dpdk.git
2 years agovhost: fix physical address mapping
Xuan Ding [Tue, 15 Feb 2022 15:02:26 +0000 (15:02 +0000)]
vhost: fix physical address mapping

When choosing IOVA as PA mode, IOVA is likely to be discontinuous,
which requires page by page mapping for DMA devices. To be consistent,
this patch implements page by page mapping instead of mapping at the
region granularity for both IOVA as VA and PA mode.

Fixes: 7c61fa08b716 ("vhost: enable IOMMU for async vhost")
Cc: stable@dpdk.org
Signed-off-by: Xuan Ding <xuan.ding@intel.com>
Signed-off-by: Yuan Wang <yuanx.wang@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
2 years agovhost: fix field naming in guest page struct
Xuan Ding [Tue, 15 Feb 2022 15:02:25 +0000 (15:02 +0000)]
vhost: fix field naming in guest page struct

This patch renames the host_phys_addr to host_iova in guest_page
struct. The host_phys_addr is iova, it depends on the DPDK
IOVA mode.

Fixes: e246896178e6 ("vhost: get guest/host physical address mappings")
Cc: stable@dpdk.org
Signed-off-by: Xuan Ding <xuan.ding@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
2 years agocommon/cnxk: enable NIX Tx interrupts errata
Harman Kalra [Mon, 31 Jan 2022 10:52:10 +0000 (16:22 +0530)]
common/cnxk: enable NIX Tx interrupts errata

An errata exists whereby NIX may incorrectly overwrite the value in
NIX_SQ_CTX_S[SQ_INT]. This may cause set interrupts to get cleared or
causing an QINT when no error is outstanding.
As a workaround, software should always read all SQ debug registers
and not just rely on NIX_SQINT_E bits set in NIX_SQ_CTX_S[SQ_INT].
Also for detecting SQB faults software must read SQ context and
check id next_sqb is NULL.

Signed-off-by: Harman Kalra <hkalra@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
2 years agonet/cnxk: fix Rx/Tx function update
Satheesh Paul [Tue, 25 Jan 2022 09:39:40 +0000 (15:09 +0530)]
net/cnxk: fix Rx/Tx function update

When Rx/Tx functions are updated such as in a rte flow
rule creation with VLAN stripping or marking action,
update the fastpath ops table as well.

Fixes: b951c2efcb15 ("net/cnxk: add Rx burst for CN9K")
Fixes: f742a9a66813 ("net/cnxk: add Rx burst for CN10K")
Cc: stable@dpdk.org
Signed-off-by: Satheesh Paul <psatheesh@marvell.com>
Reviewed-by: Jerin Jacob <jerinj@marvell.com>
2 years agocommon/cnxk: fix base rule merge
Satheesh Paul [Tue, 25 Jan 2022 04:17:59 +0000 (09:47 +0530)]
common/cnxk: fix base rule merge

Egress rules do not have a base steering rule and the
current base steering rule is applicable only for ingress
rules. Hence, when creating a flow rule, merge base steering
rule only for ingress rules.

Fixes: f9af90807466 ("common/cnxk: add mcam utility API")
Cc: stable@dpdk.org
Signed-off-by: Satheesh Paul <psatheesh@marvell.com>
Reviewed-by: Kiran Kumar K <kirankumark@marvell.com>
2 years agocommon/cnxk: fix log level during MCAM allocation
Satheesh Paul [Tue, 25 Jan 2022 04:16:43 +0000 (09:46 +0530)]
common/cnxk: fix log level during MCAM allocation

Changed log level from info to debug for a couple of log
messages printed during MCAM allocation.

Fixes: 1f66919817ee ("common/cnxk: improve MCAM entries management")
Cc: stable@dpdk.org
Signed-off-by: Satheesh Paul <psatheesh@marvell.com>
Reviewed-by: Kiran Kumar K <kirankumark@marvell.com>
2 years agocommon/cnxk: fix flow deletion
Satheesh Paul [Tue, 25 Jan 2022 04:15:26 +0000 (09:45 +0530)]
common/cnxk: fix flow deletion

When a flow is deleted, the counters are deleted without
being cleared first. This results in counter values being
retained and shown in flows created later. Hence, counter
is cleared before being deleted.

Updated driver to clear and delete counters upon flow deletion.

Fixes: a07f7ced436d ("common/cnxk: add NPC init and fini")
Cc: stable@dpdk.org
Signed-off-by: Satheesh Paul <psatheesh@marvell.com>
Reviewed-by: Jerin Jacob <jerinj@marvell.com>
2 years agocommon/cnxk: fix null access in IPv6 flow parsing
Satheesh Paul [Tue, 25 Jan 2022 04:12:48 +0000 (09:42 +0530)]
common/cnxk: fix null access in IPv6 flow parsing

Added null check in IPv6 flow pattern parsing.

Fixes: 474e275b1bc6 ("common/cnxk: support extensions attributes in IPv6 item")

Signed-off-by: Satheesh Paul <psatheesh@marvell.com>
Reviewed-by: Jerin Jacob <jerinj@marvell.com>
2 years agoapp/testpmd: check starting port is not in bonding
Min Hu (Connor) [Thu, 17 Feb 2022 11:36:55 +0000 (19:36 +0800)]
app/testpmd: check starting port is not in bonding

In bond, start or stop slave port should be operated by bonding port.
This patch add port_is_bonding_slave in start_port function.

Fixes: 0e545d3047fe ("app/testpmd: check stopping port is not in bonding")
Cc: stable@dpdk.org
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
2 years agonet/bonding: fix slaves initializing on MTU setting
Junjie Wan [Tue, 15 Feb 2022 10:59:40 +0000 (18:59 +0800)]
net/bonding: fix slaves initializing on MTU setting

If a initial process for the bonding device is like:
rte_eth_dev_configure
rte_eth_dev_set_mtu
queue setup and start, etc.

Pass the vdev args to application, and init bonding device only.
-a 0000:af:00.0 --vdev="net_bonding0,mode=2,slave=0000:af:00.0"

It will fail and complain for the slave device
"Port 0 must be configured before MTU set"

Test can be reproduced with ovs.

Fixes: b26bee10ee37 ("ethdev: forbid MTU set before device configure")
Cc: stable@dpdk.org
Signed-off-by: Junjie Wan <wanjunjie@bytedance.com>
Tested-by: Ferruh Yigit <ferruh.yigit@intel.com>
Acked-by: Min Hu (Connor) <humin29@huawei.com>
2 years agonet/bonding: log on invalid device argument
Junjie Wan [Tue, 15 Feb 2022 10:56:12 +0000 (18:56 +0800)]
net/bonding: log on invalid device argument

When invalid args exist, application exits with no error hint.
Adding a log message here will help users to know the reason.

Signed-off-by: Junjie Wan <wanjunjie@bytedance.com>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
Acked-by: Min Hu (Connor) <humin29@huawei.com>
2 years agonet/cnxk: fix mbuf data length
Rahul Bhansali [Wed, 19 Jan 2022 10:52:03 +0000 (16:22 +0530)]
net/cnxk: fix mbuf data length

If multi-segment is enabled and single segment/packet
is received, then mbuf data_len is not updated in
cn9k_nix_cqe_to_mbuf() function.
Also, in case of timestamp is enabled, mbuf data_len
and pkt_len will be updated for all packets including
multi segmented packets.

Fixes: 7c6bee34064f ("net/cnxk: enable PTP processing in vector Rx")
Cc: stable@dpdk.org
Signed-off-by: Rahul Bhansali <rbhansali@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
2 years agoethdev: fix MAC address in telemetry device info
David Marchand [Wed, 16 Feb 2022 14:13:16 +0000 (15:13 +0100)]
ethdev: fix MAC address in telemetry device info

The right size for a human readable MAC is RTE_ETHER_ADDR_FMT_SIZE.
While at it, the net library provides a helper for MAC address
formatting. Prefer it.

Fixes: 58b43c1ddfd1 ("ethdev: add telemetry endpoint for device info")
Cc: stable@dpdk.org
Reported-by: Christophe Fontaine <cfontain@redhat.com>
Signed-off-by: David Marchand <david.marchand@redhat.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
2 years agonet/iavf: reset security context pointer on stop
Yunjian Wang [Tue, 15 Feb 2022 13:02:27 +0000 (21:02 +0800)]
net/iavf: reset security context pointer on stop

When iavf_security_ctx_destroy() is cleaning up, it does not correctly
set the security_ctx variable to NULL, which will lead to wild pointer.

Fixes: 6bc987ecb860 ("net/iavf: support IPsec inline crypto")
Cc: stable@dpdk.org
Signed-off-by: Yunjian Wang <wangyunjian@huawei.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
2 years agonet/txgbe: reset security context pointer on close
Yunjian Wang [Tue, 15 Feb 2022 13:02:01 +0000 (21:02 +0800)]
net/txgbe: reset security context pointer on close

When txgbe_dev_close() is cleaning up, it does not correctly set
the security_ctx variable to NULL, which will lead to wild pointer.

Fixes: f437d97c3df1 ("net/txgbe: add IPsec context creation")
Cc: stable@dpdk.org
Signed-off-by: Yunjian Wang <wangyunjian@huawei.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
2 years agonet/ixgbe: reset security context pointer on close
Yunjian Wang [Tue, 15 Feb 2022 13:01:51 +0000 (21:01 +0800)]
net/ixgbe: reset security context pointer on close

When ixgbe_dev_close() is cleaning up, it does not correctly set
the security_ctx variable to NULL, which will lead to wild pointer.

Fixes: 9a0752f498d2 ("net/ixgbe: enable inline IPsec")
Cc: stable@dpdk.org
Signed-off-by: Yunjian Wang <wangyunjian@huawei.com>
Acked-by: Haiyue Wang <haiyue.wang@intel.com>
2 years agonet/ark: add performance optimizations
John Miller [Tue, 15 Feb 2022 22:19:26 +0000 (16:19 -0600)]
net/ark: add performance optimizations

Added software register writes for hw optimization and
performance fixes.

Signed-off-by: John Miller <john.miller@atomicrules.com>
2 years agonet/ark: add memory write barriers in critical paths
John Miller [Tue, 15 Feb 2022 22:19:25 +0000 (16:19 -0600)]
net/ark: add memory write barriers in critical paths

Add memory write barriers for read and wait status functions
in ddm, udm and mpu.

Signed-off-by: John Miller <john.miller@atomicrules.com>
2 years agonet/ark: support chunk DMA transfers
John Miller [Tue, 15 Feb 2022 22:19:24 +0000 (16:19 -0600)]
net/ark: support chunk DMA transfers

Add support for chunk DMA transfers.

Chunk mpu transfer use 64 objects (512 byte) to maintain memory
read alignment.

Align mpu memory allocation to be at 512 byte boundaries.

Reduce force-close allocation from 10000 objects to 64 objects.

Signed-off-by: John Miller <john.miller@atomicrules.com>
2 years agonet/ark: update packet generator and checker config status
John Miller [Tue, 15 Feb 2022 22:19:23 +0000 (16:19 -0600)]
net/ark: update packet generator and checker config status

Configuration status updates for internal packet checker and
generator.

Signed-off-by: John Miller <john.miller@atomicrules.com>
2 years agonet/ark: support arbitrary mbuf size
John Miller [Tue, 15 Feb 2022 22:19:22 +0000 (16:19 -0600)]
net/ark: support arbitrary mbuf size

Added arbitrary mbuf size per queue capability.
Updated ARK_UDM_CONST3 value to reflect the version number
read from the HW that is required to support this change.

Signed-off-by: John Miller <john.miller@atomicrules.com>
2 years agonet/ark: support new devices
John Miller [Tue, 15 Feb 2022 22:19:21 +0000 (16:19 -0600)]
net/ark: support new devices

Add two new supported device ID's.
Add documentation for new devices.

Signed-off-by: John Miller <john.miller@atomicrules.com>
2 years agonet/ark: add device capabilities record
John Miller [Tue, 15 Feb 2022 22:19:20 +0000 (16:19 -0600)]
net/ark: add device capabilities record

Add a device capabilities record for supported features.
Certain variants require that PCIe read-requests be correctly
throttled. This is called "rqpacing" in Arkville, and has to do
with credit and flow control on certain Arkville implementations.

Signed-off-by: John Miller <john.miller@atomicrules.com>
2 years agonet/nfb: support more MAC addresses
Martin Spinler [Tue, 15 Feb 2022 12:55:43 +0000 (13:55 +0100)]
net/nfb: support more MAC addresses

Extend the eth_dev_ops by add/remove MAC address functions.

Signed-off-by: Martin Spinler <spinler@cesnet.cz>
2 years agonet/nfb: fix multicast/promiscuous mode switching
Martin Spinler [Tue, 15 Feb 2022 12:55:42 +0000 (13:55 +0100)]
net/nfb: fix multicast/promiscuous mode switching

In the firmware, the promisc mode overrides the multicast mode.
So when the promisc mode is turned off, driver must check if the
multicast mode was active before and conditionally reactivate it.

Fixes: 6435f9a0ac22 ("net/nfb: add new netcope driver")
Cc: stable@dpdk.org
Signed-off-by: Martin Spinler <spinler@cesnet.cz>
2 years agonet/nfb: use timestamp offload flag
Martin Spinler [Tue, 15 Feb 2022 12:55:41 +0000 (13:55 +0100)]
net/nfb: use timestamp offload flag

Rewrite the RX timestamp setup code to use standard offload flag.

Signed-off-by: Martin Spinler <spinler@cesnet.cz>
2 years agonet/nfb: skip log for zero-sized Tx burst
Martin Spinler [Tue, 15 Feb 2022 12:55:40 +0000 (13:55 +0100)]
net/nfb: skip log for zero-sized Tx burst

Zero-sized TX burst floods the log no more.

Signed-off-by: Martin Spinler <spinler@cesnet.cz>
2 years agonet/nfb: fix array indexes in deinit functions
Martin Spinler [Tue, 15 Feb 2022 12:55:39 +0000 (13:55 +0100)]
net/nfb: fix array indexes in deinit functions

The indexes in the for cycle were wrongly used and
the code accessed outside of the rxmac/txmac array.

Fixes: 6435f9a0ac22 ("net/nfb: add new netcope driver")
Cc: stable@dpdk.org
Signed-off-by: Martin Spinler <spinler@cesnet.cz>
2 years agonet/enic: fix dereference before null check
John Daley [Mon, 14 Feb 2022 20:46:10 +0000 (12:46 -0800)]
net/enic: fix dereference before null check

This patch fixes an issue found by coverity. It checks for a possible
null value of "error" before dereferencing it.

CID: 375064: Dereference after null check
Either the check against null is unnecessary, or there may be a null
pointer dereference.

Coverity issue: 375064
Fixes: ee806eea59fe ("net/enic: support GENEVE flow item")

Signed-off-by: John Daley <johndale@cisco.com>
Reviewed-by: Hyong Youb Kim <hyonkim@cisco.com>
2 years agonet/iavf: support L2TPv2 for flow director
Jie Wang [Tue, 15 Feb 2022 08:25:37 +0000 (16:25 +0800)]
net/iavf: support L2TPv2 for flow director

Add support for L2TPv2(include PPP over L2TPv2) protocols FDIR
based on outer MAC src/dst address and L2TPv2 session ID.

Add support for PPPoL2TPv2oUDP protocols FDIR based on inner IP
src/dst address and UDP/TCP src/dst port.

Patterns are listed below:
eth/ipv4(6)/udp/l2tpv2
eth/ipv4(6)/udp/l2tpv2/ppp

eth/ipv4(6)/udp/l2tpv2/ppp/ipv4(6)
eth/ipv4(6)/udp/l2tpv2/ppp/ipv4(6)/udp
eth/ipv4(6)/udp/l2tpv2/ppp/ipv4(6)/tcp

Signed-off-by: Jie Wang <jie1x.wang@intel.com>
Acked-by: Beilei Xing <beilei.xing@intel.com>
2 years agonet/iavf: support L2TPv2 for RSS
Jie Wang [Tue, 15 Feb 2022 08:25:36 +0000 (16:25 +0800)]
net/iavf: support L2TPv2 for RSS

Add support for L2TPv2(include PPP over L2TPv2) protocols RSS based
on outer MAC src/dst address and L2TPv2 session ID.

Patterns are listed below:
eth/ipv4/udp/l2tpv2
eth/ipv4/udp/l2tpv2/ppp
eth/ipv6/udp/l2tpv2
eth/ipv6/udp/l2tpv2/ppp

Signed-off-by: Jie Wang <jie1x.wang@intel.com>
Acked-by: Beilei Xing <beilei.xing@intel.com>
2 years agonet/i40e: add flow mark capability to NEON Rx
Joyce Kong [Fri, 17 Dec 2021 05:36:00 +0000 (05:36 +0000)]
net/i40e: add flow mark capability to NEON Rx

This commit adds a flow director support to i40e NEON
vector RX path.

I40e can have 16 and 32 byte descriptors, and the Flow
Director ID data and indication-bit are in different
locations for each size descriptor. The support is
implemented in two separate functions as they require
vastly different operations.

The 16B descriptor re-purposes the "filter-status" u32
field to indicate FDIR ID when the FLM bit is set. No
extra loads are required, however we do have to store
to mbuf->fdir.hi, which is not stored to in the RX path
before this patch.

The 32B descriptor requires loading the 2nd 16 bytes of
each descriptor, to get the FLEXBH_STAT and FD Filter ID
from qword3. The resulting data must also be stored to
mbuf->fdir.hi, same as the 16B code path.

Signed-off-by: Joyce Kong <joyce.kong@arm.com>
Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
2 years agodoc: support IPsec Multi-buffer lib v1.2
Pablo de Lara [Wed, 23 Feb 2022 18:22:45 +0000 (18:22 +0000)]
doc: support IPsec Multi-buffer lib v1.2

Updated AESNI MB and AESNI GCM, KASUMI, ZUC and SNOW3G PMD documentation
guides with information about the latest Intel IPSec Multi-buffer
library supported.

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
2 years agocrypto/ipsec_mb: fix length and offset settings
Pablo de Lara [Wed, 23 Feb 2022 16:01:16 +0000 (16:01 +0000)]
crypto/ipsec_mb: fix length and offset settings

KASUMI, SNOW3G and ZUC require lengths and offsets to
be set in bits or bytes depending on the algorithm.
There were some algorithms that were mixing these two,
so this commit is fixing this issue.

Fixes: ae8e085c608d ("crypto/aesni_mb: support KASUMI F8/F9")
Fixes: 6c42e0cf4d12 ("crypto/aesni_mb: support SNOW3G-UEA2/UIA2")
Fixes: fd8df85487c4 ("crypto/aesni_mb: support ZUC-EEA3/EIA3")
Fixes: 8c835018de84 ("crypto/ipsec_mb: support ZUC-256 for aesni_mb")
Cc: stable@dpdk.org
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
2 years agocrypto/ipsec_mb: fix ZUC operation overwrite
Pablo de Lara [Wed, 23 Feb 2022 16:01:15 +0000 (16:01 +0000)]
crypto/ipsec_mb: fix ZUC operation overwrite

ZUC PMD batches crypto operations depending on their type
(encryption + tag generation, tag verification + decryption, etc),
to allow parallelization.
The array used to store the pointers to these operations was
always the same array provided by dequeue_burst() function,
and it was looping around the same positions (from 0 to ZUC_MAX_BURST - 1).

A new internal array is used to avoid overwriting the pointers
of the array provided by dequeue_burst() function.

Fixes: cf7685d68f00 ("crypto/zuc: add driver for ZUC library")
Cc: stable@dpdk.org
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
2 years agocrypto/ipsec_mb: fix ZUC authentication verify
Pablo de Lara [Wed, 23 Feb 2022 16:01:14 +0000 (16:01 +0000)]
crypto/ipsec_mb: fix ZUC authentication verify

ZUC authentication is done over multiple buffers at a time.
When authentication verification is done, multiple scratch buffers
are using to generate the tags that will be compared afterwards.
However, the same scratch buffer was used always, instead of having
different ones for each crypto operation.

Fixes: 0b133c36ad7d ("crypto/zuc: support IPsec Multi-buffer lib v0.54")
Cc: stable@dpdk.org
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
2 years agocrypto/ipsec_mb: check missing operation types
Pablo de Lara [Wed, 23 Feb 2022 16:01:13 +0000 (16:01 +0000)]
crypto/ipsec_mb: check missing operation types

When processing crypto operations in ZUC PMD,
there were two operation types that were set at session level,
but not checked when the operations are enqueued and processed,
leaving the buffers untouched silently.

Fixes: cde8df1bda9d ("crypto/ipsec_mb: move zuc PMD")
Cc: stable@dpdk.org
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
2 years agocrypto/virtio: fix out-of-bounds access
Brian Dooley [Tue, 22 Feb 2022 09:54:51 +0000 (09:54 +0000)]
crypto/virtio: fix out-of-bounds access

Coverity flags an untrusted loop bound. Check length of session iv.

Coverity issue: 375802
Fixes: b063e843fa03 ("crypto/virtio: fix IV physical address")
Cc: stable@dpdk.org
Signed-off-by: Brian Dooley <brian.dooley@intel.com>
Acked-by: Fan Zhang <roy.fan.zhang@intel.com>
2 years agobaseband/acc100: avoid out-of-bounds access
Nicolas Chautru [Tue, 22 Feb 2022 23:43:36 +0000 (15:43 -0800)]
baseband/acc100: avoid out-of-bounds access

The actual maximum number of gather pointers should have been limited

Coverity issue: 375803, 375813, 375819, 375827, 375831
Fixes: 5ad5060f8f7a ("baseband/acc100: add LDPC processing functions")
Cc: stable@dpdk.org
Signed-off-by: Nicolas Chautru <nicolas.chautru@intel.com>
2 years agoexamples/l2fwd-crypto: fix port mask overflow
Brian Dooley [Mon, 21 Feb 2022 18:06:58 +0000 (18:06 +0000)]
examples/l2fwd-crypto: fix port mask overflow

Coverity flags an issue with 32-bit value. If max ethports value is
configured with a value larger than 32 there will be an issue.
Coverity issue: 375863 Unintentional integer overflow

Fixes: 387259bd6c67 ("examples/l2fwd-crypto: add sample application")
Cc: stable@dpdk.org
Signed-off-by: Brian Dooley <brian.dooley@intel.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
2 years agotest/crypto: add L4 checksum case for transport mode
Volodymyr Fialko [Wed, 23 Feb 2022 10:40:46 +0000 (11:40 +0100)]
test/crypto: add L4 checksum case for transport mode

Added test case to verify L4 checksum offload in IPsec transport mode.

Signed-off-by: Volodymyr Fialko <vfialko@marvell.com>
Acked-by: Anoob Joseph <anoobj@marvell.com>
2 years agotest/crypto: add TTL and hop limit decrement cases
Volodymyr Fialko [Wed, 23 Feb 2022 10:40:45 +0000 (11:40 +0100)]
test/crypto: add TTL and hop limit decrement cases

Add test cases to verify TTL and hop limit decrement with lookaside
IPsec offload.

Signed-off-by: Volodymyr Fialko <vfialko@marvell.com>
Acked-by: Anoob Joseph <anoobj@marvell.com>
2 years agodoc: fix FIPS guide
Jakub Poczatek [Wed, 9 Feb 2022 11:36:23 +0000 (11:36 +0000)]
doc: fix FIPS guide

Added missing step for converting SHA request files to correct
format. Replaced AES_GCM with GCM to follow the correct
naming format.

Fixes: 3d0fad56b74 ("examples/fips_validation: add crypto FIPS application")
Cc: stable@dpdk.org
Signed-off-by: Jakub Poczatek <jakub.poczatek@intel.com>
Acked-by: Fan Zhang <roy.fan.zhang@intel.com>
2 years agoexamples/ipsec-secgw: add pool size parameters
Nithin Dabilpuram [Wed, 23 Feb 2022 09:53:53 +0000 (15:23 +0530)]
examples/ipsec-secgw: add pool size parameters

Add support to enable per port packet pool and also override
vector pool size from command line args. This is useful
on some HW to tune performance based on usecase.

Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
2 years agoexamples/ipsec-secgw: fix buffer freeing in vector mode
Nithin Dabilpuram [Wed, 23 Feb 2022 09:53:52 +0000 (15:23 +0530)]
examples/ipsec-secgw: fix buffer freeing in vector mode

Fix packet processing to skip after mbuf is freed instead of
touching and Tx'ing it.

Also free vector event buffer in event worker when after processing
there is no pkt to be enqueued to Tx adapter.

Fixes: 86738ebe1e3d ("examples/ipsec-secgw: support event vector")
Cc: stable@dpdk.org
Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
2 years agoexamples/ipsec-secgw: avoid logs in data path
Nithin Dabilpuram [Wed, 23 Feb 2022 09:53:51 +0000 (15:23 +0530)]
examples/ipsec-secgw: avoid logs in data path

Update error prints in data path to RTE_LOG_DP().
Error prints in fast path are not good for performance
as they slow down the application when few bad packets are
received.

Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
2 years agoexamples/ipsec-secgw: add check for unprotected port mask
Volodymyr Fialko [Wed, 9 Feb 2022 11:52:13 +0000 (12:52 +0100)]
examples/ipsec-secgw: add check for unprotected port mask

Usage of unprotected port mask without any configured SA inbound, will
cause use of uninitialized SA context, so disallow such configuration.

Signed-off-by: Volodymyr Fialko <vfialko@marvell.com>
Acked-by: Anoob Joseph <anoobj@marvell.com>
2 years agocrypto/qat: refactor asymmetric session
Arek Kusztal [Mon, 21 Feb 2022 10:48:31 +0000 (10:48 +0000)]
crypto/qat: refactor asymmetric session

This patch refactors asymmetric session in Intel
QuickAssist Technology PMD and fixes some issues
with xform. Code will be now bit more scalable,
and easier readable.

Signed-off-by: Arek Kusztal <arkadiuszx.kusztal@intel.com>
2 years agocrypto/qat: add ECPM algorithm
Arek Kusztal [Mon, 21 Feb 2022 10:48:30 +0000 (10:48 +0000)]
crypto/qat: add ECPM algorithm

This patch adds Elliptic Curve Multiplication
algorithm to Intel QuickAssist Technology PMD.

Signed-off-by: Arek Kusztal <arkadiuszx.kusztal@intel.com>
2 years agocrypto/qat: add ECDSA algorithm
Arek Kusztal [Mon, 21 Feb 2022 10:48:29 +0000 (10:48 +0000)]
crypto/qat: add ECDSA algorithm

This patch adds ECDSA algorithm to Intel
QuickAssist Technology PMD.

Signed-off-by: Arek Kusztal <arkadiuszx.kusztal@intel.com>
2 years agocrypto/qat: add named elliptic curves
Arek Kusztal [Mon, 21 Feb 2022 10:48:28 +0000 (10:48 +0000)]
crypto/qat: add named elliptic curves

This patch adds secp256r1 and secp521r1 elliptic curves to Intel
QuickAssist Technology PMD.

Signed-off-by: Arek Kusztal <arkadiuszx.kusztal@intel.com>
2 years agocrypto/qat: refactor asymmetric crypto functions
Arek Kusztal [Mon, 21 Feb 2022 10:48:27 +0000 (10:48 +0000)]
crypto/qat: refactor asymmetric crypto functions

This commit refactors asummetric crypto functions
in Intel QuickAssist Technology PMD.
Functions right now are shorter and far easier readable,
plus it facilitates addition of new algorithms.

Signed-off-by: Arek Kusztal <arkadiuszx.kusztal@intel.com>
2 years agocrypto/qat: support out-of-place SG list
Kai Ji [Wed, 23 Feb 2022 00:50:06 +0000 (08:50 +0800)]
crypto/qat: support out-of-place SG list

This patch adds the SGL out of place support to QAT PMD

Signed-off-by: Kai Ji <kai.ji@intel.com>
Acked-by: Fan Zhang <roy.fan.zhang@intel.com>
2 years agocrypto/qat: unify raw data path functions
Kai Ji [Wed, 23 Feb 2022 00:50:05 +0000 (08:50 +0800)]
crypto/qat: unify raw data path functions

This patch unifies QAT's raw dp api implementations
to the same enqueue/dequeue methods used in crypto operations.
The specific functions for different QAT generation are updated
respectively. The qat_sym_hw_dp.c is removed as no longer required.

Signed-off-by: Kai Ji <kai.ji@intel.com>
Acked-by: Fan Zhang <roy.fan.zhang@intel.com>
2 years agocrypto/qat: rework burst data path
Kai Ji [Wed, 23 Feb 2022 00:50:04 +0000 (08:50 +0800)]
crypto/qat: rework burst data path

This patch enable the op_build_request function in
qat_enqueue_op_burst, and the qat_dequeue_process_response
function in qat_dequeue_op_burst.
The op_build_request invoked in crypto build request op is based
on crypto operations setup'd during session init.

Signed-off-by: Kai Ji <kai.ji@intel.com>
Acked-by: Fan Zhang <roy.fan.zhang@intel.com>
2 years agocrypto/qat: unify asymmetric functions
Kai Ji [Wed, 23 Feb 2022 00:50:03 +0000 (08:50 +0800)]
crypto/qat: unify asymmetric functions

This patch removes qat_asym_pmd.c and integrates all the
functions into qat_asym.c. The unified/integrated asym crypto
pmd functions should make them easier to maintain.

Signed-off-by: Kai Ji <kai.ji@intel.com>
Acked-by: Fan Zhang <roy.fan.zhang@intel.com>
2 years agocrypto/qat: unify symmetric functions
Kai Ji [Wed, 23 Feb 2022 00:50:02 +0000 (08:50 +0800)]
crypto/qat: unify symmetric functions

This patch removes qat_sym_pmd.c and integrates all the functions into
qat_sym.c. The unified/integrated qat sym crypto pmd functions should
make them easier to maintain.

Signed-off-by: Kai Ji <kai.ji@intel.com>
Acked-by: Fan Zhang <roy.fan.zhang@intel.com>
2 years agocrypto/qat: rework asymmetric op build operation
Kai Ji [Wed, 23 Feb 2022 00:50:01 +0000 (08:50 +0800)]
crypto/qat: rework asymmetric op build operation

This patch reworks the asymmetric crypto data path
implementation in QAT driver. The changes include asymmetric
crypto data path separation for QAT hardware generations, and
code optimisation of the device capabilities declaration.

Signed-off-by: Kai Ji <kai.ji@intel.com>
Acked-by: Fan Zhang <roy.fan.zhang@intel.com>
2 years agocrypto/qat: rework session functions
Kai Ji [Wed, 23 Feb 2022 00:50:00 +0000 (08:50 +0800)]
crypto/qat: rework session functions

This patch introduces a set of set_session methods to QAT
generations. In addition, the reuse of QAT session between
generations is prohibit as the support of min_qat_dev_gen_id'
is removed.

Signed-off-by: Kai Ji <kai.ji@intel.com>
Acked-by: Fan Zhang <roy.fan.zhang@intel.com>
2 years agocrypto/qat: support symmetric build op request
Kai Ji [Wed, 23 Feb 2022 00:49:59 +0000 (08:49 +0800)]
crypto/qat: support symmetric build op request

This patch adds common inline functions for QAT symmetric
crypto driver to process crypto op, and the implementation of
build op request function for QAT generation 1.

Signed-off-by: Kai Ji <kai.ji@intel.com>
Acked-by: Fan Zhang <roy.fan.zhang@intel.com>
2 years agocommon/qat: define build request and dequeue operations
Kai Ji [Wed, 23 Feb 2022 00:49:58 +0000 (08:49 +0800)]
common/qat: define build request and dequeue operations

This patch introduce build request and dequeue op function
pointers to the qat queue pair implementation. The function
pointers are assigned during qat session generation based on input
crypto operation request.

Signed-off-by: Kai Ji <kai.ji@intel.com>
Acked-by: Fan Zhang <roy.fan.zhang@intel.com>
2 years agocrypto/cnxk: support TTL or hop limit decrement
Volodymyr Fialko [Fri, 18 Feb 2022 12:49:23 +0000 (13:49 +0100)]
crypto/cnxk: support TTL or hop limit decrement

Add support for TTL if IPv4 or hop limit if IPv6 decrement for cn10k.

Signed-off-by: Volodymyr Fialko <vfialko@marvell.com>
Acked-by: Anoob Joseph <anoobj@marvell.com>
2 years agocryptodev: fix RSA key type name
Arek Kusztal [Fri, 18 Feb 2022 07:53:32 +0000 (07:53 +0000)]
cryptodev: fix RSA key type name

This patch fixes misspelled RTE_RSA_KEY_TYPE_QT,
this will prevent checkpach from complaining wherever
change to RSA is being made.

Fixes: 26008aaed14c ("cryptodev: add asymmetric xform and op definitions")
Cc: stable@dpdk.org
Signed-off-by: Arek Kusztal <arkadiuszx.kusztal@intel.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
2 years agocrypto/ipsec_mb: support null auth/cipher for aesni_mb
Fan Zhang [Mon, 21 Feb 2022 13:39:46 +0000 (13:39 +0000)]
crypto/ipsec_mb: support null auth/cipher for aesni_mb

Add NULL cipher and auth support to AESNI-MB PMD type.

Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
Signed-off-by: Declan Doherty <declan.doherty@intel.com>
Signed-off-by: Radu Nicolau <radu.nicolau@intel.com>
2 years agocryptodev: fix asymmetric private session variable size
Ciara Power [Fri, 18 Feb 2022 12:57:52 +0000 (12:57 +0000)]
cryptodev: fix asymmetric private session variable size

When creating the asymmetric session mempool, the maximum private
session size of all devices is used when creating the mempool
object size.
The return value for ``rte_cryptodev_asym_get_private_session_size``
is unsigned int, whereas the variable was uint8_t, leading to a
possible overflow issue.

To fix this, the variable for private session size is now changed to
unsigned int to match the function return type.

Fixes: 1f1e4b7cbaad ("cryptodev: use single mempool for asymmetric session")

Reported-by: Arek Kusztal <arkadiuszx.kusztal@intel.com>
Signed-off-by: Ciara Power <ciara.power@intel.com>
Acked-by: Fan Zhang <roy.fan.zhang@intel.com>
Acked-by: Arek Kusztal <arkadiuszx.kusztal@intel.com>
2 years agocrypto/ipsec_mb: remove useless check
Ciara Power [Tue, 15 Feb 2022 11:07:58 +0000 (11:07 +0000)]
crypto/ipsec_mb: remove useless check

The qp is never NULL when it reaches the qp_setup_cleanup error handling
code. This is due to checks earlier in the function that return an error
if qp is NULL.

Coverity issue: 375013
Fixes: 581c39b1cb7d ("crypto/ipsec_mb: fix queue cleanup null pointer dereference")
Cc: stable@dpdk.org
Signed-off-by: Ciara Power <ciara.power@intel.com>
2 years agotest/crypto: add ESN and antireplay cases
Tejasree Kondoj [Mon, 31 Jan 2022 16:43:57 +0000 (22:13 +0530)]
test/crypto: add ESN and antireplay cases

Added test cases for IPsec ESN and Antireplay.

Signed-off-by: Tejasree Kondoj <ktejasree@marvell.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
2 years agotest/crypto: add copy and set DSCP cases
Anoob Joseph [Mon, 31 Jan 2022 16:43:56 +0000 (22:13 +0530)]
test/crypto: add copy and set DSCP cases

Add test cases to verify copy and set DSCP with
IPv4 and IPv6 tunnels.

Signed-off-by: Anoob Joseph <anoobj@marvell.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
2 years agoapp/eventdev: add crypto producer mode
Shijith Thotton [Thu, 24 Feb 2022 04:46:07 +0000 (10:16 +0530)]
app/eventdev: add crypto producer mode

In crypto producer mode, producer core enqueues cryptodev with software
generated crypto ops and worker core dequeues crypto completion events
from the eventdev. Event crypto metadata used for above processing is
pre-populated in each crypto session.

Parameter --prod_type_cryptodev can be used to enable crypto producer
mode. Parameter --crypto_adptr_mode can be set to select the crypto
adapter mode, 0 for OP_NEW and 1 for OP_FORWARD.

This mode can be used to measure the performance of crypto adapter.

Example:
  ./dpdk-test-eventdev -l 0-2 -w <EVENTDEV> -w <CRYPTODEV> -- \
  --prod_type_cryptodev --crypto_adptr_mode 1 --test=perf_atq \
  --stlist=a --wlcores 1 --plcores 2

Signed-off-by: Shijith Thotton <sthotton@marvell.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
Acked-by: Abhinandan Gujjar <abhinandan.gujjar@intel.com>
2 years agoeventdev/crypto: remove useless check
Ganapati Kundapura [Wed, 23 Feb 2022 07:34:43 +0000 (01:34 -0600)]
eventdev/crypto: remove useless check

eca_cryptodev_cdev_flush() is internal function and called with
valid range of cdevs.

crypto_cdev_info structure is allocated at adapter creation time
and retrieved from the adapter for a valid cdevs which cannot be NULL
and hence no need for NULL check.

Fixes: 2ae84b39ae7b ("eventdev/crypto: store operations in circular buffer")

Signed-off-by: Ganapati Kundapura <ganapati.kundapura@intel.com>
Acked-by: Abhinandan Gujjar <abhinandan.gujjar@intel.com>
2 years agoevent/cnxk: update SQB FC check for Tx adapter
Pavan Nikhilesh [Sat, 19 Feb 2022 12:13:38 +0000 (17:43 +0530)]
event/cnxk: update SQB FC check for Tx adapter

Update SQB limit to include CPT queue size when Security
offload is enabled.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
2 years agoevent/cnxk: remove deschedule usage in CN9K
Pavan Nikhilesh [Sat, 19 Feb 2022 12:13:37 +0000 (17:43 +0530)]
event/cnxk: remove deschedule usage in CN9K

Using deschedule cmd might incorrectly ignore updates
to WQE, GGRP on CN9K.
Use addwork to pipeline work instead.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
2 years agoevent/cnxk: fix uninitialized local variables
Gowrishankar Muthukrishnan [Sat, 12 Feb 2022 12:39:48 +0000 (18:09 +0530)]
event/cnxk: fix uninitialized local variables

Fix uninitialized local variable as reported in coverity scan.

Coverity issue: 370578, 370579, 370587
Fixes: 300b796262a ("event/cnxk: add timer arm routine")
Cc: stable@dpdk.org
Signed-off-by: Gowrishankar Muthukrishnan <gmuthukrishn@marvell.com>
2 years agoevent/cnxk: fix memory leaks
Gowrishankar Muthukrishnan [Sat, 12 Feb 2022 12:39:47 +0000 (18:09 +0530)]
event/cnxk: fix memory leaks

Fix memory leaks reported by coverity scan in
cnxk_tim_parse_clk_list()  and cnxk_tim_parse_ring_ctl_list()
functions on not freeing strduped memory.

Coverity issue: 374990 374991
Fixes: 8a3d58c189f ("event/cnxk: add option to control timer adapters")
Cc: stable@dpdk.org
Signed-off-by: Gowrishankar Muthukrishnan <gmuthukrishn@marvell.com>
2 years agoevent/cnxk: fix variables casting
Gowrishankar Muthukrishnan [Sat, 12 Feb 2022 12:39:46 +0000 (18:09 +0530)]
event/cnxk: fix variables casting

Fix incompatible casting on variables reported by coverity scan.

Coverity issue: 370582 370584
Fixes: 6223ede2036 ("event/cnxk: add event port link and unlink")
Cc: stable@dpdk.org
Signed-off-by: Gowrishankar Muthukrishnan <gmuthukrishn@marvell.com>
2 years agoevent/dlb2: poll HW CQ inflights before mapping queue
Timothy McDaniel [Wed, 16 Feb 2022 19:38:27 +0000 (13:38 -0600)]
event/dlb2: poll HW CQ inflights before mapping queue

When attempting to link a port and queue immediately after unlinking,
the CQ inflights may not all be processed. Poll the h/w register for
outstanding inflights instead of reading once, in case the inflights
are still being processed. Also return EBUSY if the inflight
processing is not completed in a suitable amount of time.

Fixes: 1857f1922ce2 ("event/dlb2: use new implementation of resource file")
Cc: stable@dpdk.org
Signed-off-by: Timothy McDaniel <timothy.mcdaniel@intel.com>
2 years agoevent/dlb2: update rolling mask used for dequeue
Timothy McDaniel [Wed, 16 Feb 2022 19:38:26 +0000 (13:38 -0600)]
event/dlb2: update rolling mask used for dequeue

Update the rolling mask used in dequeue operations to
fix the vector optimized dequeue.

Fixes: 000a7b8e7582 ("event/dlb2: optimize dequeue operation")
Cc: stable@dpdk.org
Signed-off-by: Timothy McDaniel <timothy.mcdaniel@intel.com>
2 years agoeventdev/eth_rx: fix queue config query
Naga Harish K S V [Mon, 14 Feb 2022 17:10:16 +0000 (11:10 -0600)]
eventdev/eth_rx: fix queue config query

Remove the memcpy usage in queue config get function for
`event` variable which is 8 byte size and use direct copy.
Also provide vector information and event buffer size in the
queue config info.

Fixes: da781e6488 ("eventdev/eth_rx: support Rx queue config get")
Cc: stable@dpdk.org
Signed-off-by: Naga Harish K S V <s.v.naga.harish.k@intel.com>
Acked-by: Jay Jayatheerthan <jay.jayatheerthan@intel.com>
2 years agoeventdev/eth_rx: fix parameters parsing memory leak
Weiguo Li [Fri, 4 Feb 2022 08:21:28 +0000 (16:21 +0800)]
eventdev/eth_rx: fix parameters parsing memory leak

The memory get from strdup should be freed when parameter parsing
finished, and also should be freed when error occurs.

Fixes: 814d01709328 ("eventdev/eth_rx: support telemetry")
Fixes: 9e583185318f ("eventdev/eth_rx: support telemetry")
Cc: stable@dpdk.org
Signed-off-by: Weiguo Li <liwg06@foxmail.com>
Acked-by: Ganapati Kundapura <ganapati.kundapura@intel.com>
Acked-by: Jay Jayatheerthan <jay.jayatheerthan@intel.com>
2 years agoexamples/qos_sched: fix core mask overflow
Megha Ajmera [Wed, 23 Feb 2022 17:36:30 +0000 (17:36 +0000)]
examples/qos_sched: fix core mask overflow

Masking of core mask was incorrect. Instead of using 1U for shifting, it
should be using 1LU as the result is assigned to uint64.

CID 375859: Potentially overflowing expression "1U << app_main_core" with
type "unsigned int" (32 bits, unsigned) is evaluated using 32-bit
arithmetic, and then used in a context that expects an expression of
type "uint64_t" (64 bits, unsigned).

Coverity issue: 375859
Fixes: de3cfa2c9823 ("sched: initial import")
Cc: stable@dpdk.org
Signed-off-by: Megha Ajmera <megha.ajmera@intel.com>
Acked-by: Morten Brørup <mb@smartsharesystems.com>
2 years agosched: enable traffic class oversubscription unconditionally
Megha Ajmera [Tue, 22 Feb 2022 12:57:45 +0000 (12:57 +0000)]
sched: enable traffic class oversubscription unconditionally

Removed RTE_SCHED_SUBPORT_TC_OV from rte_config.h.
Best effort traffic class oversubscription is always enabled.

Signed-off-by: Megha Ajmera <megha.ajmera@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
2 years agosched: enable statistics unconditionally
Megha Ajmera [Tue, 22 Feb 2022 12:57:44 +0000 (12:57 +0000)]
sched: enable statistics unconditionally

Removed RTE_SCHED_COLLECT_STATS flag from rte_config.h.
Stats collection is always enabled.

Signed-off-by: Megha Ajmera <megha.ajmera@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
2 years agosched: move grinder configuration
Megha Ajmera [Tue, 22 Feb 2022 12:57:43 +0000 (12:57 +0000)]
sched: move grinder configuration

Grinder configuration is now moved to sched library.

Number of grinders can also modified by specifying
RTE_SCHED_PORT_N_GRINDERS=N in CFLAGS, where N is number of grinders.

Signed-off-by: Megha Ajmera <megha.ajmera@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
2 years agosched: remove vector functions
Megha Ajmera [Tue, 22 Feb 2022 12:57:42 +0000 (12:57 +0000)]
sched: remove vector functions

Remove RTE_SCHED_VECTOR flag from rte_config.h.
This flag is no longer useful. Only scalar version is supported.

Signed-off-by: Megha Ajmera <megha.ajmera@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
2 years agoraw/cnxk_gpio: check pointer before use in self test
Tomasz Duszynski [Tue, 22 Feb 2022 08:28:17 +0000 (09:28 +0100)]
raw/cnxk_gpio: check pointer before use in self test

Even though nonexistent cnxk_gpio is unlikely at this point
it's a good practice to check pointers before using them.

Fixes: 0e6557b448fa ("raw/cnxk_gpio: add self test")
Coverity issue: 376502

Signed-off-by: Tomasz Duszynski <tduszynski@marvell.com>
Reviewed-by: Jerin Jacob <jerinj@marvell.com>
2 years agoraw/cnxk_gpio: fix resource leak
Tomasz Duszynski [Tue, 22 Feb 2022 08:28:16 +0000 (09:28 +0100)]
raw/cnxk_gpio: fix resource leak

All used resources need to be properly cleaned up in error path.

Fixes: 0e6557b448fa ("raw/cnxk_gpio: add self test")
Coverity issue: 376504

Signed-off-by: Tomasz Duszynski <tduszynski@marvell.com>
Reviewed-by: Jerin Jacob <jerinj@marvell.com>
2 years agoraw/cnxk_gpio: fix ABI version
Tomasz Duszynski [Wed, 23 Feb 2022 13:32:10 +0000 (14:32 +0100)]
raw/cnxk_gpio: fix ABI version

PMD driver got merged during 22.03 merge window and number in map file
should reflect that.

Fixes: d0b8a4e19131 ("raw/cnxk_gpio: add GPIO driver skeleton")

Reported-by: Ferruh Yigit <ferruh.yigit@intel.com>
Signed-off-by: Tomasz Duszynski <tduszynski@marvell.com>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
2 years agodoc: improve configuration examples in idxd guide
Bruce Richardson [Wed, 16 Feb 2022 16:06:09 +0000 (16:06 +0000)]
doc: improve configuration examples in idxd guide

The documentation on how to configure device instances using
accel-config can be improved by a number of changes:

* For initial example, when only configuring one queue, omit
  configuration of a second engine, which is unused later.
* Add the "max-batch-size" setting to the options being configured for
  each queue
* Add a final, more complete example, showing configuration of multiple
  queues on a device.

Cc: stable@dpdk.org
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Reviewed-by: Kevin Laatz <kevin.laatz@intel.com>
2 years agodma/idxd: configure maximum batch size to high value
Bruce Richardson [Wed, 16 Feb 2022 16:06:08 +0000 (16:06 +0000)]
dma/idxd: configure maximum batch size to high value

When configuring an Intel DSA instance using the utility script
dpdk_idxd_cfg.py, explicitly set the max supported batch size value to a
high value, to ensure large bursts are supported if so desired. The
default in the linux kernel is now just 32 [1], which may not be
sufficient for all DPDK apps.

[1] https://lore.kernel.org/r/163528473483.3926048.7950067926287180976.stgit@djiang5-desk3.ch.intel.com

Cc: stable@dpdk.org
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Reviewed-by: Kevin Laatz <kevin.laatz@intel.com>
2 years agotest/dma: fix missing checks for device capacity
Bruce Richardson [Wed, 16 Feb 2022 16:06:07 +0000 (16:06 +0000)]
test/dma: fix missing checks for device capacity

For some DMA HW devices, e.g. those using the idxd driver, the maximum
burst size is configurable, which can lead to test failures if the value
is set too small. Add explicit check for this to give reasonable error
messages for devices which need their config adjusted.

Fixes: 1b86a66a30c2 ("test/dma: add more comprehensive copy tests")
Fixes: 8fa5d2683940 ("test/dma: add burst capacity test")
Cc: stable@dpdk.org
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Reviewed-by: Conor Walsh <conor.walsh@intel.com>
Reviewed-by: Kevin Laatz <kevin.laatz@intel.com>
2 years agodma/hisilicon: use common PCI device naming
Chengwen Feng [Thu, 17 Feb 2022 02:59:11 +0000 (10:59 +0800)]
dma/hisilicon: use common PCI device naming

For DMA device 0000:7d:0.0, the original generated dmadev name starts
with the "7d:0.0", which is not expected.
This patch uses rte_pci_device_name API to generates the dmadev name.

Cc: stable@dpdk.org
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
2 years agodma/hisilicon: add queue full statistics
Chengwen Feng [Thu, 17 Feb 2022 02:59:10 +0000 (10:59 +0800)]
dma/hisilicon: add queue full statistics

This patch adds queue full statistics for HiSilicon DMA PMD.

Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
2 years agodma/hisilicon: support registers dump for Kunpeng 930
Chengwen Feng [Thu, 17 Feb 2022 02:59:09 +0000 (10:59 +0800)]
dma/hisilicon: support registers dump for Kunpeng 930

This patch supports dump Kunpeng930 DMA registers.

Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
2 years agodma/hisilicon: support error handling with Kunpeng 930
Chengwen Feng [Thu, 17 Feb 2022 02:59:08 +0000 (10:59 +0800)]
dma/hisilicon: support error handling with Kunpeng 930

The Kunpeng930 DMA supports the capability of handles errors.

Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
2 years agodma/hisilicon: support Kunpeng 930
Chengwen Feng [Thu, 17 Feb 2022 02:59:07 +0000 (10:59 +0800)]
dma/hisilicon: support Kunpeng 930

The Kunpeng930 DMA devices have the same PCI device id with Kunpeng920,
but with different PCI revision and register layout. This patch
introduces the basic initialization for Kunpeng930 DMA devices.

Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
2 years agogpudev: use CPU mapping in communication list
Elena Agostini [Tue, 22 Feb 2022 00:22:33 +0000 (00:22 +0000)]
gpudev: use CPU mapping in communication list

rte_gpu_mem_cpu_map() exposes a GPU memory area to the CPU.
In gpudev communication list this is useful to store the
status flag.

A communication list status flag allocated on GPU memory
and mapped for CPU visibility can be updated by CPU and polled
by a GPU workload.

The polling operation is more frequent than the CPU update operation.
Having the status flag in GPU memory reduces the GPU workload polling
latency.

If CPU mapping feature is not enabled, status flag resides in
CPU memory registered so it's visible from the GPU.

To facilitate the interaction with the status flag, this patch
provides also the set/get functions for it.

Signed-off-by: Elena Agostini <eagostini@nvidia.com>
2 years agogpudev: use device memory pointer for CPU unmap
Elena Agostini [Mon, 21 Feb 2022 23:07:06 +0000 (23:07 +0000)]
gpudev: use device memory pointer for CPU unmap

Update rte_gpu_mem_cpu_unmap() header documentation
and the test application to use GPU pointer when unmapping.

Signed-off-by: Elena Agostini <eagostini@nvidia.com>
2 years agobuild: make ring mempool driver mandatory
Bruce Richardson [Mon, 14 Feb 2022 15:30:38 +0000 (15:30 +0000)]
build: make ring mempool driver mandatory

The default mempool driver is one based on the rte_ring, and as such it
needs to be present to have just about any app (which doesn't override
the mempool) run. Given this state of affairs it is probably best to add
this default mempool driver to the always-enabled list to ensure we get
a runnable build in all cases. This means that, for example, to run some
NIC tests with testpmd in a minimal build, in most cases, all the user
should need to do is specify "-Denable_drivers=net/<nic_drv>" for the
build.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Morten Brørup <mb@smartsharesystems.com>
2 years agobuildtools/chkincs: check missing C++ guards
Bruce Richardson [Tue, 15 Feb 2022 17:30:29 +0000 (17:30 +0000)]
buildtools/chkincs: check missing C++ guards

Simply compiling a C header with a C++ compiler is not enough to flag
missing 'extern "C"' guards. To catch missing guards, we can just use a
simple grep for the 'extern "C"' part, and error out if any files have a
miss.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Tyler Retzlaff <roretzla@linux.microsoft.com>
2 years agobuildtools/chkincs: check SDK headers for C++ compatibility
Bruce Richardson [Tue, 15 Feb 2022 17:30:28 +0000 (17:30 +0000)]
buildtools/chkincs: check SDK headers for C++ compatibility

With a one-line change to the lib meson.build file we can add the SDK
headers to the list of files to be checked using the chkincs binary.
Unfortunately, many of those SDK header depend upon headers in the PCI
and vdev bus drivers, so we need to update chkincs build to ensure those
dependencies are added. We also need to allow internal APIs to be
present in these SDK headers.

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Tyler Retzlaff <roretzla@linux.microsoft.com>