dpdk.git
3 years agonet/mlx5: separate Rx function declarations to another file
Michael Baum [Mon, 12 Apr 2021 06:32:20 +0000 (06:32 +0000)]
net/mlx5: separate Rx function declarations to another file

The mlx5_rxtx.c file contains a lot of Tx burst functions, each of those
is performance-optimized for the specific set of requested offloads.
These ones are generated on the basis of the template function and it
takes significant time to compile, just due to a large number of giant
functions generated in the same file and this compilation is not being
done in parallel with using multithreading.

Therefore we can split the mlx5_rxtx.c file into several separate files
to allow different functions to be compiled simultaneously.
In this patch, we separate Rx function declarations to different header
file in preparation for removing them from the source file and as an
optional preparation step for further consolidation of Rx burst
functions.

Signed-off-by: Michael Baum <michaelba@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
3 years agonet/hns3: rename Rx burst function
Chengwen Feng [Thu, 15 Apr 2021 08:35:21 +0000 (16:35 +0800)]
net/hns3: rename Rx burst function

Currently, user could use runtime config "rx_func_hint=simple" to
select the hns3_recv_pkts API, but the API's name get from
rte_eth_rx_burst_mode_get is "Scalar" which has not reflected "simple".

So this patch renames hns3_recv_pkts to hns3_recv_pkts_simple, and
also change it's name which gets from rte_eth_rx_burst_mode_get to
"Scalar Simple" to maintain conceptual consistency.

Fixes: 521ab3e93361 ("net/hns3: add simple Rx path")
Cc: stable@dpdk.org
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
3 years agonet/virtio: fix kernel set owner for multi-queue devices
Thierry Herbelot [Thu, 8 Apr 2021 12:37:29 +0000 (14:37 +0200)]
net/virtio: fix kernel set owner for multi-queue devices

Restore the original code, where VHOST_SET_OWNER is applied to
all vhostfds of the device.

Fixes: 06856cabb89b ("net/virtio: add virtio-user ops to set owner")
Cc: stable@dpdk.org
Signed-off-by: Thierry Herbelot <thierry.herbelot@6wind.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
3 years agonet/hns3: support fixed link speed
Huisong Li [Tue, 13 Apr 2021 13:47:19 +0000 (21:47 +0800)]
net/hns3: support fixed link speed

This patch adds the configuration of fixed speed for the PF device.

Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agonet/hns3: support flow control autoneg for copper port
Huisong Li [Tue, 13 Apr 2021 13:47:18 +0000 (21:47 +0800)]
net/hns3: support flow control autoneg for copper port

If the flow control auto-negotiation is not supported and the flow
control modes on the local and link partner is asymmetric, the flow
control on the NIC does not take effect. The support of the
auto-negotiation capability requires the cooperation of the firmware
and driver.

This patch supports the flow control auto-negotiation only for copper
port. For optical ports, the forced flow control mode is still used.

Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agonet/hns3: support link speed autoneg for PF
Huisong Li [Tue, 13 Apr 2021 13:47:17 +0000 (21:47 +0800)]
net/hns3: support link speed autoneg for PF

This patch supports link speed auto-negotiation for PF. If the
device supports auto-negotiation, the device negotiates with
the link partner at all speeds supported by the device.

Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agonet/hns3: report speed capability for PF
Huisong Li [Tue, 13 Apr 2021 13:47:16 +0000 (21:47 +0800)]
net/hns3: report speed capability for PF

The speed capability of the device can be reported to the upper-layer app
in rte_eth_dev_info_get API. In this API, the speed capability is derived
from the 'supported_speed', which is the speed capability actually
supported by the NIC. The value of the 'supported_speed' is obtained
once in the probe stage and may be updated in the scheduled task to deal
with the change of the transmission interface.

Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agonet/hns3: obtain supported speed for fiber port
Huisong Li [Tue, 13 Apr 2021 13:47:15 +0000 (21:47 +0800)]
net/hns3: obtain supported speed for fiber port

Currently, the speed of fiber port is obtained by using the default
query type of HNS3_OPC_GET_SFP_INFO opcode. In this way, only
the speed of the optical module can be obtained. In fact, the opcode
also supports an active query type, which is a channel for obtaining
information such as the speed, the supported speed, auto-negotiation
capability, and FEC mode. This patch changes the query type of the
opcode from the default query type to the active query type to obtain
the supported speed of fiber port.

Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agonet/hns3: fix firmware compatibility configuration
Huisong Li [Tue, 13 Apr 2021 13:47:14 +0000 (21:47 +0800)]
net/hns3: fix firmware compatibility configuration

The firmware compatibility configuration in PF driver is used to
maintain the compatibility of some features of the driver and
firmware, and requires firmware to enable these features. Currently,
the configuration is in hns3_init_hardware(), which is a little back.
Because firmware may clear some configurations (such as, MAC related)
after receiving the command. And firmware can not be aware of some
default initializations (such as, flow control) before executing the
command to set the copper PHY when the PHY is controlled by firmware.
Therefore, it is recommended that no other hardware resources are
configured before the compatibility configuration. And it should be
moved to hns3_cmd_init(), which is responsible for the firmware
command initialization of driver.

In addition, the driver needs to perform corresponding processing
if the command fails to be sent.
1) If firmware fails to take over the copper PHY, the copper port fails
   to initialize.
2) If fails to enable the report of link events, the device does not
   support the LSC capability.

Fixes: bff6ebfe30d4 ("net/hns3: refactor PF LSC event report")
Fixes: bac6a0644121 ("net/hns3: fix link status change from firmware")
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 flow control mode
Huisong Li [Tue, 13 Apr 2021 13:47:13 +0000 (21:47 +0800)]
net/hns3: fix flow control mode

Currently, hns3 driver doesn't support to flow control auto-negotiation.
The FC mode requested by user is the same as the current FC mode. It is
not necessary to maintain the current FC mode. We only report the current
FC mode based on actual flow control mode in hns3_flow_ctrl_get().

This patch removes this redundant field. In addition, "requested_mode" in
hns3_hw struct indicates the FC mode requested by user, and the name is
unreasonable. It needs to be modified to "requested_fc_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: add 1000M speed bit for copper PHYs
Huisong Li [Tue, 13 Apr 2021 13:47:12 +0000 (21:47 +0800)]
net/hns3: add 1000M speed bit for copper PHYs

The bit(5) of supported, advertising and lp_advertising for copper
PHYs obtained from the firmware indicates 1000M full-duplex. This
speed capability bit is missing in the current codes.

Fixes: 2e4859f3b362 ("net/hns3: support PF device with copper PHYs")

Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agonet/hns3: fix supported speed of copper ports
Huisong Li [Tue, 13 Apr 2021 13:47:11 +0000 (21:47 +0800)]
net/hns3: fix supported speed of copper ports

The "supported capability" obtained from firmware on copper ports
includes the speed capability, auto-negotiation capability, and flow
control capability. Therefore, this patch changes "supported_capa" to
"supported_speed" and parses the speed capability supported by the
driver from the "supported capability".

Fixes: 2e4859f3b362 ("net/hns3: support PF device with copper PHYs")

Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agonet/hns3: remove VLAN/QinQ ptypes from support list
Chengwen Feng [Tue, 13 Apr 2021 11:50:11 +0000 (19:50 +0800)]
net/hns3: remove VLAN/QinQ ptypes from support list

In the previous patch, driver will calculate packet type by ignoring
VLAN information because the packet type may calculate error when
exist VLAN and VLAN strip.

So here remove the following ptypes from support list:
1) RTE_PTYPE_L2_ETHER_VLAN
2) RTE_PTYPE_L2_ETHER_QINQ
3) RTE_PTYPE_INNER_L2_ETHER_VLAN
4) RTE_PTYPE_INNER_L2_ETHER_QINQ

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>
3 years agonet/hns3: list supported ptypes for advanced Rx descriptor
Chengwen Feng [Tue, 13 Apr 2021 11:50:10 +0000 (19:50 +0800)]
net/hns3: list supported ptypes for advanced Rx descriptor

Kunpeng 930 supports RXD advanced layout. If enabled the layout, the
hardware will report packet type by 8-bit PTYPE filed in the Rx
descriptor, and the supported ptypes are different from original
scheme. So this patch adds supported list for RXD advanced layout.

Fixes: fb5e90694022 ("net/hns3: support Rx descriptor advanced layout")

Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agonet/hns3: fix PTP capability report
Chengwen Feng [Tue, 13 Apr 2021 11:50:08 +0000 (19:50 +0800)]
net/hns3: fix PTP capability report

The PTP depends on special packet type reported by hardware which
enabled rxd advanced layout, so if the hardware doesn't support rxd
advanced layout, driver should ignore the PTP capability.

Fixes: 438752358158 ("net/hns3: get device capability from firmware")
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: add reporting tunnel GRE packet type
Chengwen Feng [Tue, 13 Apr 2021 11:50:07 +0000 (19:50 +0800)]
net/hns3: add reporting tunnel GRE packet type

This patch supports reporting TUNNEL GRE packet type when rxd advanced
layout enabled.

Fixes: fb5e90694022 ("net/hns3: support Rx descriptor advanced layout")

Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agonet/hns3: fix missing outer L4 UDP flag for VXLAN
Chengwen Feng [Tue, 13 Apr 2021 11:50:06 +0000 (19:50 +0800)]
net/hns3: fix missing outer L4 UDP flag for VXLAN

This patch adds RTE_PTYPE_L4_UDP flag when parsed tunnel vxlan packet.

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>
3 years agonet/hns3: fix verification of NEON support
Chengwen Feng [Tue, 13 Apr 2021 11:50:05 +0000 (19:50 +0800)]
net/hns3: fix verification of NEON support

This patch adds verification of whether NEON supported.

Fixes: a3d4f4d291d7 ("net/hns3: support NEON Rx")
Fixes: e31f123db06b ("net/hns3: support NEON Tx")
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 use of command status enumeration
Chengchang Tang [Tue, 13 Apr 2021 11:50:04 +0000 (19:50 +0800)]
net/hns3: fix use of command status enumeration

The type of return value of hns3_cmd_send is int, some function declare
the return value as hns3_cmd_status.

This patch fix the incorrect use of the enum hns3_cmd_status.

Fixes: 737f30e1c3ab ("net/hns3: support command interface with firmware")
Fixes: 02a7b55657b2 ("net/hns3: support Rx interrupt")
Cc: stable@dpdk.org
Signed-off-by: Chengchang Tang <tangchengchang@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agonet/hns3: fix timing in mailbox
Chengchang Tang [Tue, 13 Apr 2021 11:50:03 +0000 (19:50 +0800)]
net/hns3: fix timing in mailbox

Currently, when processing MBX messages, the system timestamp is obtained
to determine whether timeout occurs. However, the gettimeofday function
is not monotonically increasing. Therefore, this may lead to incorrect
judgment or difficulty exiting the loop. And actually, in this scenario,
it is not necessary to obtain the timestamp.

This patch deletes the call to the gettimeofday function during MBX
message processing.

Fixes: 463e748964f5 ("net/hns3: support mailbox")
Cc: stable@dpdk.org
Signed-off-by: Chengchang Tang <tangchengchang@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
3 years agonet/hns3: fix VF handling LSC event in secondary process
Chengwen Feng [Tue, 13 Apr 2021 11:50:02 +0000 (19:50 +0800)]
net/hns3: fix VF handling LSC event in secondary process

VF will build two queues (csq: command send queue, crq: command receive
queue) with firmware, the crq may contain the following messages:
1) mailbox response message which was the ack of mailbox sync request.
2) PF's link status change message which may send by PF at anytime;

Currently, any threads in the primary and secondary processes could
send mailbox sync request, so it will need to process the crq messages
in there own thread context.

If the crq hold two messages: a) PF's link status change message, b)
mailbox response message when secondary process deals with the crq
messages, it will lead to report lsc event in secondary process
because it uses the policy of processing all pending messages at once.

We use the following scheme to solve it:
1) threads in secondary process could only process specifics messages
   (eg. mailbox response message) in crq, if the message processed, its
   opcode will rewrite with zero, then the intr thread in primary
   process will not process again.
2) threads other than intr thread in the primary process use the same
   processing logic as the threads in secondary process.
3) intr thread in the primary process could process all messages.

Fixes: 76a3836b98c4 ("net/hns3: fix setting default MAC address in bonding of VF")
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 possible mismatched response of mailbox
Chengwen Feng [Tue, 13 Apr 2021 11:50:01 +0000 (19:50 +0800)]
net/hns3: fix possible mismatched response of mailbox

Currently, the mailbox synchronous communication between VF and PF use
the following fields to maintain communication:
1. Req_msg_data which was combined by message code and subcode, used to
   match request and response.
2. Head which means the number of requests successfully sent by VF.
3. Tail which means the number of responses successfully received by VF.
4. Lost which means the number of requests which are timeout.

There may possible mismatches of the following situation:
1. VF sends message A with code=1 subcode=1.
Then head=1, tail=0, lost=0.
2. PF was blocked about 500ms when processing the message A.
3. VF will detect message A timeout because it can't get the response
within 500ms.
Then head=1, tail=0, lost=1.
4. VF sends message B with code=1 subcode=1 which equal message A.
Then head=2, tail=0, lost=1.
5. PF processes the first message A and send the response message to VF.
6. VF will update tail field to 1, but the lost field will remain
   unchanged because the code/subcode equal message B's, so driver will
   return success because now the head(2) equals tail(1) plus lost(1).
   This will lead to mismatch of request and response.

To fix the above bug, we use the following scheme:
1. The message sent from VF was labelled with match_id which was a
   unique 16-bit non-zero value.
2. The response sent from PF will label with match_id which got from the
   request.
3. The VF uses the match_id to match request and response message.

This scheme depends on the PF driver, if the PF driver don't support
then VF will uses the original scheme.

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: delete mailbox arq ring
Chengwen Feng [Tue, 13 Apr 2021 11:50:00 +0000 (19:50 +0800)]
net/hns3: delete mailbox arq ring

Currently, driver will copy mailbox messages body into arq ring when
process HNS3_MBX_LINK_STAT_CHANGE and HNS3_MBX_LINK_STAT_CHANGE
message, and then call hns3_mbx_handler API which will direct process
pre-copy messages. In the whole process, the arq ring don't have a
substantial effect.

Note: The arq ring is designed for kernel environment which could not
do much job in interrupt context, but for DPDK it's not required.

Also we rename hns3_handle_link_change_event to
hns3pf_handle_link_change_event which add 'pf' suffix to make it
better to distinguish.

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/iavf: add offload path for Rx AVX512 flex descriptor
Wenzhuo Lu [Wed, 14 Apr 2021 07:34:09 +0000 (15:34 +0800)]
net/iavf: add offload path for Rx AVX512 flex descriptor

Add a specific path for RX AVX512 (flexible descriptor).
In this path, support the HW offload features, like,
checksum, VLAN stripping, RSS hash.
This path is chosen automatically according to the
configuration.

'inline' is used, then the duplicate code is generated
by the compiler.

Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
3 years agonet/iavf: add offload path for Rx AVX512
Wenzhuo Lu [Wed, 14 Apr 2021 07:34:08 +0000 (15:34 +0800)]
net/iavf: add offload path for Rx AVX512

Add a specific path for RX AVX512 (traditional).
In this path, support the HW offload features, like,
checksum, VLAN stripping, RSS hash.
This path is chosen automatically according to the
configuration.

'inline' is used, then the duplicate code is generated
by the compiler.

Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
3 years agonet/iavf: add offload path for Tx AVX512
Wenzhuo Lu [Wed, 14 Apr 2021 07:34:07 +0000 (15:34 +0800)]
net/iavf: add offload path for Tx AVX512

Add a specific path for TX AVX512.
In this path, support the HW offload features, like,
checksum insertion, VLAN insertion.
This path is chosen automatically according to the
configuration.

'inline' is used, then the duplicate code is generated
by the compiler.

Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
3 years agonet/iavf: store offload flag of Rx queue
Wenzhuo Lu [Wed, 14 Apr 2021 07:34:06 +0000 (15:34 +0800)]
net/iavf: store offload flag of Rx queue

Add the offload flag for RX queues to know which offload
features are set.

Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
3 years agonet/i40e: fix crash in AVX512
Wenzhuo Lu [Wed, 14 Apr 2021 07:25:26 +0000 (15:25 +0800)]
net/i40e: fix crash in AVX512

Fix segment fault when failing to get the memory from the pool.
If there's no memory in the default cache, fall back to the
previous process.

The previous AVX2 rearm function is changed to add some AVX512
instructions and changed to a callee of the AVX2 and AVX512
rearm functions.

Fixes: e6a6a138919f ("net/i40e: add AVX512 vector path")
Cc: stable@dpdk.org
Reported-by: David Coyle <david.coyle@intel.com>
Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Tested-by: David Coyle <david.coyle@intel.com>
3 years agonet/ice: fix crash in AVX512
Wenzhuo Lu [Wed, 14 Apr 2021 07:25:25 +0000 (15:25 +0800)]
net/ice: fix crash in AVX512

Fix segment fault when failing to get the memory from the pool.
If there's no memory in the default cache, fall back to the
previous process.

The previous AVX2 rearm function is changed to add some AVX512
instructions and changed to a callee of the AVX2 and AVX512
rearm functions.

Fixes: 7f85d5ebcfe1 ("net/ice: add AVX512 vector path")
Cc: stable@dpdk.org
Reported-by: David Coyle <david.coyle@intel.com>
Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Tested-by: David Coyle <david.coyle@intel.com>
3 years agonet/iavf: fix crash in AVX512
Wenzhuo Lu [Wed, 14 Apr 2021 07:25:24 +0000 (15:25 +0800)]
net/iavf: fix crash in AVX512

Fix segment fault when failing to get the memory from the pool.
If there's no memory in the default cache, fall back to the
previous process.

The previous AVX2 rearm function is changed to add some AVX512
instructions and changed to a callee of the AVX2 and AVX512
rearm functions.

Fixes: 31737f2b66fb ("net/iavf: enable AVX512 for legacy Rx")
Cc: stable@dpdk.org
Reported-by: David Coyle <david.coyle@intel.com>
Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Tested-by: David Coyle <david.coyle@intel.com>
3 years agoethdev: update flow item GTP QFI definition
Raslan Darawsheh [Tue, 23 Mar 2021 12:11:34 +0000 (14:11 +0200)]
ethdev: update flow item GTP QFI definition

'qfi' field is 8 bits which represent single bit for
PPP (paging Policy Presence) single bit for RQI
(Reflective QoS Indicator) and 6 bits for QFI
(QoS Flow Identifier)
This is based on RFC 38415-g30
https://www.3gpp.org/ftp/Specs/archive/38_series/38.415/38415-g30.zip

Updated the doxygen comment and the mask for 'qfi'
to properly identify the full 8 bits of the field.

note: changing the default mask would cause different
patterns generated by testpmd.

Fixes: 346553db5bd1 ("ethdev: add GTP extension header to flow API")
Cc: stable@dpdk.org
Signed-off-by: Raslan Darawsheh <rasland@nvidia.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
3 years agoapp/testpmd: fix bitmap of link speeds when force speed
Huisong Li [Wed, 14 Apr 2021 03:02:05 +0000 (11:02 +0800)]
app/testpmd: fix bitmap of link speeds when force speed

Currently, when the user sets force link speed through 'link_speeds',
bit(0) of 'link_speeds' is not set to 1(ETH_LINK_SPEED_FIXED),
which conflicts with the definition.

Fixes: 88fbedcd5e5a ("app/testpmd: move speed and duplex parsing in a function")
Cc: stable@dpdk.org
Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
3 years agoapp/testpmd: add command for single flow dump
Haifei Luo [Wed, 14 Apr 2021 10:20:00 +0000 (13:20 +0300)]
app/testpmd: add command for single flow dump

Add support for single flow dump.
The CLIs to dump one rule: flow dump PORT rule ID
to dump all: flow dump PORT all
Examples:
testpmd> flow dump 0 all
testpmd> flow dump 0 rule 0

Signed-off-by: Haifei Luo <haifeil@nvidia.com>
Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
3 years agoethdev: dump single flow rule
Haifei Luo [Wed, 14 Apr 2021 10:19:59 +0000 (13:19 +0300)]
ethdev: dump single flow rule

Previous implementations support dump all the flows. Add new arg
rte_flow in rte_flow_dev_dump to dump one flow.

Signed-off-by: Haifei Luo <haifeil@nvidia.com>
Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Acked-by: Ori Kam <orika@nvidia.com>
3 years agodrivers: add missing includes
Ferruh Yigit [Fri, 26 Mar 2021 16:46:15 +0000 (16:46 +0000)]
drivers: add missing includes

These headers are used but not included explicitly, including them.

"arpa/inet.h" is included for 'htons' and friends.
"netinet/in.h" is included for 'IPPROTO_IP'.

Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>
Acked-by: Rasesh Mody <rmody@marvell.com>
3 years agonet/mvpp2: check meter packet mode
Li Zhang [Tue, 13 Apr 2021 15:59:57 +0000 (18:59 +0300)]
net/mvpp2: check meter packet mode

Currently meter algorithms only supports bytes per second(BPS).
Check packet_mode set to TRUE are rejected.

Signed-off-by: Li Zhang <lizh@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
Acked-by: Liron Himi <lironh@marvell.com>
3 years agonet/softnic: check meter packet mode
Li Zhang [Tue, 13 Apr 2021 15:59:56 +0000 (18:59 +0300)]
net/softnic: check meter packet mode

Currently meter algorithms only supports bytes per second(BPS).
Check packet_mode set to TRUE are rejected.

Signed-off-by: Li Zhang <lizh@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
3 years agoapp/testpmd: add meter profile packet mode option
Li Zhang [Tue, 13 Apr 2021 15:59:55 +0000 (18:59 +0300)]
app/testpmd: add meter profile packet mode option

add meter profile packet_mode to the ethernet device.
One example:
add port meter profile rfc2697 (port_id) (profile_id)
(cir) (cbs) (ebs) (packet_mode)

Signed-off-by: Li Zhang <lizh@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
3 years agoethdev: add packet mode in meter profile structure
Li Zhang [Tue, 13 Apr 2021 15:59:54 +0000 (18:59 +0300)]
ethdev: add packet mode in meter profile structure

Currently meter algorithms only supports rate is bytes per second (BPS).
Add packet_mode flag in meter profile parameters data structure.
So that it can meter traffic by packet per second.

When packet_mode is 0, the profile rates and bucket sizes are
specified in bytes per second and bytes
when packet_mode is not 0, the profile rates and bucket sizes are
specified in packets and packets per second.

The below structure will be extended:
rte_mtr_meter_profile
rte_mtr_capabilities

Signed-off-by: Li Zhang <lizh@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
3 years agonet/ice: update QinQ switch filter handling
Haiyue Wang [Tue, 13 Apr 2021 05:14:59 +0000 (13:14 +0800)]
net/ice: update QinQ switch filter handling

The hardware outer/inner VLAN protocol types are now updated to map to
new interface VLAN protocol types, so update the application to use new
VLAN protocol types when the rte_flow is QinQ filter type.

Signed-off-by: Haiyue Wang <haiyue.wang@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
3 years agonet/ice/base: allow GTP-U filter using only inner protocols
Qi Zhang [Tue, 13 Apr 2021 05:06:40 +0000 (13:06 +0800)]
net/ice/base: allow GTP-U filter using only inner protocols

Adds a support for switch filter: GTP-U using just inner fields.
If user doesn't specify outer protocol and its fields but wants to
add switch filter for GTP-U using inner protocols and related fields
such as inner L3 and/or inner L4, this patch enables such filtering.

Signed-off-by: Kiran Patil <kiran.patil@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Qiming Yang <qiming.yang@intel.com>
3 years agonet/ice/base: add packet type for PPPoL2TPv2oUDP
Qi Zhang [Tue, 13 Apr 2021 05:06:39 +0000 (13:06 +0800)]
net/ice/base: add packet type for PPPoL2TPv2oUDP

Add some new macros of PTYPE values to support PPPoL2TPv2oUDP.

Signed-off-by: Ting Xu <ting.xu@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Qiming Yang <qiming.yang@intel.com>
3 years agonet/ice/base: fix QinQ PPPoE dummy packet selection
Qi Zhang [Tue, 13 Apr 2021 05:06:38 +0000 (13:06 +0800)]
net/ice/base: fix QinQ PPPoE dummy packet selection

The dummy packet should be QinQ PPPoE ipv6 when ppp protocol is ipv6.

Fixes: bb3386f348dd ("net/ice: enable QinQ filter for switch")
Cc: stable@dpdk.org
Signed-off-by: Yuying Zhang <yuying.zhang@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Qiming Yang <qiming.yang@intel.com>
3 years agonet/ice/base: add inner VLAN protocol type for QinQ filter
Qi Zhang [Tue, 13 Apr 2021 05:06:37 +0000 (13:06 +0800)]
net/ice/base: add inner VLAN protocol type for QinQ filter

Since VLAN protocol type 'ICE_VLAN_OFOS' has been changed to map
the hardware VLAN protocol ID to 'ICE_VLAN_OF_HW (16)' when in Double
VLAN mode, and to 'ICE_VLAN_OL_HW (17)' when in Single VLAN mode.

So 'ICE_VLAN_OFOS' can't be used with 'ICE_VLAN_EX' which is outer VLAN
hardware protocol ID 'ICE_VLAN_OF_HW (16)' to do the QinQ VLAN pattern.

Introduce the new inner VLAN protocol type 'ICE_VLAN_IN', which is inner
VLAN hardware protocol ID 'ICE_VLAN_OL_HW (17)'.

Now for QinQ VLAN pattern, the protocol 'ICE_VLAN_EX' and 'ICE_VLAN_IN'
should be used to set the related protocol header fields like VLAN ID.

Signed-off-by: Haiyue Wang <haiyue.wang@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Qiming Yang <qiming.yang@intel.com>
3 years agonet/ice/base: add set/get GPIO helper functions
Qi Zhang [Tue, 13 Apr 2021 05:06:35 +0000 (13:06 +0800)]
net/ice/base: add set/get GPIO helper functions

Add helper functions to set the GPIO pin state or get the value of a
GPIO signal that's the part of the topology based on AQ commands.
This change is needed to setup GPIO pins state for PTP, SyncE etc.

Signed-off-by: Karol Kolacinski <karol.kolacinski@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Qiming Yang <qiming.yang@intel.com>
3 years agonet/ice/base: enable I2C read/write commands
Qi Zhang [Tue, 13 Apr 2021 05:06:33 +0000 (13:06 +0800)]
net/ice/base: enable I2C read/write commands

Enable I2C read/write AQ commands. They are now required for
controlling the external physical connectors via external I2C
port expander on E810-T adapters.

Signed-off-by: Maciej Machnikowski <maciej.machnikowski@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Qiming Yang <qiming.yang@intel.com>
3 years agonet/ice/base: add priority check of matching recipe
Qi Zhang [Tue, 13 Apr 2021 05:06:36 +0000 (13:06 +0800)]
net/ice/base: add priority check of matching recipe

Check priority when look for a recipe which matches our request
to enable flow priority for switch filter.

Signed-off-by: Yuying Zhang <yuying.zhang@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Qiming Yang <qiming.yang@intel.com>
3 years agonet/ice/base: change protocol ID for VLAN in DVM
Qi Zhang [Tue, 13 Apr 2021 05:06:32 +0000 (13:06 +0800)]
net/ice/base: change protocol ID for VLAN in DVM

Protocol id for first vlan in Double VLAN Mode (DVM) should be
ICE_VLAN_OF_HW = 16, but for Single VLAN Mode (SVM) this should be
ICE_VLAN_OL_HW = 17.

Change protocol id in type to id translation array for outer vlan
to 17 when DVM is enabled, which means the driver, package,
and firmware support DVM.

Signed-off-by: Michal Swiatkowski <michal.swiatkowski@intel.com>
Signed-off-by: Haiyue Wang <haiyue.wang@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Qiming Yang <qiming.yang@intel.com>
3 years agonet/ice/base: support PPPoL2TPv2oUDP RSS hash
Qi Zhang [Tue, 13 Apr 2021 05:06:34 +0000 (13:06 +0800)]
net/ice/base: support PPPoL2TPv2oUDP RSS hash

Add support for PPPoL2TPv2oUDP RSS hash. L2TPv2 and PPP ptypes
and flow headers are added. Protocol id for PPP is added.

Signed-off-by: Ting Xu <ting.xu@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Qiming Yang <qiming.yang@intel.com>
3 years agonet/ice/base: set MAC type for E823C device
Qi Zhang [Tue, 13 Apr 2021 05:06:31 +0000 (13:06 +0800)]
net/ice/base: set MAC type for E823C device

Set E823C device's MAC type as generic.

Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Qiming Yang <qiming.yang@intel.com>
3 years agonet/ice/base: remove unused ptype field in PTT definition
Qi Zhang [Tue, 13 Apr 2021 05:06:30 +0000 (13:06 +0800)]
net/ice/base: remove unused ptype field in PTT definition

Remove the unused ptype entry, and use the gcc extension for
ranged initializers in arrays for Linux, and explicitly target
each table entry by index when initializing under Linux.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Qiming Yang <qiming.yang@intel.com>
3 years agonet/ice/base: print link configure error
Qi Zhang [Tue, 13 Apr 2021 05:06:29 +0000 (13:06 +0800)]
net/ice/base: print link configure error

Newer NVMs return link_cfg_err for get_link_status AQ. Print it
for debug use.

Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Qiming Yang <qiming.yang@intel.com>
3 years agonet/ice/base: support removing VSI from flow profile
Qi Zhang [Tue, 13 Apr 2021 05:06:28 +0000 (13:06 +0800)]
net/ice/base: support removing VSI from flow profile

Adding a function ice_flow_rem_vsi_prof() to remove flow entries
associated to the SW VSI handle. Once complete, clear the vsi index from
the flow profile bitmap. This will ensure that a VSI once removed
can be re-added and the package block rules will be added again.

Signed-off-by: Vignesh Sridhar <vignesh.sridhar@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Qiming Yang <qiming.yang@intel.com>
3 years agonet/ice/base: cleanup code
Qi Zhang [Tue, 13 Apr 2021 05:06:27 +0000 (13:06 +0800)]
net/ice/base: cleanup code

1. There are a lots of function header mismatch its function name.
2. remove unnecessary header file include.
3. remove unnecessary macro.
4. remove unnecessary comment.

Signed-off-by: Dave Ertman <david.m.ertman@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Qiming Yang <qiming.yang@intel.com>
3 years agonet/ice: use write combining store for tail on AVX512
Radu Nicolau [Fri, 9 Apr 2021 10:09:03 +0000 (10:09 +0000)]
net/ice: use write combining store for tail on AVX512

Performance improvement: use a write combining store
instead of a regular mmio write to update queue tail
registers.

Signed-off-by: Radu Nicolau <radu.nicolau@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
3 years agonet/ixgbe: fix Rx errors statistics for UDP checksum
Haiyue Wang [Thu, 8 Apr 2021 06:30:00 +0000 (14:30 +0800)]
net/ixgbe: fix Rx errors statistics for UDP checksum

Restrict the "remove l3_l4_xsum_errors from rx_errors" to 82599 only for
hardware errata.

Fixes: 256ff05a9cae ("ixgbe: fix Rx errors statistics for UDP checksum")
Cc: stable@dpdk.org
Signed-off-by: Haiyue Wang <haiyue.wang@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
3 years agonet/mlx5: fix resource release for mirror flow
Jiawei Wang [Fri, 9 Apr 2021 12:33:28 +0000 (15:33 +0300)]
net/mlx5: fix resource release for mirror flow

The mlx5 PMD allocated the resources of the sample actions, and then
moved these ones to the destination actions array. The original indices
were not cleared and the resources were referenced twice in the
flow object - as the fate actions and in the destination actions array.

This causes the failure on flow destroy because PMD tried to release the
same objects twice.

The patch clears the original indices, add the missed checking for zero
and eliminates multiple object releasing.

Fixes: 00c10c22118a ("net/mlx5: update translate function for mirroring")
Cc: stable@dpdk.org
Signed-off-by: Jiawei Wang <jiaweiw@nvidia.com>
Reviewed-by: Suanming Mou <suanmingm@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
3 years agonet/mlx4: fix RSS action with null hash key
Viacheslav Ovsiienko [Wed, 7 Apr 2021 15:30:15 +0000 (15:30 +0000)]
net/mlx4: fix RSS action with null hash key

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>
3 years agodoc: update push/pop VLAN support in mlx5 guide
Dong Zhou [Thu, 1 Apr 2021 13:22:47 +0000 (16:22 +0300)]
doc: update push/pop VLAN support in mlx5 guide

Updates the documentation for push/pop VLAN support. In E-Switch
mode, push VLAN on ingress traffic and pop VLAN in egress traffic
are both support.

Signed-off-by: Dong Zhou <dongzhou@nvidia.com>
Reviewed-by: Asaf Penso <asafp@nvidia.com>
3 years agonet/mlx5: fix redundant flow after RSS expansion
Xiaoyu Min [Tue, 30 Mar 2021 13:40:32 +0000 (21:40 +0800)]
net/mlx5: fix redundant flow after RSS expansion

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>
3 years agonet/mlx5: support 64-bit for modify field action
Alexander Kozyrev [Wed, 7 Apr 2021 15:18:29 +0000 (15:18 +0000)]
net/mlx5: support 64-bit for modify field action

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>
3 years agonet/mlx5: fix modify field action endianness
Alexander Kozyrev [Fri, 2 Apr 2021 02:07:41 +0000 (02:07 +0000)]
net/mlx5: fix modify field action endianness

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>
3 years agonet/mlx5: fix modify field action order for IPv6
Alexander Kozyrev [Wed, 7 Apr 2021 01:14:33 +0000 (01:14 +0000)]
net/mlx5: fix modify field action order for IPv6

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>
3 years agonet/mlx5: fix link speed calculation on Windows
Tal Shnaiderman [Tue, 6 Apr 2021 07:31:59 +0000 (10:31 +0300)]
net/mlx5: fix link speed calculation on Windows

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>
3 years agonet/mlx5: check extended metadata for meta modification
Alexander Kozyrev [Wed, 7 Apr 2021 01:13:57 +0000 (01:13 +0000)]
net/mlx5: check extended metadata for meta modification

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>
3 years agonet/iavf: support flow director for IP fragment
Jeff Guo [Tue, 13 Apr 2021 08:10:32 +0000 (16:10 +0800)]
net/iavf: support flow director for IP fragment

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>
3 years agonet/iavf: support RSS hash for IP fragment
Jeff Guo [Tue, 13 Apr 2021 08:10:31 +0000 (16:10 +0800)]
net/iavf: support RSS hash for IP fragment

New pattern and RSS hash flow parsing are added to handle fragmented
IPv4/IPv6 packets.

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>
3 years agocommon/iavf: add protocol header for IP fragment
Jeff Guo [Tue, 13 Apr 2021 08:10:30 +0000 (16:10 +0800)]
common/iavf: add protocol header for IP fragment

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>
3 years agonet/ice: fix illegal access when removing MAC filter
Wenwu Ma [Fri, 2 Apr 2021 13:52:47 +0000 (13:52 +0000)]
net/ice: fix illegal access when removing MAC filter

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>
3 years agonet/e1000: fix max Rx packet size
Alvin Zhang [Thu, 8 Apr 2021 07:41:02 +0000 (15:41 +0800)]
net/e1000: fix max Rx packet size

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>
3 years agonet/ice: extend invalid RSS combinations
Lingyu Liu [Thu, 8 Apr 2021 15:54:36 +0000 (15:54 +0000)]
net/ice: extend invalid RSS combinations

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>
3 years agonet/enic: fix completion pointer calculation
John Daley [Mon, 12 Apr 2021 20:50:21 +0000 (13:50 -0700)]
net/enic: fix completion pointer calculation

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>
3 years agocommon/sfc_efx/base: fix indication of MAE encap support
Ivan Malov [Sat, 10 Apr 2021 00:51:43 +0000 (03:51 +0300)]
common/sfc_efx/base: fix indication of MAE encap support

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>
3 years agonet/hns3: fix configure FEC when concurrent with reset
Chengchang Tang [Sat, 10 Apr 2021 01:11:20 +0000 (09:11 +0800)]
net/hns3: fix configure FEC when concurrent with reset

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>
3 years agonet/hns3: fix queue state when concurrent with reset
Chengchang Tang [Sat, 10 Apr 2021 01:11:19 +0000 (09:11 +0800)]
net/hns3: fix queue state when concurrent with reset

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>
3 years agonet/hns3: fix timing in resetting queues
Chengchang Tang [Sat, 10 Apr 2021 01:11:18 +0000 (09:11 +0800)]
net/hns3: fix timing in resetting queues

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>
3 years agonet/hns3: fix some packet types
Chengwen Feng [Sat, 10 Apr 2021 01:11:17 +0000 (09:11 +0800)]
net/hns3: fix some packet types

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>
3 years agonet/hns3: fix concurrent interrupt handling
Hongbo Zheng [Sat, 10 Apr 2021 01:11:16 +0000 (09:11 +0800)]
net/hns3: fix concurrent interrupt handling

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>
3 years agonet/hns3: fix rollback in PF init
Min Hu (Connor) [Sat, 10 Apr 2021 01:11:15 +0000 (09:11 +0800)]
net/hns3: fix rollback in PF init

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>
3 years agonet/hns3: simplify selecting Rx/Tx function
Chengwen Feng [Sat, 10 Apr 2021 01:11:14 +0000 (09:11 +0800)]
net/hns3: simplify selecting Rx/Tx function

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>
3 years agonet/hns3: log selected datapath
Chengwen Feng [Fri, 9 Apr 2021 10:26:43 +0000 (18:26 +0800)]
net/hns3: log selected datapath

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>
3 years agonet/hns3: refactor PF LSC event report
Chengwen Feng [Fri, 9 Apr 2021 04:45:22 +0000 (12:45 +0800)]
net/hns3: refactor PF LSC event report

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>
3 years agonet/hns3: refactor VF LSC event report
Chengwen Feng [Fri, 9 Apr 2021 04:45:21 +0000 (12:45 +0800)]
net/hns3: refactor VF LSC event report

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>
3 years agotest: fix TCP header initialization
Lance Richardson [Tue, 30 Mar 2021 13:23:28 +0000 (09:23 -0400)]
test: fix TCP header initialization

Initialize TCP data offset field with TCP header length, this
field is used to derive L4 header length and by hardware to
validate a TCP header.

Fixes: 41f72ec94074 ("test: add packet burst generator functions")
Cc: stable@dpdk.org
Signed-off-by: Lance Richardson <lance.richardson@broadcom.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
3 years agoapp/testpmd: fix missing MPLS tokens for RSS
Hemant Agrawal [Thu, 8 Apr 2021 09:17:44 +0000 (14:47 +0530)]
app/testpmd: fix missing MPLS tokens for RSS

This patch adds missing MPLS tokens in for RSS config.

Fixes: d810252857c9 ("ethdev: add MPLS RSS offload type")
Cc: stable@dpdk.org
Signed-off-by: Hemant Agrawal <hemant.agrawal@nxp.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
3 years agoethdev: fix VXLAN mask initialization
Gregory Etelson [Thu, 8 Apr 2021 06:48:22 +0000 (09:48 +0300)]
ethdev: fix VXLAN mask initialization

In GCC compiler, __builtin_constant_p(exp) is a function.
The function returns the integer 1 if the argument is known to be
a compile-time constant.
Therefore, __builtin_constant_p(0xffffff << 8) returned 1.
As the result, rte_flow_item_vxlan_mask was initiated to
{{
  {flags = 0x0, rsvd0 = {0x0, 0x0, 0x0},
   vni = {0x0, 0x0, 0x0}, rsvd1 = 0x1},
  hdr = {vx_flags = 0x0, vx_vni = 0x1000000}}}
}}
GCC fails initialization
rte_flow_item_vxlan_mask.hdr.vni = (0xffffff << 8)
with "initializer element is not a constant expression" error.
Use immediate 0xffffff00 value instead.

Fixes: 43af98e687cf ("ethdev: reuse VXLAN header definition in flow item")

Signed-off-by: Gregory Etelson <getelson@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
Reviewed-by: Ivan Malov <ivan.malov@oktetlabs.ru>
3 years agobuildtools: fix all drivers disabled on Windows
Dmitry Kozlyuk [Fri, 16 Apr 2021 20:48:52 +0000 (23:48 +0300)]
buildtools: fix all drivers disabled on Windows

buildtools/list-dir-globs.py printed paths with OS directory separator,
which is "/" on Unices and "\" on Windows, while Meson code always
expected "/". This resulted in all drivers being disabled on Windows.

Replace "\" with "/" in script output. Forward slash is a valid,
although non-default, separator on Windows, so no paths can be broken
by this substitution.

Fixes: ab9407c3addd ("build: allow using wildcards to disable drivers")
Cc: stable@dpdk.org
Signed-off-by: Dmitry Kozlyuk <dmitry.kozliuk@gmail.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
3 years agodoc: add links for build requirements per OS
Asaf Penso [Wed, 24 Feb 2021 08:53:24 +0000 (08:53 +0000)]
doc: add links for build requirements per OS

To compile with meson some dependencies should be installed.
Section "Getting the Tools" describes what needed, but per
OS there are additional steps to do.

Add links to Linux, FreeBSD, and Windows guide for more info.

Signed-off-by: Asaf Penso <asafp@nvidia.com>
3 years agodoc: update minimum required Meson version for Windows
Jie Zhou [Wed, 24 Feb 2021 23:36:52 +0000 (15:36 -0800)]
doc: update minimum required Meson version for Windows

Meson with Windows clang generates incorrect linker flag
"--subsystem,console" instead of "/subsystem:console" which
will fail the DPDK build. This is discovered at porting testpmd.

Meson 0.57.0 has the fix and should be used for DPDK Windows build.
Update the WindowsGSG DPDK Build document for the proper meson version.

Signed-off-by: Jie Zhou <jizh@microsoft.com>
Acked-by: Dmitry Kozlyuk <dmitry.kozliuk@gmail.com>
3 years agobuild: update minimum required Meson version
Gabriel Ganne [Tue, 23 Mar 2021 09:52:19 +0000 (10:52 +0100)]
build: update minimum required Meson version

Bump Meson required version to 0.49.2 which is chosen so as
to be provided by both redhat-8 and debian-10.

Update documentation and travis setup script accordingly.

This fixes the following warning:
WARNING: Project targeting '>= 0.47.1' but tried to use feature introduced
         in '0.48.0': console arg in custom_target

'console' argument is used within kernel/linux/kni/meson.build

Signed-off-by: Gabriel Ganne <gabriel.ganne@6wind.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
3 years agoexamples/l2fwd-crypto: remove key size validation
Shiri Kuzin [Tue, 13 Apr 2021 16:24:52 +0000 (19:24 +0300)]
examples/l2fwd-crypto: remove key size validation

In the example application the key can be provided by the user or
generated randomly by the example application.

Then a validation is done in order to check if the key size is
supported in the algorithm capabilities.

A new feature flag is added in crypto PMDs to allow wrapped keys,
hence, to allow wrapped keys, app should remove the validation of
key size in the application and rely on a PMD key size validation.

The validation is removed in case the key is provided by user and
the RTE_CRYPTODEV_FF_CIPHER_WRAPPED_KEY feature flag is set, and
kept in case the key should be generated by the application or
RTE_CRYPTODEV_FF_CIPHER_WRAPPED_KEY is not set.

Signed-off-by: Shiri Kuzin <shirik@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
3 years agocryptodev: formalize key wrap method in API
Matan Azrad [Tue, 13 Apr 2021 06:37:18 +0000 (09:37 +0300)]
cryptodev: formalize key wrap method in API

The Key Wrap approach is used by applications in order to protect keys
located in untrusted storage or transmitted over untrusted
communications networks. The constructions are typically built from
standard primitives such as block ciphers and cryptographic hash
functions.

The Key Wrap method and its parameters are a secret between the keys
provider and the device, means that the device is preconfigured for
this method using very secured way.

The key wrap method may change the key length and layout.

Add a description for the cipher transformation key to allow wrapped key
to be forwarded by the same API.

Add a new feature flag RTE_CRYPTODEV_FF_CIPHER_WRAPPED_KEY to be enabled
by PMDs support wrapped key in cipher trasformation.

Signed-off-by: Matan Azrad <matan@nvidia.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
3 years agocryptodev: add dequeue count parameter in raw API
Fan Zhang [Wed, 31 Mar 2021 17:20:38 +0000 (18:20 +0100)]
cryptodev: add dequeue count parameter in raw API

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>
3 years agocrypto/octeontx2: support lookaside IPv4 transport mode
Tejasree Kondoj [Thu, 15 Apr 2021 07:22:05 +0000 (12:52 +0530)]
crypto/octeontx2: support lookaside IPv4 transport mode

Adding support for IPv4 lookaside IPsec transport mode.

Signed-off-by: Tejasree Kondoj <ktejasree@marvell.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
3 years agoexamples/ipsec-secgw: support UDP encapsulation
Tejasree Kondoj [Thu, 15 Apr 2021 07:22:04 +0000 (12:52 +0530)]
examples/ipsec-secgw: support UDP encapsulation

Adding lookaside IPsec UDP encapsulation support
for NAT traversal.
Application has to add udp-encap option to sa config file
to enable UDP encapsulation on the SA.

Signed-off-by: Tejasree Kondoj <ktejasree@marvell.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
3 years agocrypto/octeontx2: support UDP encapsulation
Tejasree Kondoj [Thu, 15 Apr 2021 07:22:03 +0000 (12:52 +0530)]
crypto/octeontx2: support UDP encapsulation

Adding UDP encapsulation support for IPsec in
lookaside protocol mode.

Signed-off-by: Tejasree Kondoj <ktejasree@marvell.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
3 years agocryptodev: support multiple cipher data-units
Matan Azrad [Wed, 14 Apr 2021 20:21:58 +0000 (22:21 +0200)]
cryptodev: support multiple cipher data-units

In cryptography, a block cipher is a deterministic algorithm operating
on fixed-length groups of bits, called blocks.

A block cipher consists of two paired algorithms, one for encryption
and the other for decryption. Both algorithms accept two inputs:
an input block of size n bits and a key of size k bits; and both yield
an n-bit output block. The decryption algorithm is defined to be the
inverse function of the encryption.

For AES standard the block size is 16 bytes.
For AES in XTS mode, the data to be encrypted\decrypted does not have to
be multiple of 16B size, the unit of data is called data-unit.
The data-unit size can be any size in range [16B, 2^24B], so, in this
case, a data stream is divided into N amount of equal data-units and
must be encrypted\decrypted in the same data-unit resolution.

For ABI compatibility reason, the size is limited to 64K (16-bit field).
The new field dataunit_len is inserted in a struct padding hole,
which is only 2 bytes long in 32-bit build.
It could be moved and extended later during an ABI-breakage window.

The current cryptodev API doesn't allow the user to select a specific
data-unit length supported by the devices.
In addition, there is no definition how the IV is detected per data-unit
when single operation includes more than one data-unit.

That causes applications to use single operation per data-unit even though
all the data is continuous in memory what reduces datapath performance.

Add a new feature flag to support multiple data-unit sizes, called
RTE_CRYPTODEV_FF_CIPHER_MULTIPLE_DATA_UNITS.
Add a new field in cipher capability, called dataunit_set,
where the devices can report the range of the supported data-unit sizes.
Add a new cipher transformation field, called dataunit_len, where the user
can select the data-unit length for all the operations.

All the new fields do not change the size of their structures,
by filling some struct padding holes.
They are added as exceptions in the ABI check file libabigail.abignore.

Using a bitmap to report the supported data-unit sizes capability allows
the devices to report a range simply as same as the user to read it
simply. also, thus sizes are usually common and probably will be shared
among different devices.

Signed-off-by: Matan Azrad <matan@nvidia.com>
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Akhil Goyal <gakhil@marvell.com>
3 years agocrypto/qat: fix offset for out-of-place scatter-gather
Arek Kusztal [Thu, 18 Mar 2021 13:16:17 +0000 (13:16 +0000)]
crypto/qat: fix offset for out-of-place scatter-gather

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>
3 years agocrypto/qat: support single-pass GMAC on GEN3
Adam Dybkowski [Wed, 14 Apr 2021 11:33:21 +0000 (12:33 +0100)]
crypto/qat: support single-pass GMAC on GEN3

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>
3 years agoapp/crypto-perf: close device after benchmark run
Adam Dybkowski [Fri, 9 Apr 2021 12:13:01 +0000 (13:13 +0100)]
app/crypto-perf: close device after benchmark run

This patch adds closing of the PMD after running the benchmark.

Signed-off-by: Adam Dybkowski <adamx.dybkowski@intel.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>