bnxt_start_nic. Also helps avoid confusion with some of
the other bnxt_init_xxx routines.
Rename bnxt_init_fw() to bnxt_get_config() as that is what that routine
is doing mostly functionality wise.
Michael Baum [Wed, 20 Jan 2021 08:14:51 +0000 (08:14 +0000)]
net/mlx4: fix handling of probing failure
In mlx4 PCI probing, there are some validations for the Ethernet device
configuration.
From each PCI device the function creates one or two Ethernet devices.
When one of validations fails during the creation of the second device,
the first device is not freed what caused a memory leak.
Free it.
Fixes: 7fae69eeff13 ("mlx4: new poll mode driver") Cc: stable@dpdk.org Signed-off-by: Michael Baum <michaelba@nvidia.com> Acked-by: Matan Azrad <matan@nvidia.com> Tested-by: David Marchand <david.marchand@redhat.com>
Dekel Peled [Sun, 17 Jan 2021 09:40:45 +0000 (11:40 +0200)]
net/mlx5: fix flow split combined with age action
Currently, for a flow containing an age action, if flow is split to
sub-flows, a new age action will be created for each sub-flow.
However only the action created for the last sub-flow will be queried
on flow query and cleared on flow removal.
This behavior is wrong, causing a leak of resources.
Need to create just one action per flow, and use it for all sub-flows.
This patch adds the required check to make sure an age action is
created just once per flow, and used by all sub-flows.
Fixes: f935ed4b645a ("net/mlx5: support flow hit action for aging") Cc: stable@dpdk.org Signed-off-by: Dekel Peled <dekelp@nvidia.com> Acked-by: Matan Azrad <matan@nvidia.com>
Dekel Peled [Sun, 17 Jan 2021 09:40:46 +0000 (11:40 +0200)]
net/mlx5: fix flow split combined with counter
Currently, for a flow containing a count action, if flow is split to
sub-flows, a new counter will be created for each sub-flow.
However only the counter created for the last sub-flow will be queried
on flow query and cleared on flow removal.
This behavior is wrong, causing a leak of resources.
Need to create just one counter per flow, and use it for all sub-flows.
This patch adds the required check to make sure a counter is
created just once per flow, and used by all sub-flows.
Dong Zhou [Tue, 12 Jan 2021 06:40:00 +0000 (08:40 +0200)]
net/mlx5: enlarge maximal flow priority
Currently, the maximal flow priority in non-root table to user
is 4, it's not enough for user to do some flow match by priority,
such as LPM, for one IPV4 address, we need 32 priorities for each
bit of 32 mask length.
PMD will manage 3 sub-priorities per user priority according to L2,
L3 and L4. The internal priority is 16 bits, user can use priorities
from 0 - 21843.
Those enlarged flow priorities are only used for ingress or egress
flow groups greater than 0 and for any transfer flow group.
Jiawei Wang [Tue, 12 Jan 2021 10:29:18 +0000 (12:29 +0200)]
net/mlx5: support E-Switch mirroring with modify action
While there's the modify action and sample action with ratio=1
in the E-Switch flow, and modify action is after the sample
action, means that the modify should only impact on after sample.
MLX5 PMD will monitor the above case and split the E-Switch flow
into two sub flows, similar as sample flow did before:
- the prefix sub flow with all actions preceding the sample and the
sample action itself, also append the new jump action after sample
in the prefix sub flow;
- the suffix sub flow with the modify action and other actions
following the sample action.
The flow split as below:
Original flow: items / actions pre / sample / modify / actions sfx
prefix sub flow -
items / actions pre / set_tag action / sample / jump
suffix sub flow -
tag_item / modify / actions sfx
Signed-off-by: Jiawei Wang <jiaweiw@nvidia.com> Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
Jiawei Wang [Tue, 12 Jan 2021 10:29:17 +0000 (12:29 +0200)]
net/mlx5: extend skip scale flag
The sampling feature introduces the scale flow group with factor,
then the scaled table value can be used for the normal path table
due to this table be created implicitly.
But if the input group value already be scaled, for example the
group value of sampling suffix flow, then use 'skip_scale" flag
to skip the scale twice in the translation action.
Consider the flow with jump action and this jump action could be
created implicitly, PMD may only scale the original flow group
value or scale the jump group value or both, so extend the
'skip_scale' flag to two bits:
If bit0 of 'skip_scale' flag is set to 1, then skip the scale the
original flow group;
If bit1 of 'skip_scale' flag is set to 1, then skip the scale the
jump flow group.
Signed-off-by: Jiawei Wang <jiaweiw@nvidia.com> Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
Jiawei Wang [Tue, 12 Jan 2021 10:29:16 +0000 (12:29 +0200)]
net/mlx5: support E-Switch mirroring and jump in one flow
mlx5 E-Switch mirroring is implemented as multiple destination array in
one steering table. The array currently supports only port ID as
destination actions.
This patch adds the jump action support to the array as one of
destination.
The packets can be mirrored to the port and jump to the next table in
the same destination array allowing to continue handling in the new
table.
For example:
set sample_actions 0 port_id id 1 / end
flow create 0 ingress transfer pattern eth / end actions
sample ratio 1 index 0 / jump group 1 / end
flow create 1 ingress transfer group 1 pattern eth / end actions
set_mac_dst mac_addr 00:aa:bb:cc:dd:ee / port_id id 2 / end
The flow results all the matched ingress packets are mirrored
to port id 1 and go to group 1. In the group 1, packets are modified
with the destination mac and sent to port id 2.
Signed-off-by: Jiawei Wang <jiaweiw@nvidia.com> Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
Jiawei Wang [Tue, 12 Jan 2021 10:29:15 +0000 (12:29 +0200)]
common/mlx5: query preserve capability via DevX
Update function mlx5_devx_cmd_query_hca_attr() to add the
reg_c_preserve bit query.
The stored metadata in register C may be lost in NIC Tx and
FDB egress while doing one of the following operations:
- packet encapsulation.
- packet mirroring (multiple processing paths).
- packet sampling (using Flow Sampler).
If the reg_c_preserve bit is set to 1, then the above
limitation is obsolete, the all metadata registers Cx
preserve their values even through the operations mentioned
above.
Signed-off-by: Jiawei Wang <jiaweiw@nvidia.com> Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
git.droids-corp.org / dpdk.git / log