dpdk.git
2 years agoapp/testpmd: use per-core variable in flowgen
Zhihong Wang [Fri, 13 Aug 2021 08:05:48 +0000 (16:05 +0800)]
app/testpmd: use per-core variable in flowgen

Use per-core variable for flow indexing to solve cache contention in
multi-core scenarios.

Signed-off-by: Zhihong Wang <wangzhihong.wzh@bytedance.com>
Acked-by: Xiaoyun Li <xiaoyun.li@intel.com>
2 years agoapp/testpmd: record Rx and dropped stats in flowgen
Zhihong Wang [Fri, 13 Aug 2021 08:05:47 +0000 (16:05 +0800)]
app/testpmd: record Rx and dropped stats in flowgen

Call inc_rx_burst_stats for rx operation, and record fwd_dropped.

Signed-off-by: Zhihong Wang <wangzhihong.wzh@bytedance.com>
Acked-by: Xiaoyun Li <xiaoyun.li@intel.com>
2 years agoapp/testpmd: use existing checksum API in flowgen engine
Zhihong Wang [Fri, 13 Aug 2021 08:05:46 +0000 (16:05 +0800)]
app/testpmd: use existing checksum API in flowgen engine

Use the rte_ipv4_cksum API to replace local ip_sum implementation.

Signed-off-by: Zhihong Wang <wangzhihong.wzh@bytedance.com>
Acked-by: Xiaoyun Li <xiaoyun.li@intel.com>
2 years agoapp/testpmd: fix Tx retry in flowgen engine
Zhihong Wang [Fri, 13 Aug 2021 08:05:45 +0000 (16:05 +0800)]
app/testpmd: fix Tx retry in flowgen engine

Fix tx_pkt number in tx retry logic.

Fixes: bf56fce1fb45 ("app/testpmd: add retry option")
Cc: stable@dpdk.org
Signed-off-by: Zhihong Wang <wangzhihong.wzh@bytedance.com>
Acked-by: Xiaoyun Li <xiaoyun.li@intel.com>
2 years agonet/mlx5: update GENEVE TLV option matching
Shiri Kuzin [Mon, 31 May 2021 11:45:43 +0000 (14:45 +0300)]
net/mlx5: update GENEVE TLV option matching

The GENEVE TLV option matching is done using a flex parser.

Recent update in firmware, requires that in order to match on the
GENEVE TLV option the "geneve_tlv_option_0_exist" bit should be set.

Add the new "geneve_tlv_option_0_exist" setting when translating the
GENEVE TLV option item.

Signed-off-by: Shiri Kuzin <shirik@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
2 years agonet/bnxt: update ring group after ring stop start
Ajit Khaparde [Mon, 2 Aug 2021 04:44:30 +0000 (21:44 -0700)]
net/bnxt: update ring group after ring stop start

A Rx ring stop start sequence may result in the FW returning
a different set of Rx ring and AGG ring IDs. If the ring group
is not updated with the new IDs, the HW sees the host driver using
incorrect BD types for the Rx ring and AGG ring. This can cause
the chip to go into a bad state or encounter RE_flush issue
or leak mbufs in the HW.

Fix this by issuing a bnxt_hwrm_ring_grp_free() and an
bnxt_hwrm_ring_grp_alloc() to refresh the ring group information.

Fixes: 9b63c6fd70e3 ("net/bnxt: support Rx/Tx queue start/stop")
Cc: stable@dpdk.org
Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
2 years agonet/mlx5: fix eCPRI matching
Dmitry Kozlyuk [Mon, 9 Aug 2021 14:26:46 +0000 (17:26 +0300)]
net/mlx5: fix eCPRI matching

When an ETH or VLAN flow item directly preceding ECPRI (i. e. a pattern
for eCPRI over Ethernet) did not specify the eCPRI protocol, matches
were not restricted to eCPRI traffic. For example, "eth / ecpri / end"
pattern behaved as "eth / end". Implicitly add Ethernet type condition,
so that "eth / ecpri / end" behaves as "eth type is 0xAEFE / end".

Fixes: daa38a8924a0 ("net/mlx5: add flow translation of eCPRI header")
Cc: stable@dpdk.org
Signed-off-by: Dmitry Kozlyuk <dkozlyuk@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
2 years agonet/mlx5: fix mbuf replenishment check for zipped CQE
Alexander Kozyrev [Wed, 4 Aug 2021 06:23:16 +0000 (09:23 +0300)]
net/mlx5: fix mbuf replenishment check for zipped CQE

A core dump is being generated with the following call stack:
0 _mm256_storeu_si256 (__A=..., __P=0x80)
1 rte_mov32 (src=0x2299c9140 "", dst=0x80)
2 rte_memcpy_aligned (n=60, src=0x2299c9140, dst=0x80)
3 rte_memcpy (n=60, src=0x2299c9140, dst=0x80)
4 mprq_buf_to_pkt (strd_cnt=1, strd_idx=0, buf=0x2299c8a00, len=60,
pkt=0x18345f0c0, rxq=0x18345ef40)
5 rxq_copy_mprq_mbuf_v (rxq=0x18345ef40, pkts=0x7f76e0ff6d18, pkts_n=5)
6 rxq_burst_mprq_v (rxq=0x18345ef40, pkts=0x7f76e0ff6d18, pkts_n=46,
err=0x7f76e0ff6a28, no_cq=0x7f76e0ff6a27)
7 mlx5_rx_burst_mprq_vec (dpdk_rxq=0x18345ef40, pkts=0x7f76e0ff6a88,
pkts_n=128)
8 rte_eth_rx_burst (nb_pkts=128, rx_pkts=0x7f76e0ff6a88,
queue_id=<optimized out>, port_id=<optimized out>)

This crash is caused by an attempt to copy previously uncompressed CQEs
into non-allocated mbufs. There is a check to make sure we only use
allocated mbufs in the rxq_burst_mprq_v() function, but it is done only
before the main processing loop. Leftovers of compressed CQEs session are
handled before that loop and may lead to the mbufs overflow as seen.

Move the check for replenished mbufs up to protect uncompressed CQEs
session leftovers from accessing non-allocated mbufs after the
mlx5_rx_mprq_replenish_bulk_mbuf() function is invoked.

Bugzilla ID: 746
Fixes: 0f20acbf5eda ("net/mlx5: implement vectorized MPRQ burst")
Cc: stable@dpdk.org
Signed-off-by: Alexander Kozyrev <akozyrev@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
2 years agonet/mlx5: fix RSS expansion for inner tunnel VLAN
Lior Margalit [Tue, 3 Aug 2021 18:13:22 +0000 (21:13 +0300)]
net/mlx5: fix RSS expansion for inner tunnel VLAN

The RSS expansion algorithm is using a graph to find the possible
expansion paths. The VLAN item in the flow pattern requires special
treatment, because it should not be added implicitly by the expansion
algorithm.  If the flow pattern ends with ETH item, the pattern will be
expanded with IPv4 and IPv6.
For example:
testpmd> flow create ... eth / end actions rss / end
ETH END
ETH IPV4 END
ETH IPV6 END
If a VLAN item follows the ETH item in the flow pattern, the pattern
will be expanded with IPv4 and IPv6 following the VLAN item.
For example:
testpmd> flow create ... eth / vlan / end actions rss level 1 / end
ETH VLAN END
ETH VLAN IPV4 END
ETH VLAN IPV6 END

The case of inner tunnel VLAN item was not taken care of so the flow
pattern did not expand with IPv6 and IPv4 as expected.
Example with inner VLAN:
testpmd> flow create ... / vxlan / eth / vlan / end actions rss level 2
/ end
The current result of the expansion alg:
ETH IPV6 UDP VXLAN ETH VLAN END
The expected result of the expansion alg:
ETH IPV6 UDP VXLAN ETH VLAN END
ETH IPV6 UDP VXLAN ETH VLAN IPV4 END
ETH IPV6 UDP VXLAN ETH VLAN IPV6 END

The fix is to introduce a new flag to set on a graph expansion node
to apply the 'explicit' behavior, meaning the node is not added to
the expanded pattern, if it is not found in the flow pattern, but the
expansion alg can go deeper to its next nodes.

Fixes: c7870bfe09dc ("ethdev: move RSS expansion code to mlx5 driver")
Cc: stable@dpdk.org
Signed-off-by: Lior Margalit <lmargalit@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
2 years agonet/txgbe: fix L4 port mask in flow director
Jiawen Wu [Thu, 12 Aug 2021 02:00:20 +0000 (10:00 +0800)]
net/txgbe: fix L4 port mask in flow director

Remove bit reverse for TCP/UDP port mask, since it causes the flows with
some TCP/UDP ports to disobey the flow director rules.

Fixes: ea230dda16ad ("net/txgbe: configure flow director filter")
Cc: stable@dpdk.org
Signed-off-by: Jiawen Wu <jiawenwu@trustnetic.com>
2 years agonet/txgbe: fix reading SFP module SFF-8472 data
Jiawen Wu [Thu, 12 Aug 2021 02:00:19 +0000 (10:00 +0800)]
net/txgbe: fix reading SFP module SFF-8472 data

Fix the I2C target address selection to read SFP module's SFF-8472 data.

Fixes: 8f09fb4642fa ("net/txgbe: add module identify")
Cc: stable@dpdk.org
Signed-off-by: Jiawen Wu <jiawenwu@trustnetic.com>
2 years agonet/txgbe: fix link status when device stopped
Jiawen Wu [Thu, 12 Aug 2021 02:00:18 +0000 (10:00 +0800)]
net/txgbe: fix link status when device stopped

When device is stopped, the port status is not changed and only the Tx
laser is turned off by hardware design.

Fixes: 0c061eadec59 ("net/txgbe: add link status change")
Cc: stable@dpdk.org
Signed-off-by: Jiawen Wu <jiawenwu@trustnetic.com>
2 years agonet/nfp: rename files for consistency
Heinrich Kuhn [Thu, 29 Jul 2021 13:47:11 +0000 (15:47 +0200)]
net/nfp: rename files for consistency

Rename the nfp_net.c file to nfp_common as it now contains functions
common to VF and PF functionality. Rename the header file too to be
consistent. Also remove the "net" naming from the _ctrl and _logs files
for consistency across the PMD.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
2 years agonet/nfp: move PF functions into its own file
Heinrich Kuhn [Thu, 29 Jul 2021 13:47:10 +0000 (15:47 +0200)]
net/nfp: move PF functions into its own file

Similar to the last commit, this changeset moves all the PF specific
functions to a new file called nfp_ethdev.c.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
2 years agonet/nfp: move VF functions into its own file
Heinrich Kuhn [Thu, 29 Jul 2021 13:47:09 +0000 (15:47 +0200)]
net/nfp: move VF functions into its own file

Move any ethdev functionality specific to VF devices into a new file
called nfp_ethdev_vf.c.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
2 years agonet/nfp: move common function prototypes
Heinrich Kuhn [Thu, 29 Jul 2021 13:47:08 +0000 (15:47 +0200)]
net/nfp: move common function prototypes

The majority of "ethdev" type functions are used for both PF devices and
VF devices. Prototype these functions in the nfp_net_pmd header file in
preparation of splitting PF and VF specific functions.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
2 years agonet/nfp: move CPP bridge to separate file
Heinrich Kuhn [Thu, 29 Jul 2021 13:47:07 +0000 (15:47 +0200)]
net/nfp: move CPP bridge to separate file

This commit moves the CPP bridge logic to a separate file. A new
corresponding header file is also created.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
2 years agonet/nfp: move datapath functions to their own file
Heinrich Kuhn [Thu, 29 Jul 2021 13:47:06 +0000 (15:47 +0200)]
net/nfp: move datapath functions to their own file

Create a new rxtx file and move the Rx/Tx functions to this file. This
commit will also move the needed shared functions to the nfp_net_pmd.h
file as needed.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
2 years agonet/nfp: split datapath structs into separate file
Heinrich Kuhn [Thu, 29 Jul 2021 13:47:05 +0000 (15:47 +0200)]
net/nfp: split datapath structs into separate file

This change splits out the rx/tx specific structs and defines from the
main nfp_net_pmd header file and into their own header file.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
2 years agonet/hns3: support set link up/down for PF
Huisong Li [Mon, 26 Jul 2021 10:59:40 +0000 (18:59 +0800)]
net/hns3: support set link up/down for PF

This patch adds set link up/down feature. RxTx datapath and link status
will be disabled when dev_set_link_down() is called, and can be enabled by
dev_start() or dev_set_link_up().

Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
2 years agonet/hns3: add Tx start/stop multi-process handling
Huisong Li [Mon, 26 Jul 2021 10:59:39 +0000 (18:59 +0800)]
net/hns3: add Tx start/stop multi-process handling

Currently, hns3 PMD has supported start/stop RxTx datapath request message
between the primary and secondary processes. However, there are some cases
only to start/stop Tx datapath. This patch adds start/stop Tx datapath
request for MP.

Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
2 years agonet/nfp: remove compile time log
Ferruh Yigit [Tue, 18 May 2021 10:41:15 +0000 (11:41 +0100)]
net/nfp: remove compile time log

Logging should be converted to dynamic log.

Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Reviewed-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
2 years agonet/ice: fix bandwidth config size in memory copy
Ting Xu [Tue, 27 Jul 2021 10:55:08 +0000 (18:55 +0800)]
net/ice: fix bandwidth config size in memory copy

The memory size of bandwidth config parameters is not set correctly in
memory copy process, which leads to the wrong values. This patch fixed
the size to the correct value.

Fixes: 3a6bfc37eaf4 ("net/ice: support QoS config VF bandwidth in DCF")
Cc: stable@dpdk.org
Signed-off-by: Ting Xu <ting.xu@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
2 years agonet/ice: fix max entry number for ACL normal priority
Simei Su [Wed, 28 Jul 2021 02:24:29 +0000 (10:24 +0800)]
net/ice: fix max entry number for ACL normal priority

For ACL, there are three entry priorities: LOW, NORMAL, HIGH.
Low priority starts from the highest index, 25% of total entries;
Normal priority starts from the highest index, 50% of total entries;
High priority starts from the lowest index, 25% of total entries.

Each TCAM block has 512 entries of 40 bits. Currently, there is a
scenario in which multiple TCAM blocks are cascaded. It means the
total entries are 512. The default priority is NORMAL, so the max
entry is 256, not 512. This patch changes the max entry number for
NORMAL priority.

Fixes: 40d466fa9f76 ("net/ice: support ACL filter in DCF")
Cc: stable@dpdk.org
Signed-off-by: Simei Su <simei.su@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
2 years agonet/ice/base: increase maximum TCAM/PTG per profile
Qi Zhang [Tue, 10 Aug 2021 02:51:40 +0000 (10:51 +0800)]
net/ice/base: increase maximum TCAM/PTG per profile

For GTPoGRE protocol in AVF FDIR/RSS, the number of associated PTGs
of one Profile may exceed the defined ICE_MAX_PTG_PER_PROFILE and
ICE_MAX_TCAM_PER_PROFILE. In those cases, some PTGs may be missed,
and therefore, the related and received packets will not have hash
values. Thus, this patch updated the ICE_MAX_PTG_PER_PROFILE and
ICE_MAX_TCAM_PER_PROFILE to a larger number 64.

Signed-off-by: Junfeng Guo <junfeng.guo@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: fix typo in comment
Qi Zhang [Tue, 10 Aug 2021 02:51:39 +0000 (10:51 +0800)]
net/ice/base: fix typo in comment

Correct spelling of word data instead of date.

Fixes: 453d087ccaff ("net/ice/base: add common functions")
Cc: stable@dpdk.org
Signed-off-by: Kevin Scott <kevin.c.scott@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: rename and add setter for unicast MAC flag
Qi Zhang [Tue, 10 Aug 2021 02:51:38 +0000 (10:51 +0800)]
net/ice/base: rename and add setter for unicast MAC flag

Rename ucast_shared to umac_shared, as "umac" is a more widely
used shorthand for "unicast MAC".

Also add a helper function to set this flag. This helper is
expected to be called by core drivers.

Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: support flow director for GTPU UL/DL with QFI
Qi Zhang [Tue, 10 Aug 2021 02:51:37 +0000 (10:51 +0800)]
net/ice/base: support flow director for GTPU UL/DL with QFI

Enable Flow Director filtering for GTPU UL/DL QFI field matching.

Signed-off-by: Junfeng Guo <junfeng.guo@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: enable jumbo frame during HW init
Qi Zhang [Tue, 10 Aug 2021 02:51:36 +0000 (10:51 +0800)]
net/ice/base: enable jumbo frame during HW init

Call ice_aq_set_mac_cfg in ice_hw_init to enable jumbo frame support.

Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: support RSS for IPv4/L4 checksum
Qi Zhang [Tue, 10 Aug 2021 02:51:35 +0000 (10:51 +0800)]
net/ice/base: support RSS for IPv4/L4 checksum

The IPv4/TCP/UDP/SCTP header checksum fields are defined in this
patch and can be used as RSS input sets.

Signed-off-by: Alvin Zhang <alvinx.zhang@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: support flow director for GTPoGRE
Qi Zhang [Tue, 10 Aug 2021 02:51:34 +0000 (10:51 +0800)]
net/ice/base: support flow director for GTPoGRE

Enable Flow Director filtering for GTPoGRE inner/outer fields
matching.

Signed-off-by: Junfeng Guo <junfeng.guo@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: enable NVM update reset capabilities
Qi Zhang [Tue, 10 Aug 2021 02:51:33 +0000 (10:51 +0800)]
net/ice/base: enable NVM update reset capabilities

Add logic to parse capabilities relating to the firmware update reset
requirements. This includes both capability 0x76, which informs the
driver if the firmware can sometimes skip PCIe resets, and 0x77, which
informs the driver if the firmware might potentially restrict EMP
resets.

For capability 0x76, if the number is 1, the firmware will report the
required reset level for a given update as part of its response to the
last command sent to program the NVM bank. (Otherwise, if the firmware
does not support this capability then it will always send a 0 in the
field of the response).

For capability 0x77, if the number is 1, the firmware will report when
EMP reset is available as part of the response to the command for
switching flash banks. (Otherwise, if the firmware does not support this
capability, it will always send a 0 in the field of the response
message).

These capabilities are required to implement immediate firmware
activation. If the capabilities are set, software can read the response
data and determine what reset level is required to activate the firmware
image. If only an EMP reset is required, and if the EMP reset is not
restricted by firmware, then the driver can issue an EMP reset to
immediately activate the new firmware.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: support RSS for GTPoGRE
Qi Zhang [Tue, 10 Aug 2021 02:51:32 +0000 (10:51 +0800)]
net/ice/base: support RSS for GTPoGRE

Support RSS for GTPoGRE inner fields hash.

Signed-off-by: Junfeng Guo <junfeng.guo@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: support flow director for GTPU EH inner IPv6
Qi Zhang [Tue, 10 Aug 2021 02:51:31 +0000 (10:51 +0800)]
net/ice/base: support flow director for GTPU EH inner IPv6

Support FDIR filtering for IPV4_GTPU_EH_IPV6 with inner
IPV6/UDP/TCP fields matching.

Signed-off-by: Junfeng Guo <junfeng.guo@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: support RSS for GRE tunnel
Qi Zhang [Tue, 10 Aug 2021 02:51:30 +0000 (10:51 +0800)]
net/ice/base: support RSS for GRE tunnel

Support RSS of inner headers for GRE tunnel packet.

Signed-off-by: Wenjun Wu <wenjun1.wu@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: support flow director for GRE tunnel
Qi Zhang [Tue, 10 Aug 2021 02:51:29 +0000 (10:51 +0800)]
net/ice/base: support flow director for GRE tunnel

Support IPV4_GRE and IPV6_GRE with inner IPV4/IPV6/UDP/TCP for
FDIR.

Signed-off-by: Wenjun Wu <wenjun1.wu@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: support TC nodes PIR configuration
Qi Zhang [Tue, 10 Aug 2021 02:51:28 +0000 (10:51 +0800)]
net/ice/base: support TC nodes PIR configuration

TC nodes CIR configuration is not supported. In order to configure PIR,
the corresponding adminq command should not include the flag for CIR.
Since the TC node info has this flag by default, it is supposed to delete
this flag for TC nodes before sending the adminq command.

Signed-off-by: Ting Xu <ting.xu@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: refine MAC rule adding
Qi Zhang [Tue, 10 Aug 2021 02:51:27 +0000 (10:51 +0800)]
net/ice/base: refine MAC rule adding

Move replay_pre_init function to interface.
Add further MAC rules, despite unicast address is already on list.

Signed-off-by: Marcin Domagala <marcinx.domagala@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: add new AQ description
Qi Zhang [Tue, 10 Aug 2021 02:51:26 +0000 (10:51 +0800)]
net/ice/base: add new AQ description

Add ice_aqc_sw_gpio struct to ice_aq_desc
This change allows us to do SW_GPIO AQ cmd transactions
over ice_aq_send_cmd() interface.

Signed-off-by: Siddaraju DH <siddaraju.dh@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: implement firmware debug dump
Qi Zhang [Tue, 10 Aug 2021 02:51:25 +0000 (10:51 +0800)]
net/ice/base: implement firmware debug dump

Basic implementation of FW Debug Dump.

Signed-off-by: Marcin Domagala <marcinx.domagala@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: add E810T check function
Qi Zhang [Tue, 10 Aug 2021 02:51:24 +0000 (10:51 +0800)]
net/ice/base: add E810T check function

Add function ice_is_e810t() to be able to distinguish if hardware is
E810T based or not.

Signed-off-by: Michal Michalik <michal.michalik@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: support starting PHY in bypass mode
Qi Zhang [Tue, 10 Aug 2021 02:51:23 +0000 (10:51 +0800)]
net/ice/base: support starting PHY in bypass mode

After starting the timestamping block, hardware begins calculating
precise offsets through a process of vernier calibration. This process
measures the effective phase offset of the various internal clocks used
in the PHY.

Once hardware completes these measurements, the P_REG_TX_OV_STATUS and
P_REG_RX_OV_STATUS registers are updated to indicate that the hardware
offset measurements are done.

This process does not happen immediately, but requires that at least one
packet be sent or received in order for the offset in that direction to
be calculated.

This poses a problem in some setups, because software expects the first
packet sent to be timestamped. This most often occurs if the clock time
is set by an application during startup. This set time command triggers
a PHY restart. Because of this, the timestamping block is reset, and
timestamps are not enabled until vernier calibration is complete. Since
this process won't complete until at least one packet is sent through
the PHY, timestamps of the very first packet sent will not be obtained.

This can result in the application failing due to missing timestamps.

To avoid this, allow starting the PHY in bypass mode. This mode enables
timestamps immediately, and skips adding the precise offset measurement.
This reduces the accuracy of the timestamp slightly, but ensures that we
get a reasonable value for the first packet.

The driver can continue monitoring the P_REG_TX_OV_STATUS and
P_REG_RX_OV_STATUS registers and exit bypass mode once the total
calibration is completed. In this way, once calibration is complete, the
timestamps will have the precise offset, but we do not break
applications which expect to be able to timestamp immediately.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: clarify comments on checking PFC mode
Qi Zhang [Tue, 10 Aug 2021 02:51:22 +0000 (10:51 +0800)]
net/ice/base: clarify comments on checking PFC mode

Rework the comment around checking PFC mode to make it clear why we are
checking the mode after sending the command.

Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: implement Vernier calibration for E822
Qi Zhang [Tue, 10 Aug 2021 02:51:21 +0000 (10:51 +0800)]
net/ice/base: implement Vernier calibration for E822

Move the implementation of Vernier calibration from Linux core ice_ptp.c
into the shared ice_ptp_hw.c file.

This implementation was recently refactored in Linux, so the move should
be verbatim with the latest Linux code that we had implemented.

This includes a new constant table with pre-determined values based on
link speed, new functions to aide in reading the multi-register values
from the PHY, functions to program the PAR/PCS conversion ratios, and
the UIX conversion ratios, functions to program the total Tx and Rx
offset after vernier calibration in the hardware completes, and finally
a function to start and stop the PHY timestamping block.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: print human-friendly PHY types
Qi Zhang [Tue, 10 Aug 2021 02:51:20 +0000 (10:51 +0800)]
net/ice/base: print human-friendly PHY types

Add functions to print PHY types in human-friendly form

Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: add accessors to get/set time reference
Qi Zhang [Tue, 10 Aug 2021 02:51:19 +0000 (10:51 +0800)]
net/ice/base: add accessors to get/set time reference

The E822 device clock might come from a variety of different sources,
called TIME_REFs. The firmware reports the current TIME_REF as part of
its function capabilities, which the driver caches when it loads.

Add an accessor function to look up the current TIME_REF from the
capabilities. This reduces line length significantly and also avoids
a tight coupling to the capabilities structure.

In some cases, TIME_REF might change at run time. This can occur in the
event that the CGU registers are updated. When this happens, its
possible that the capabilities structure can be out of date until the
capabilities are re-read.

Add a setter function to update the TIME_REF when this occurs. The
driver can call this function after updating the CGU to ensure that the
TIME_REF in the capabilities structure is up to date, without needing to
re-read the entire capabilities from firmware.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: add clock initialization function
Qi Zhang [Tue, 10 Aug 2021 02:51:18 +0000 (10:51 +0800)]
net/ice/base: add clock initialization function

Before the device PTP hardware clock can be initialized, some steps must
be taken by the driver. This includes writing some registers and
initializing the PHY.

Some of these steps are distinct depending on the device type (E810 or
E822). Additionally, a future change will introduce more steps for E822
devices to program the Clock Generation Unit.

Introduce ice_ptp_init_phc as well as device-specific sub-functions for
e810 and e822 devices.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: add timestamp masks
Qi Zhang [Tue, 10 Aug 2021 02:51:17 +0000 (10:51 +0800)]
net/ice/base: add timestamp masks

Adding macros for shift and masking of the lower timestamp work in the
Rx flex descriptor. The LSB of the timestamp-low word indicates the
validity of the timestamp while the rest 7 bits contain the timestamp.

Signed-off-by: Vignesh Sridhar <vignesh.sridhar@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: change dummy packets with VLAN
Qi Zhang [Tue, 10 Aug 2021 02:51:16 +0000 (10:51 +0800)]
net/ice/base: change dummy packets with VLAN

Ethertype was traded as VLAN tpid in dummy packets with VLAN.
This led to a problem when user wanted to add filter for VLAN and
ethertype.

Change ice_vlan_hdr to reflect correct order of VLAN fields in
packets (VLAN tpid, VLAN id). Correct all dummy packets with VLAN.
Move VLAN fields before ethertype and change offsets. Leave values
from dummy packets unchanged as they fit to new VLAN layout.

Order of offsets in ice_prot_ext_tbl_entry for VLAN protocol should
reflect order of fields in ice_vlan_hdr. However, hardware doesn't
support matching on all tpid. This should be done by matching on
packet flags. There is no FV word with protocol for VLAN and offset
2. Because of that, adding vlan tpid with not zero mask will lead
to error in creating recipe.

Signed-off-by: Michal Swiatkowski <michal.swiatkowski@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: add ethertype IPv6 check for dummy packet
Qi Zhang [Tue, 10 Aug 2021 02:51:15 +0000 (10:51 +0800)]
net/ice/base: add ethertype IPv6 check for dummy packet

In order to support switch rule for ethertype filter
with ipv6 ethertype id, it has to check ethertype and
then find a proper dummy packet. There was a silent
assumption that packet is ipv4, unless src or dst ipv6
address is specified in a flow.

Signed-off-by: Grzegorz Nitka <grzegorz.nitka@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: add functions for device clock control
Qi Zhang [Tue, 10 Aug 2021 02:51:14 +0000 (10:51 +0800)]
net/ice/base: add functions for device clock control

The ice hardware supports exposing a hardware clock for high precision
timestamping. This is primarily intended for accelerating the Precision
Time Protocol.

Add several low level functions intended to be used as the basis for
enabling the device clock, and ensuring that the port timers are
synchronized properly.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/ice/base: add IEEE 1588 capability probing
Qi Zhang [Tue, 10 Aug 2021 02:51:13 +0000 (10:51 +0800)]
net/ice/base: add IEEE 1588 capability probing

Parse 1588 timesync capability during device capability probing.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Junfeng Guo <junfeng.guo@intel.com>
2 years agonet/i40e: fix clang warning on non-x86
Ruifeng Wang [Fri, 30 Jul 2021 09:32:58 +0000 (17:32 +0800)]
net/i40e: fix clang warning on non-x86

Build on aarch64 with clang-10 has warning:
i40e_rxtx.c:3228:1:
warning: unused function 'get_avx_supported' [-Wunused-function]

The function is used in x86 specific path. Moved it into ifdef
to fix build on non-x86.

Fixes: c30751afc360 ("net/i40e: fix data path selection in secondary process")
Cc: stable@dpdk.org
Signed-off-by: Ruifeng Wang <ruifeng.wang@arm.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
2 years agonet/ice: fix priority of DCF switch rule
Wenjun Wu [Mon, 2 Aug 2021 07:25:17 +0000 (15:25 +0800)]
net/ice: fix priority of DCF switch rule

This patch fixes the reversed priority of DCF switch rule. Priority 0
and 1 are supported, and priority 0 should be the highest priority.

Fixes: 2321e34c23b3 ("net/ice: support flow priority for DCF switch filter")
Cc: stable@dpdk.org
Signed-off-by: Wenjun Wu <wenjun1.wu@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
2 years agonet/i40e: reduce L1 cache misses in NEON Rx
Feifei Wang [Fri, 23 Jul 2021 03:10:49 +0000 (11:10 +0800)]
net/i40e: reduce L1 cache misses in NEON Rx

For N1 platform, packet mbuf load and descs load are hot spots to limit
the performance for "desc_to_ptype_v" and "desc_to_olflags_v" functions
in i40e rx NEON path. This is because packet mbuf and descs are evicted
from l1d-cache to l2d-cache.

To reduce l1d-cache-misses and improve the performance, change the code
order and move "desc_to_ptype_v" and "desc_to_olflags_v" functions
forward to the location, where packet mbuf and descs are just loaded.

Test Result:
dpdk:21.08-rc1
gcc-9
For n1sdp, the patch improves the performance by 1.8%.
For thunderx2, no performance changes.

Signed-off-by: Feifei Wang <feifei.wang2@arm.com>
Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
2 years agonet/i40e: increase readability in NEON Rx
Feifei Wang [Fri, 23 Jul 2021 03:10:48 +0000 (11:10 +0800)]
net/i40e: increase readability in NEON Rx

Rearrange the code in logical order for better readability and
maintenance convenience in Rx NEON path.

No performance change with this patch in arm platform.

Suggested-by: Joyce Kong <joyce.kong@arm.com>
Signed-off-by: Feifei Wang <feifei.wang2@arm.com>
Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
2 years agodrivers/net: fix vector Rx comments
Feifei Wang [Fri, 23 Jul 2021 03:10:47 +0000 (11:10 +0800)]
drivers/net: fix vector Rx comments

For the loop to process packets in Rx vector path, some notes for the
code are wrong, fix these errors.

Fixes: 7092be8437bd ("fm10k: add vector Rx")
Fixes: c3def6a8724c ("net/i40e: implement vector PMD for altivec")
Fixes: ae0eb310f253 ("net/i40e: implement vector PMD for ARM")
Fixes: 9ed94e5bb04e ("i40e: add vector Rx")
Fixes: 319c421f3890 ("net/avf: enable SSE Rx Tx")
Fixes: 1162f5a0ef31 ("net/iavf: support flexible Rx descriptor in SSE path")
Fixes: c68a52b8b38c ("net/ice: support vector SSE in Rx")
Fixes: cf4b4708a88a ("ixgbe: improve slow-path perf with vector scattered Rx")
Cc: stable@dpdk.org
Suggested-by: Ruifeng Wang <ruifeng.wang@arm.com>
Signed-off-by: Feifei Wang <feifei.wang2@arm.com>
Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
2 years agodrivers/net: fix typo in vector Rx comment
Feifei Wang [Fri, 23 Jul 2021 03:10:46 +0000 (11:10 +0800)]
drivers/net: fix typo in vector Rx comment

In Rx vec path, for extracting and recording EOP bit, comment has
redundant "count" word, removing it.

Fixes: 7092be8437bd ("fm10k: add vector Rx")
Fixes: c3def6a8724c ("net/i40e: implement vector PMD for altivec")
Fixes: ae0eb310f253 ("net/i40e: implement vector PMD for ARM")
Fixes: 9ed94e5bb04e ("i40e: add vector Rx")
Fixes: 319c421f3890 ("net/avf: enable SSE Rx Tx")
Fixes: 1162f5a0ef31 ("net/iavf: support flexible Rx descriptor in SSE path")
Fixes: c68a52b8b38c ("net/ice: support vector SSE in Rx")
Fixes: cf4b4708a88a ("ixgbe: improve slow-path perf with vector scattered Rx")
Cc: stable@dpdk.org
Signed-off-by: Feifei Wang <feifei.wang2@arm.com>
Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
2 years agobuild: fix install from any directory with Meson 0.55
Dmitry Kozlyuk [Tue, 10 Aug 2021 23:03:22 +0000 (02:03 +0300)]
build: fix install from any directory with Meson 0.55

Install command on Windows for Meson >= 0.55.0 referenced the script
by a plain string, assuming the build directory to be directly under
the source tree root.
This resulted in an error when the assumption did not hold:

    c:\python\python.exe: can't open file
    '../buildtools/symlink-drivers-solibs.py':
    [Errno 2] No such file or directory

Use files() to make a valid script path for any build directory.

Fixes: cd27047dbee1 ("build: support drivers symlink on Windows")
Cc: stable@dpdk.org
Signed-off-by: Dmitry Kozlyuk <dmitry.kozliuk@gmail.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Nick Connolly <nick.connolly@mayadata.io>
2 years agodrivers: remove warning with Meson 0.59
Jerin Jacob [Fri, 30 Jul 2021 07:35:48 +0000 (13:05 +0530)]
drivers: remove warning with Meson 0.59

Since meson 0.59.0 version, the extract_all_objects() API
need to pass explicit boolean value.

To remove the following warning[1], added explicit `true` for
extract_all_objects() use in codebase whever there is
no argument.

[1]
WARNING: extract_all_objects called without setting recursive
keyword argument. Meson currently defaults to
non-recursive to maintain backward compatibility but
the default will be changed in the future.

Signed-off-by: Jerin Jacob <jerinj@marvell.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
2 years agoexamples/performance-thread: fix build with clang 12.0.1
Jerin Jacob [Mon, 16 Aug 2021 13:19:14 +0000 (18:49 +0530)]
examples/performance-thread: fix build with clang 12.0.1

In clang 12.0.1 version, the use of pthread_yield() is deprecated,
use sched_yield() instead.

log:
    examples/performance-thread/pthread_shim/main.c:75:9: warning:
    'pthread_yield' is deprecated: pthread_yield is deprecated,
    use sched_yield instead [-Wdeprecated-declarations]

Bugzilla ID: 745
Fixes: 433ba6228f9a ("examples/performance-thread: add pthread_shim app")
Cc: stable@dpdk.org
Signed-off-by: Jerin Jacob <jerinj@marvell.com>
Tested-by: Ali Alnubani <alialnu@nvidia.com>
2 years agoversion: 21.11-rc0
Thomas Monjalon [Sun, 8 Aug 2021 19:26:58 +0000 (21:26 +0200)]
version: 21.11-rc0

Start a new release cycle with empty release notes.

The ABI version becomes 22.0.
The map files are updated to the new ABI major number (22).
The ABI exceptions are dropped and CI ABI checks are disabled because
compatibility is not preserved.

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
Acked-by: David Marchand <david.marchand@redhat.com>
2 years agoversion: 21.08.0
Thomas Monjalon [Sun, 8 Aug 2021 15:23:21 +0000 (17:23 +0200)]
version: 21.08.0

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
2 years agodoc: announce changes in IPsec xform structure
Radu Nicolau [Thu, 5 Aug 2021 10:20:55 +0000 (11:20 +0100)]
doc: announce changes in IPsec xform structure

Signed-off-by: Radu Nicolau <radu.nicolau@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Anoob Joseph <anoobj@marvell.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
2 years agodoc: announce change in IPsec SA structure
Radu Nicolau [Thu, 5 Aug 2021 10:20:54 +0000 (11:20 +0100)]
doc: announce change in IPsec SA structure

Signed-off-by: Radu Nicolau <radu.nicolau@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
2 years agodoc: announce hiding crypto session structures
Akhil Goyal [Tue, 3 Aug 2021 12:01:12 +0000 (17:31 +0530)]
doc: announce hiding crypto session structures

The structures rte_cryptodev_sym_session and
rte_cryptodev_asym_session are not used by the
application directly. The application just need
an opaque pointer which it can attach to rte_crypto_op
while enqueue.
Hence, these structures can be internal to library
hidden from the user.

Signed-off-by: Akhil Goyal <gakhil@marvell.com>
Acked-by: Anoob Joseph <anoobj@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Acked-by: Matan Azrad <matan@nvidia.com>
2 years agodoc: announce clarification of implicit filter by port
Andrew Rybchenko [Mon, 2 Aug 2021 19:57:23 +0000 (22:57 +0300)]
doc: announce clarification of implicit filter by port

Transfer flow rules may be applied to traffic entering switch from
many sources. There are flow API pattern items which allow to specify
ingress port match criteria explicitly, but it is not documented
if ethdev port used to create flow rule adds any implicit match
criteria and how it coexists with explicit ones.

These aspects should be documented and drivers and applications
which use it in a different way must be fixed.

Signed-off-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Acked-by: Ori Kam <orika@nvidia.com>
Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Acked-by: Thomas Monjalon <thomas@monjalon.net>
2 years agodoc: announce change in ethdev flow action port ID
Andrew Rybchenko [Mon, 2 Aug 2021 19:57:22 +0000 (22:57 +0300)]
doc: announce change in ethdev flow action port ID

By its very name, action PORT_ID means that packets hit an ethdev with the
given DPDK port ID. At least the current comments don't state the opposite.

However some drivers implement it in a different way and direct traffic to
the opposite end of the "wire" plugged to the given ethdev. For example in
the case of a VF representor traffic is redirected to the corresponding VF
itself rather than to the representor ethdev and OvS uses PORT_ID action
this way.

The documentation must be clarified and, likely, rte_flow_action_port_id
structure should be extended to support both meanings.

Signed-off-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Acked-by: Ori Kam <orika@nvidia.com>
Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Acked-by: Thomas Monjalon <thomas@monjalon.net>
2 years agodoc: announce moving ethdev actions to general modify
Ori Kam [Tue, 3 Aug 2021 08:57:54 +0000 (11:57 +0300)]
doc: announce moving ethdev actions to general modify

Currently there is a dedicated modify action for each
packet field that the application wants to change.
For example:
RTE_FLOW_ACTION_TYPE_SET_IPV4_DST to modify destination of IPv4.

A new action RTE_FLOW_ACTION_TYPE_MODIFY_FIELD added the ability
to use the same action to modify any field, in addition to be able to
modify the value based on different field and not just immediate value.

Signed-off-by: Ori Kam <orika@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
2 years agodoc: announce change to ethdev modify action data
Ori Kam [Tue, 3 Aug 2021 08:57:53 +0000 (11:57 +0300)]
doc: announce change to ethdev modify action data

In the current implementation,
the action rte_flow_action_modify_field is not well defined
for fields larger than 64 bits (for example IPv6 source)
In addition, the byte order is also not well defined.

Both of those issue should be fixed.

Signed-off-by: Ori Kam <orika@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Acked-by: Jerin Jacob <jerinj@marvell.com>
2 years agodoc: announce ethdev flag and field for shared queue
Xueming Li [Mon, 2 Aug 2021 13:10:55 +0000 (16:10 +0300)]
doc: announce ethdev flag and field for shared queue

To support shared Rx queue, this patch announces new offload flag
RTE_ETH_RX_OFFLOAD_SHARED_RXQ and new shared_group field to struct
rte_eth_rxconf in DPDK v21.11.

[1] mail list discussion:
https://mails.dpdk.org/archives/dev/2021-July/215575.html

Signed-off-by: Xueming Li <xuemingl@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Acked-by: Jerin Jacob <jerinj@marvell.com>
2 years agodoc: announce renaming of vhost operations struct
Maxime Coquelin [Fri, 30 Jul 2021 08:12:30 +0000 (10:12 +0200)]
doc: announce renaming of vhost operations struct

This patch announces the renaming of struct
vhost_device_ops to rte_vhost_device_ops in DPDK v21.11.

Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Acked-by: Chenbo Xia <chenbo.xia@intel.com>
Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Acked-by: Adrian Moreno <amorenoz@redhat.com>
Acked-by: Marvin Liu <yong.liu@intel.com>
2 years agodoc: announce marking vDPA driver interface as internal
Maxime Coquelin [Fri, 30 Jul 2021 08:12:29 +0000 (10:12 +0200)]
doc: announce marking vDPA driver interface as internal

This patch announces the marking of all the vDPA driver APIs
as internal.

Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Acked-by: Chenbo Xia <chenbo.xia@intel.com>
Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Acked-by: Marvin Liu <yong.liu@intel.com>
2 years agodoc: announce promoting some vhost API to stable
Chenbo Xia [Fri, 30 Jul 2021 08:19:25 +0000 (16:19 +0800)]
doc: announce promoting some vhost API to stable

This patch announces the experimental tag removal of 10 vhost APIs,
which have been experimental for more than 2 years.
All APIs could be made stable in DPDK 21.11.

Signed-off-by: Chenbo Xia <chenbo.xia@intel.com>
Acked-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Acked-by: Fan Zhang <roy.fan.zhang@intel.com>
Acked-by: Marvin Liu <yong.liu@intel.com>
2 years agodoc: announce changes in IPsec SA config option struct
Archana Muniganti [Sat, 31 Jul 2021 17:44:28 +0000 (23:14 +0530)]
doc: announce changes in IPsec SA config option struct

Propose new fields to support offloads like
- IPsec inner checksum(L3/L4)
- IPsec tunnel header verification
- TSO
- etc
in the structure ``rte_security_ipsec_sa_options``.

Signed-off-by: Archana Muniganti <marchana@marvell.com>
Signed-off-by: Tejasree Kondoj <ktejasree@marvell.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
Acked-by: Radu Nicolau <radu.nicolau@intel.com>
Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
2 years agodoc: announce changes in IPsec xform struct
Anoob Joseph [Tue, 3 Aug 2021 06:48:37 +0000 (12:18 +0530)]
doc: announce changes in IPsec xform struct

IPsec xform struct would be updated to include IPsec SA lifetime
configuration. The existing member 'esn_soft_limit' would only track
ESN. And as sequence number control is getting introduced,
'esn_soft_limit' may not indicate the number of packets processed.
Replace that with a new structure to cover all lifetime cases with
support for specifying both soft and hard lifetimes.

ESN control introduced by https://patches.dpdk.org/patch/95808/

Signed-off-by: Anoob Joseph <anoobj@marvell.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
2 years agodoc: announce changes in security session struct
Akhil Goyal [Tue, 3 Aug 2021 12:23:35 +0000 (17:53 +0530)]
doc: announce changes in security session struct

The structure rte_security_session is not directly used
by the application. The application just need an opaque
pointer to attached to the mbuf or rte_crypto_op while
enqueue. Hence, it can be hidden inside the library
and would prevent unnecessary indirection to the priv
session data in fastpath.

Signed-off-by: Akhil Goyal <gakhil@marvell.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Acked-by: Anoob Joseph <anoobj@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
2 years agodoc: announce changes in crypto raw data vector
Hemant Agrawal [Thu, 5 Aug 2021 13:55:29 +0000 (19:25 +0530)]
doc: announce changes in crypto raw data vector

The current crypto raw data vectors need to be extended to support
out of place processing. It is proposed to add additional desl_sgl
to provide details for destination sgl.
The same is also extended to support rte_security usecases, where
we need total data length to know how much additional memory space
is available in buffer other than data length so that driver/HW
can write expanded size data after encryption.

Signed-off-by: Gagandeep Singh <g.singh@nxp.com>
Signed-off-by: Hemant Agrawal <hemant.agrawal@nxp.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Thomas Monjalon <thomas@monjalon.net>
2 years agodoc: announce change in crypto adapter metadata
Shijith Thotton [Wed, 4 Aug 2021 18:14:20 +0000 (23:44 +0530)]
doc: announce change in crypto adapter metadata

In crypto adapter metadata, first 8 bytes of request info is a space
holder for response info. For better clarity, reserved field should be
removed from request info. New space for response info can be made by
changing type of event crypto metadata to structure from union.

Signed-off-by: Shijith Thotton <sthotton@marvell.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
Acked-by: Ray Kinsella <mdr@ashroe.eu>
Acked-by: Jerin Jacob <jerinj@marvell.com>
2 years agodoc: announce removal of lcore state finished
Honnappa Nagarahalli [Fri, 30 Jul 2021 19:59:12 +0000 (14:59 -0500)]
doc: announce removal of lcore state finished

Lcore state FINISHED is used by the worker thread to indicate that
it has completed the assigned task. The state is changed to
WAIT by another thread after it observes the updated state. This
additional step is redundant. After this deprecation, the worker
thread will update the state to WAIT.

Signed-off-by: Honnappa Nagarahalli <honnappa.nagarahalli@arm.com>
Reviewed-by: Ruifeng Wang <ruifeng.wang@arm.com>
Acked-by: Feifei Wang <feifei.wang2@arm.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Acked-by: Thomas Monjalon <thomas@monjalon.net>
2 years agodoc: update release notes for 21.08
John McNamara [Thu, 5 Aug 2021 21:57:13 +0000 (21:57 +0000)]
doc: update release notes for 21.08

Fix grammar, spelling and formatting of DPDK 21.08 release notes.

Signed-off-by: John McNamara <john.mcnamara@intel.com>
2 years agodoc: add known issue with mbuf segment
Thomas Monjalon [Wed, 4 Aug 2021 13:29:52 +0000 (15:29 +0200)]
doc: add known issue with mbuf segment

A bug with segmented packets has been discovered but the agreement
to apply the fix is not concluded at the time of DPDK 21.08 release.
This bug seems to be in DPDK for many years and should be fixed in 21.11.

Suggested-by: Olivier Matz <olivier.matz@6wind.com>
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Acked-by: Morten Brørup <mb@smartsharesystems.com>
2 years agodoc: announce changes to eventdev library
Pavan Nikhilesh [Mon, 2 Aug 2021 21:09:48 +0000 (02:39 +0530)]
doc: announce changes to eventdev library

Make driver layer as internal, remove unnecessary rte_ prefix for
structures and functions that are not a part of public API.
Promote experimental trace and vector APIs to stable.
Add reserved field to `rte_event_timer` structure.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>
Acked-by: Mattias Rönnblom <mattias.ronnblom@ericsson.com>
Acked-by: Jay Jayatheerthan <jay.jayatheerthan@intel.com>
Acked-by: Abhinandan Gujjar <abhinandan.gujjar@intel.com>
2 years agonet/mlx5: fix build on Windows
Gregory Etelson [Thu, 5 Aug 2021 09:55:03 +0000 (12:55 +0300)]
net/mlx5: fix build on Windows

mlx5_dev_check_sibling_config() API was updated to allow newly
spawned port locate existing sibling devices.
PMD port initialization for Windows OS was not updated
for the new API prototype:

drivers/net/mlx5/windows/mlx5_os.c:457:50: error:
too few arguments to function call, expected 3, have 2
err = mlx5_dev_check_sibling_config(priv, config);

The patch fixes mlx5_dev_check_sibling_config call for Windows OS.

Fixes: e9d420dfc2d0 ("net/mlx5: fix find sibling devices")

Signed-off-by: Gregory Etelson <getelson@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
2 years agoversion: 21.08-rc4
Thomas Monjalon [Wed, 4 Aug 2021 16:49:16 +0000 (18:49 +0200)]
version: 21.08-rc4

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
2 years agodoc: announce cryptodev operation struct changes
Anoob Joseph [Mon, 2 Aug 2021 11:11:24 +0000 (16:41 +0530)]
doc: announce cryptodev operation struct changes

One reserved byte in rte_crypto_op struct would be used to indicate
warnings and other information from the crypto/security operation. This
field will be used to communicate events such as soft expiry with IPsec
in lookaside mode.

Signed-off-by: Anoob Joseph <anoobj@marvell.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>
2 years agodoc: announce cryptodev driver interface as internal
Akhil Goyal [Tue, 3 Aug 2021 11:44:46 +0000 (17:14 +0530)]
doc: announce cryptodev driver interface as internal

The APIs which are internal to PMD and cryptodev library
can be marked as internal so that ABI checking do not
shout for changes in interfaces which are internal to DPDK.

Signed-off-by: Akhil Goyal <gakhil@marvell.com>
Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Acked-by: Matan Azrad <matan@nvidia.com>
Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>
2 years agonet: announce changes in IPv4 header access
Gregory Etelson [Mon, 2 Aug 2021 10:38:34 +0000 (13:38 +0300)]
net: announce changes in IPv4 header access

Announce changes to add 2 unions.
The first union will provide integral and bits access to version and IHL.
The second union will provide integral and bits access to fragment flags
and offset.

Signed-off-by: Gregory Etelson <getelson@nvidia.com>
Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
2 years agodoc: announce hiding interrupt handle structure
Harman Kalra [Mon, 2 Aug 2021 16:03:52 +0000 (21:33 +0530)]
doc: announce hiding interrupt handle structure

Moving struct rte_intr_handle as an internal structure to
avoid any ABI breakages in future. Since this structure defines
some static arrays and changing respective macros breaks the ABI.
Eg:
Currently RTE_MAX_RXTX_INTR_VEC_ID imposes a limit of maximum 512
MSI-X interrupts that can be defined for a PCI device, while PCI
specification allows maximum 2048 MSI-X interrupts that can be used.
If some PCI device requires more than 512 vectors, either change the
RTE_MAX_RXTX_INTR_VEC_ID limit or dynamically allocate based on
PCI device MSI-X size on probe time. Either way its an ABI breakage.

Discussion thread:
https://mails.dpdk.org/archives/dev/2021-March/202959.html

Signed-off-by: Harman Kalra <hkalra@marvell.com>
Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Acked-by: Chenbo Xia <chenbo.xia@intel.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
2 years agodoc: announce renaming of mbuf offload flags
Olivier Matz [Fri, 30 Jul 2021 15:57:01 +0000 (17:57 +0200)]
doc: announce renaming of mbuf offload flags

The mbuf offload flags do not match the DPDK namespace (they are
not prefixed by RTE_). Announce their rename in 21.11, and the
removal of the old names in 22.11.

A draft coccinelle script is provided to anticipate what the
renaming will be.

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Acked-by: Thomas Monjalon <thomas@monjalon.net>
2 years agodoc: add tested platforms with Mellanox NICs
Raslan Darawsheh [Wed, 4 Aug 2021 12:54:47 +0000 (15:54 +0300)]
doc: add tested platforms with Mellanox NICs

Add tested platforms with Mellanox NICs to the 21.08 release notes.

Signed-off-by: Raslan Darawsheh <rasland@nvidia.com>
2 years agodoc: add tested Intel platforms with Intel NICs
Yan Xia [Wed, 4 Aug 2021 14:20:24 +0000 (14:20 +0000)]
doc: add tested Intel platforms with Intel NICs

Add tested Intel platforms with Intel NICs to v21.08 release note.

Signed-off-by: Yan Xia <yanx.xia@intel.com>
2 years agopipeline: fix table statistics
Churchill Khangar [Mon, 2 Aug 2021 05:32:12 +0000 (11:02 +0530)]
pipeline: fix table statistics

This patch fixes the memcpy function call which was incorrect and led
to memory corruption for tables with more that just a few actions.

Fixes: 742b0a57f50e4 ("pipeline: add table statistics to SWX")
Cc: stable@dpdk.org
Signed-off-by: Churchill Khangar <churchill.khangar@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
2 years agodoc: add policy for promotion of experimental API
Ray Kinsella [Wed, 4 Aug 2021 09:34:31 +0000 (10:34 +0100)]
doc: add policy for promotion of experimental API

Clarifying the ABI policy on the promotion of experimental APIs to stable.
We have a fair number of APIs that have been experimental for more than
2 years. This policy amendment indicates that these APIs should be
promoted or removed, or should at least form a conversation between the
maintainer and original contributor.

Signed-off-by: Ray Kinsella <mdr@ashroe.eu>
Acked-by: Tyler Retzlaff <roretzla@linux.microsoft.com>
Acked-by: Thomas Monjalon <thomas@monjalon.net>
2 years agodoc: update offload information for metering
Jiawei Wang [Tue, 3 Aug 2021 13:02:54 +0000 (16:02 +0300)]
doc: update offload information for metering

Update the Minimal SW and HW version offload support
information for ASO metering and metering hierarchy.

Signed-off-by: Jiawei Wang <jiaweiw@nvidia.com>
Acked-by: Asaf Penso <asafp@nvidia.com>
2 years agodoc: limit FW support for mlx5 regex driver
Ori Kam [Tue, 3 Aug 2021 19:07:49 +0000 (22:07 +0300)]
doc: limit FW support for mlx5 regex driver

MLX5 RegEx is only supported with FW version XX.30.1004 or lower.

Signed-off-by: Ori Kam <orika@nvidia.com>
Acked-by: Asaf Penso <asafp@nvidia.com>
2 years agoapp/testpmd: fix IPv4 checksum
Gregory Etelson [Mon, 2 Aug 2021 18:13:16 +0000 (21:13 +0300)]
app/testpmd: fix IPv4 checksum

UDP protocol reserves 0 checksum value for special purposes.
Other protocols, like IPv4, TCP and SCTP must calculate checksum value
in software or offload checksum calculation to hardware.

If IPv4 TX checksum offload was off and header checksum was set to 0,
testpmd csum engine did not calculate checksum value for IPv4, TCP and
SCTP.

The patch always calculates IPv4, TCP and SCTP TX checksums if it is
not offloaded.

Bugzilla ID: 768
Fixes: b2a9e4a855d0 ("app/testpmd: fix Tx checksum calculation for tunnel")
Cc: stable@dpdk.org
Signed-off-by: Gregory Etelson <getelson@nvidia.com>
Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
2 years agobus: clarify log for non-NUMA-aware devices
Dmitry Kozlyuk [Wed, 4 Aug 2021 08:03:01 +0000 (11:03 +0300)]
bus: clarify log for non-NUMA-aware devices

PCI, vmbus, and auxiliary drivers printed a warning
when NUMA node had been reported as (-1) or not reported by OS:

    EAL:   Invalid NUMA socket, default to 0

This message and its level might confuse users because the configuration
is valid and nothing happens that requires attention or intervention.
It was also printed without the device identification and with an indent
(PCI only), which is confusing unless DEBUG logging is on to print
the header message with the device name.

Reduce level to INFO, reword the message, and suppress it when there is
only one NUMA node because NUMA awareness does not matter in this case.
Also, remove the indent for PCI.

Fixes: f0e0e86aa35d ("pci: move NUMA node check from scan to probe")
Fixes: 831dba47bd36 ("bus/vmbus: add Hyper-V virtual bus support")
Fixes: 1afce3086cf4 ("bus/auxiliary: introduce auxiliary bus")
Cc: stable@dpdk.org
Signed-off-by: Dmitry Kozlyuk <dkozlyuk@nvidia.com>
Reviewed-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
Reviewed-by: Xueming Li <xuemingl@nvidia.com>
Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
2 years agonet/mlx5: fix find sibling devices
Gregory Etelson [Tue, 3 Aug 2021 15:06:58 +0000 (18:06 +0300)]
net/mlx5: fix find sibling devices

The routine mlx5_eth_find_next() and related iterating macro
MLX5_ETH_FOREACH_DEV is used to iterate through sibling devices (all
representors share the same configuration and switching domain) on top
of specified root device.

The root device parameter was specified as NULL, and it caused
missing siblings in iteration during representor device probing,
causing:

1. allocating new domain_id for the device being probed.
2. discrepancy in representor configurations and potential overall
   driver malfunctions.

Fixes: 56bb3c84e982 ("net/mlx5: reduce PCI dependency")

Signed-off-by: Gregory Etelson <getelson@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
2 years agonet/mlx5: fix domains detection in meter hierarchy
Shun Hao [Wed, 4 Aug 2021 07:26:47 +0000 (10:26 +0300)]
net/mlx5: fix domains detection in meter hierarchy

Meters in one hierarchy might support different domains. For
example, one meter may support ingress only, but the root meter
can support all the domains.

If the later meter in the meter hierarchy wrongly doesn't inherit
the first meter's domains, it will lead to invalid domain table
access.

Fix is when creating meter hierarchy, try to inherit the first meter
domains in the meter hierarchy.

Fixes: a3b7af90baba ("net/mlx5: validate meter action in policy")
Cc: stable@dpdk.org
Signed-off-by: Shun Hao <shunh@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>