Yunjian Wang [Wed, 22 Sep 2021 13:30:04 +0000 (21:30 +0800)]
net/ixgbe: fix memzone leak on queue re-configure
Normally when closing the device the queue memzone should be
freed. But the memzone will be not freed, when device setup
ops like:
rte_eth_bond_slave_remove
-->__eth_bond_slave_remove_lock_free
---->slave_remove
------>rte_eth_dev_internal_reset
-------->rte_eth_dev_rx_queue_config
---------->eth_dev_rx_queue_config
------------>ixgbe_dev_rx_queue_release
rte_eth_dev_close
-->ixgbe_dev_close
---->ixgbe_dev_free_queues
------>ixgbe_dev_rx_queue_release
(not been called due to nb_rx_queues and nb_tx_queues are 0)
And when queue number is changed to small size, the BIG memzone
queue index will be lost. This will lead to a memory leak. So we
should release the memzone when releasing queues.
Fixes: 460d1679586e ("drivers/net: delete HW rings while freeing queues") Cc: stable@dpdk.org Signed-off-by: Yunjian Wang <wangyunjian@huawei.com> Acked-by: Haiyue Wang <haiyue.wang@intel.com>
Yunjian Wang [Wed, 22 Sep 2021 13:29:36 +0000 (21:29 +0800)]
net/i40e: fix memzone leak on queue re-configure
Normally when closing the device the queue memzone should be
freed. But the memzone will be not freed, when device setup
ops like:
rte_eth_bond_slave_remove
-->__eth_bond_slave_remove_lock_free
---->slave_remove
------>rte_eth_dev_internal_reset
-------->rte_eth_dev_rx_queue_config
---------->eth_dev_rx_queue_config
------------>i40e_dev_rx_queue_release
rte_eth_dev_close
-->i40e_dev_close
---->i40e_dev_free_queues
------>i40e_dev_rx_queue_release
(not been called due to nb_rx_queues and nb_tx_queues are 0)
And when queue number is changed to small size, the BIG memzone
queue index will be lost. This will lead to a memory leak. So we
should release the memzone when releasing queues.
Fixes: 460d1679586e ("drivers/net: delete HW rings while freeing queues") Cc: stable@dpdk.org Signed-off-by: Yunjian Wang <wangyunjian@huawei.com> Acked-by: Haiyue Wang <haiyue.wang@intel.com>
Yunjian Wang [Wed, 22 Sep 2021 13:28:36 +0000 (21:28 +0800)]
net/ice: fix memzone leak on queue re-configure
Normally when closing the device the queue memzone should be
freed. But the memzone will be not freed, when device setup
ops like:
rte_eth_bond_slave_remove
-->__eth_bond_slave_remove_lock_free
---->slave_remove
------>rte_eth_dev_internal_reset
-------->rte_eth_dev_rx_queue_config
---------->eth_dev_rx_queue_config
------------>ice_rx_queue_release
rte_eth_dev_close
-->ice_dev_close
---->ice_free_queues
------>ice_rx_queue_release
(not been called due to nb_rx_queues and nb_tx_queues are 0)
And when queue number is changed to small size, the BIG memzone
queue index will be lost. This will lead to a memory leak. So we
should release the memzone when releasing queues.
Fixes: 460d1679586e ("drivers/net: delete HW rings while freeing queues") Cc: stable@dpdk.org Signed-off-by: Yunjian Wang <wangyunjian@huawei.com> Acked-by: Haiyue Wang <haiyue.wang@intel.com>
Yunjian Wang [Wed, 22 Sep 2021 13:28:20 +0000 (21:28 +0800)]
net/e1000: fix memzone leak on queue re-configure
Normally when closing the device the queue memzone should be
freed. But the memzone will be not freed, when device setup
ops like:
rte_eth_bond_slave_remove
-->__eth_bond_slave_remove_lock_free
---->slave_remove
------>rte_eth_dev_internal_reset
-------->rte_eth_dev_rx_queue_config
---------->eth_dev_rx_queue_config
------------>em_rx_queue_release
rte_eth_dev_close
-->eth_em_close
---->em_dev_free_queues
------>em_rx_queue_release
(not been called due to nb_rx_queues and nb_tx_queues are 0)
And when queue number is changed to small size, the BIG memzone
queue index will be lost. This will lead to a memory leak. So we
should release the memzone when releasing queues.
Fixes: 460d1679586e ("drivers/net: delete HW rings while freeing queues") Cc: stable@dpdk.org Signed-off-by: Yunjian Wang <wangyunjian@huawei.com> Acked-by: Haiyue Wang <haiyue.wang@intel.com>
Andrew Rybchenko [Wed, 29 Sep 2021 08:39:50 +0000 (11:39 +0300)]
ethdev: remove legacy mirroring API
A more fine-grain flow API action RTE_FLOW_ACTION_TYPE_SAMPLE should
be used instead of it.
Signed-off-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru> Acked-by: Thomas Monjalon <thomas@monjalon.net> Acked-by: Jerin Jacob <jerinj@marvell.com> Acked-by: Haiyue Wang <haiyue.wang@intel.com> Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
Heinrich Kuhn [Wed, 6 Oct 2021 06:44:55 +0000 (08:44 +0200)]
net/nfp: cancel delayed LSC work in port close
The link state change interrupt handler of the NFP PMD will delay the
actual LSC work for a short period to ensure the link is stable. If the
link of the port changes state and the port is closed immediately after
the link event then a segmentation fault will occur. This happens
because the delayed LSC work eventually triggers and this logic will try
to access private port data that had been released when the port was
closed.
Fixes: 6c53f87b3497 ("nfp: add link status interrupt") Cc: stable@dpdk.org Signed-off-by: Heinrich Kuhn <heinrich.kuhn@corigine.com> Signed-off-by: Simon Horman <simon.horman@corigine.com>
Xueming Li [Wed, 6 Oct 2021 11:18:22 +0000 (19:18 +0800)]
ethdev: change queue release callback
Currently, most ethdev callback API use queue ID as parameter, but Rx
and Tx queue release callback use queue object which is used by Rx and
Tx burst data plane callback.
To align with other eth device queue configuration callbacks:
- queue release callbacks are changed to use queue ID
- all drivers are adapted
Signed-off-by: Xueming Li <xuemingl@nvidia.com> Reviewed-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru> Acked-by: Ajit Khaparde <ajit.khaparde@broadcom.com> Acked-by: Somnath Kotur <somnath.kotur@broadcom.com> Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
Srujana Challa [Fri, 1 Oct 2021 13:40:19 +0000 (19:10 +0530)]
net/cnxk: allow zero UDP6 checksum for non-inline device
Sets IP6_UDP_OPT in NIX RX config to allow optional
UDP checksum for IPv6 in case of security offload.
Also disable drop_re when inline inbound is enabled.
Signed-off-by: Srujana Challa <schalla@marvell.com> Acked-by: Jerin Jacob <jerinj@marvell.com>
common/cnxk: support NIX inline inbound and outbound setup
Add API to support setting up NIX inline inbound and
NIX inline outbound. In case of inbound, SA base is setup
on NIX PFFUNC and in case of outbound, required number of
CPT LF's are attached to NIX PFFUNC.
Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com> Acked-by: Jerin Jacob <jerinj@marvell.com>
Ferruh Yigit [Mon, 4 Oct 2021 17:27:54 +0000 (18:27 +0100)]
net/cxgbe: remove useless address check
reported by "gcc (GCC) 12.0.0 20211003 (experimental)":
../drivers/net/cxgbe/cxgbe_ethdev.c:
In function ‘cxgbe_dev_rx_queue_setup’:
../drivers/net/cxgbe/cxgbe_ethdev.c:682:24:
error: the comparison will always evaluate as ‘true’ for the
address of ‘fl’ will never be NULL [-Werror=address]
682 | if ((&rxq->fl) != NULL)
| ^~
Ivan Ilchenko [Fri, 1 Oct 2021 09:07:21 +0000 (12:07 +0300)]
ethdev: update xstats by ID driver callbacks documentation
Update xstats by IDs callbacks documentation in accordance with
ethdev usage of these callbacks. Document valid combinations of
input arguments to make driver implementation simpler.
Signed-off-by: Ivan Ilchenko <ivan.ilchenko@oktetlabs.ru> Signed-off-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru> Reviewed-by: Andy Moreton <amoreton@xilinx.com> Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
ethdev: do not use get xstats names by IDs to obtain count
Relax requirements on get xstats names by IDs. After the patch
corresponding the driver operation is called with non-NULL ids
and xstats_names parameters only.
Signed-off-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru> Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
Tudor Cornea [Fri, 1 Oct 2021 08:35:01 +0000 (11:35 +0300)]
net/af_packet: reinsert stripped VLAN tag
The af_packet pmd driver binds to a raw socket and allows sending and
receiving of packets through the kernel.
Since commit [1], the kernel strips the vlan tags early in
__netif_receive_skb_core(), so we receive untagged packets while running
with the af_packet pmd.
Luckily for us, the skb vlan-related fields are still populated from the
stripped vlan tags, so we end up having all the information that we need
in the mbuf.
Having the pmd driver support DEV_RX_OFFLOAD_VLAN_STRIP allows the
application to control the desired vlan stripping behavior, until we
have a way to describe offloads that can't be disabled by pmd drivers.
This patch will cause a change in the default way that the af_packet pmd
treats received vlan-tagged frames. While previously, the application
was required to check the PKT_RX_VLAN_STRIPPED flag, after this patch,
the pmd will re-insert the vlan tag transparently to the user, unless
the DEV_RX_OFFLOAD_VLAN_STRIP is enabled in rxmode.offloads.
I've attempted a preliminary benchmark to understand if the change could
cause a sizable performance hit.
Setup:
Two virtual machines running on top of an ESXi hypervisor
Tx: DPDK app (running on top of vmxnet3 PMD)
Rx: af_packet (running on top of a kernel vmxnet3 interface)
Packet size :68 (packet contains a vlan tag)
Add support to fetch port and queue stats via xstats API. Also remove
queue stats from basic stats because they're now available via xstats
API for the VF.
Junxiao Shi [Tue, 28 Sep 2021 13:51:59 +0000 (13:51 +0000)]
net/memif: allocate socket hash on any NUMA socket
Previously, memif socket hash is always allocated on NUMA socket 0.
If the application is entirely running on another NUMA socket and EAL
--socket-limit prevents memory allocation on NUMA socket 0, memif
creation fails with "HASH: memory allocation failed" error.
This patch allows allocating memif socket hash on any NUMA socket.
Signed-off-by: Junxiao Shi <git@mail1.yoursunny.com> Reviewed-by: Jakub Grajciar <jgrajcia@cisco.com>
An issue exists on some HW revisions whereby if a CQ overflows
NIX may have undefined behavior, e.g. free incorrect buffers.
Implementing a workaround for this known HW issue.
Signed-off-by: Harman Kalra <hkalra@marvell.com> Acked-by: Jerin Jacob <jerinj@marvell.com>
As per an known HW issue RVUM interrupts may get dropped, If an RVUM
interrupt event occurs when PCCPF_XXX_MSIX_CAP_HDR[MSIXEN]=0 then no
interrupt is triggered, which is expected. But after MSIXEN is set to
1, subsequently if same interrupts event occurs again, still no
interrupt will be triggered.
As a workaround, all RVUM interrupt lines should be cleared between
MSIXEN=0 and MSIXEN=1.
Signed-off-by: Harman Kalra <hkalra@marvell.com> Acked-by: Jerin Jacob <jerinj@marvell.com>
Dmitry Kozlyuk [Thu, 7 Oct 2021 22:07:50 +0000 (01:07 +0300)]
doc: remove unneeded min/max fields deprecations
It was announced that `min` and `max` fields and variables in public
headers will be renamed to avoid clash with `min` and `max` macros
defined in Windows headers. However, it is unnecessary, because these
are function-like macros, which are not expanded unless the next token
is an opening brace. It cannot happen for integer fields and variables.
Remove the deprecation notices.
Fixes: c4379ee599ef ("doc: announce API changes for Windows compatibility") Signed-off-by: Dmitry Kozlyuk <dmitry.kozliuk@gmail.com>
Dmitry Kozlyuk [Thu, 7 Oct 2021 22:07:49 +0000 (01:07 +0300)]
net: rename Ethernet header fields
Definition of `rte_ether_addr` structure used a workaround allowing DPDK
and Windows SDK headers to be used in the same file, because Windows SDK
defines `s_addr` as a macro. Rename `s_addr` to `src_addr` and `d_addr`
to `dst_addr` to avoid the conflict and remove the workaround.
Deprecation notice:
https://mails.dpdk.org/archives/dev/2021-July/215270.html
William Tu [Wed, 6 Oct 2021 22:35:44 +0000 (22:35 +0000)]
net/ixgbe: build on Windows
This patch enables building the ixgbe driver for Windows.
It also enables its dependencies on security and cryptodev.
I tested on AWS using ixgbe VF device, using dpdk-testpmd.
Signed-off-by: William Tu <u9012063@gmail.com> Acked-by: Pallavi Kadam <pallavi.kadam@intel.com> Acked-by: Dmitry Kozlyuk <dmitry.kozliuk@gmail.com>
Jun Yang [Wed, 6 Oct 2021 17:01:23 +0000 (22:31 +0530)]
net/dpaa2: support Tx flow redirection action
TX redirection support by flow action RTE_FLOW_ACTION_TYPE_PHY_PORT
and RTE_FLOW_ACTION_TYPE_PORT_ID
This action is executed by HW to forward packets between ports.
If the ingress packets match the rule, the packets are switched
without software involved and perf is improved as well.
Signed-off-by: Jun Yang <jun.yang@nxp.com> Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>
Hemant Agrawal [Wed, 6 Oct 2021 17:01:22 +0000 (22:31 +0530)]
bus/fslmc: update MC FW to 10.28
Updating MC firmware support APIs to be latest. It supports
improved DPDMUX (SRIOV equivalent) for traffic split between
dpnis and additional PTP APIs.
Finding with "librt" keyword would give the output with
full path of librt such as
/usr/lib/gcc/x86_64-linux-gnu/7/../../../x86_64-linux-gnu/librt.so
instead of -lrt in libdpdk.pc pkg-config file.
Assume find_library() will prepend "lib", thus remove
"lib" from "librt" keyword. The output will shows
as -lrt.
This will cause an issue when compile DPDK app with
static library as the path of librt has been hard-coded
in the libdpdk.pc file.
Fixes: e41856b515ce ("raw/ifpga/base: enhance driver reliability in multi-process") Cc: stable@dpdk.org Signed-off-by: Mohamad Noor Alim Hussin <mohamad.noor.alim.hussin@intel.com> Acked-by: Tianfei Zhang <tianfei.zhang@intel.com> Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Raja Zidane [Tue, 5 Oct 2021 12:27:33 +0000 (12:27 +0000)]
regex/mlx5: refactor HW queue objects
The mlx5 PMD for regex class uses an MMO WQE operated by the GGA engine
in BF devices.
Currently, all the MMO WQEs are managed by the SQ object.
Starting from BF3, the queue of the MMO WQEs should be connected to the
GGA engine using a new configuration, MMO, that will be supported only
in the QP object.
The FW introduced new capabilities to define whether the MMO
configuration should be configured for the GGA queue.
Replace all the GGA queue objects to QP, set MMO configuration according
to the new FW capabilities.
Signed-off-by: Raja Zidane <rzidane@nvidia.com> Acked-by: Matan Azrad <matan@nvidia.com>
Raja Zidane [Tue, 5 Oct 2021 12:27:32 +0000 (12:27 +0000)]
compress/mlx5: refactor HW queue object
The mlx5 PMD for compress class uses an MMO WQE operated by the GGA
engine in BF devices.
Currently, all the MMO WQEs are managed by the SQ object.
Starting from BF3, the queue of the MMO WQEs should be connected to the
GGA engine using a new configuration, MMO, that will be supported only
in the QP object.
The FW introduced new capabilities to define whether the MMO
configuration should be configured for the GGA queue.
Replace all the GGA queue objects to QP, set MMO configuration according
to the new FW capabilities.
Signed-off-by: Raja Zidane <rzidane@nvidia.com> Acked-by: Matan Azrad <matan@nvidia.com>
Raja Zidane [Tue, 5 Oct 2021 12:27:31 +0000 (12:27 +0000)]
common/mlx5: add MMO configuration for DevX queue pair
A new configuration MMO was added to QP Context.
If set, MMO WQEs are supported on this QP.
For DMA MMO, supported only when dma_mmo_qp==1.
For REGEXP MMO, supported only when regexp_mmo_qp==1.
For COMPRESS MMO, supported only when compress_mmo_qp==1.
For DECOMPRESS MMO, supported only when decompress_mmo_qp==1.
Add support to DevX interface to set MMO bit.
Signed-off-by: Raja Zidane <rzidane@nvidia.com> Acked-by: Matan Azrad <matan@nvidia.com>
Raja Zidane [Tue, 5 Oct 2021 12:27:30 +0000 (12:27 +0000)]
common/mlx5: update MMO HCA capabilities
New MMO HCA capabilities were added and others were renamed.
Align hca capabilities with new prm.
Add support in DevX interface for changes in HCA capabilities.
Signed-off-by: Raja Zidane <rzidane@nvidia.com> Acked-by: Matan Azrad <matan@nvidia.com>
Raja Zidane [Tue, 5 Oct 2021 12:27:29 +0000 (12:27 +0000)]
common/mlx5: share DevX queue pair operations
Currently drivers using QP (vDPA, crypto and compress, regex soon)
manage their memory, creation, modification and destruction of the QP,
in almost identical code.
Move QP memory management, creation and destruction to common.
Add common function to change QP state to RTS.
Add user_index attribute to QP creation.
It's for better code maintenance and reuse.
Signed-off-by: Raja Zidane <rzidane@nvidia.com> Acked-by: Matan Azrad <matan@nvidia.com>
David Marchand [Sat, 2 Oct 2021 16:24:30 +0000 (18:24 +0200)]
bus/vmbus: fix leak on device scan
Caught running ASAN.
The device name was leaked on scan.
rte_device name field being a const, use a local pointer and release
in error path.
Fixes: 831dba47bd36 ("bus/vmbus: add Hyper-V virtual bus support") Cc: stable@dpdk.org Signed-off-by: David Marchand <david.marchand@redhat.com> Acked-by: Long Li <longli@microsoft.com>
David Marchand [Tue, 5 Oct 2021 09:16:26 +0000 (11:16 +0200)]
sort symbol maps
Fixed with ./devtools/update-abi.sh $(cat ABI_VERSION)
Fixes: e73a7ab22422 ("net/softnic: promote manage API") Fixes: 8f532a34c4f2 ("fib: promote API to stable") Fixes: 4aeb92396b85 ("rib: promote API to stable") Signed-off-by: David Marchand <david.marchand@redhat.com> Acked-by: Ray Kinsella <mdr@ashroe.eu>
Aman Deep Singh [Tue, 5 Oct 2021 11:57:54 +0000 (17:27 +0530)]
net/ice/base: fix strict-aliasing with GCC
Code changes done for build issue as reported in Bug 817
error: dereferencing type-punned pointer will break strict-aliasing rules.
added union to avoid pointer dereferencing
The build issue has been reported with both gcc 4.8.5 (RHEL 7) and
gcc 5.4.0 (Ubuntu 16.04).
Bugzilla ID: 817 Fixes: 39925373a333 ("net/ice/base: add parser execution main loop") Signed-off-by: Aman Deep Singh <aman.deep.singh@intel.com> Tested-by: David Marchand <david.marchand@redhat.com> Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
net/cxgbe: avoid sending link params change from VF
VFs are not allowed to change physical link params when a port
module change is detected. The firmware already returns appropriate
permission error when VF tries to change physical link params. But,
make sure to avoid sending the command to firmware from VF in the
first place and prevent flooding firmware debug logs with permission
errors.
Fixes: a83041b1e998 ("net/cxgbe: rework and simplify link handling") Cc: stable@dpdk.org Signed-off-by: Nikhil Vasoya <nikhil.vasoya@chelsio.com> Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
The metadata can be set in the mbuf dynamic field and then used in
flow rules steering for egress direction. The hardware requires
network order for both the insertion of a rule and sending a packet.
Indeed, there is no strict restriction for the endianness. The order
for sending a packet and its steering rule should be consistent.
In the past, there was no endianness conversion due to the
performance reason. The flow rule converted the metadata into little
endian for hardware (if needed) and the packet hit the flow rule also
with little endian.
After the metadata was converted to big endian, the missing adaption
in the data path resulted in a flow miss of the egress packets.
Converting the metadata to big endian before posting a WQE to the
hardware solves this issue.
Fixes: b57e414b48c0 ("net/mlx5: convert meta register to big-endian") Cc: stable@dpdk.org Signed-off-by: Bing Zhao <bingz@nvidia.com> Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
In the function mlx5_alloc_shared_dr(), there are various reasons
to result in a failure and error clean up process. While in the
caller of mlx5_dev_spawn(), once there is a error occurring after
the mlx5_alloc_shared_dr(), the mlx5_os_free_shared_dr() is called
to release all the resources.
To prevent a double release, the pointers of the resources should
be checked before the releasing and set to NULL after done.
In the mlx5_free_table_hash_list(), after the releasing, the pointer
was missed to set to NULL and a double release may cause a crash.
By setting the tables pointer to NULL as done for other resources,
the double release and crash could be solved.
Maxime Coquelin [Thu, 30 Sep 2021 08:12:59 +0000 (10:12 +0200)]
net/virtio: revert forcing IOVA as VA mode for virtio-user
This patch removes the simplification in Virtio descriptors
handling, where their buffer addresses are IOVAs for Virtio
PCI devices, and VA-only for Virtio-user devices, which
added a requirement on Virtio-user that it only supported
IOVA as VA.
This change introduced a regression for applications using
Virtio-user and other physical PMDs that require IOVA as PA
because they don't use an IOMMU.
This patch reverts to the old behaviour, but needed to be
reworked because of the refactoring that happened in v21.02.
Fixes: 17043a2909bb ("net/virtio: force IOVA as VA mode for virtio-user") Cc: stable@dpdk.org Reported-by: Olivier Matz <olivier.matz@6wind.com> Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com> Tested-by: Olivier Matz <olivier.matz@6wind.com> Reviewed-by: David Marchand <david.marchand@redhat.com>
David Marchand [Tue, 28 Sep 2021 08:51:14 +0000 (10:51 +0200)]
net/virtio-user: fix init when using existing tap
When attaching to an existing mono queue tap, the virtio-user was not
reporting that the virtio device was not properly initialised which
prevented from starting the port later.
$ ip tuntap add test mode tap
$ dpdk-testpmd --vdev \
net_virtio_user0,iface=test,path=/dev/vhost-net,queues=2 -- -i
...
virtio_user_dev_init_mac(): (/dev/vhost-net) No valid MAC in devargs or
device, use random
vhost_kernel_open_tap(): TUNSETIFF failed: Invalid argument
vhost_kernel_enable_queue_pair(): fail to open tap for vhost kernel
virtio_user_start_device(): (/dev/vhost-net) Failed to start device
...
Configuring Port 0 (socket 0)
vhost_kernel_open_tap(): TUNSETIFF failed: Invalid argument
vhost_kernel_enable_queue_pair(): fail to open tap for vhost kernel
virtio_set_multiple_queues(): Multiqueue configured but send command
failed, this is too late now...
Fail to start port 0: Invalid argument
Please stop the ports first
Done
The virtio-user with vhost-kernel backend was going through a lot
of complications to initialise tap fds only when using them.
For each qp enabled for the first time, a tapfd was created via
TUNSETIFF with unneeded additional steps (see below) and then mapped to
the right qp in the vhost-net backend.
Unneeded steps (as long as it has been done once for the port):
- tap features were queried while this is a constant on a running
system,
- the device name in DPDK was updated,
- the mac address of the tap was set,
On subsequent qps state change, the vhost-net backend fd mapping was
updated and the associated queue/tapfd were disabled/enabled via
TUNSETQUEUE.
Now, this patch simplifies the whole logic by keeping all tapfds opened
and in enabled state (from the tap point of view) at all time.
Unused ioctl defines are removed.
Tap features are validated earlier to fail initialisation asap.
Tap name discovery and mac address configuration are moved when
configuring qp 0.
To support attaching to mono queue tap, the virtio-user driver now tries
to attach in multi queue first, then fallbacks to mono queue.
Finally (but this is more for consistency), VIRTIO_NET_F_MQ feature is
exposed only if the underlying tap supports multi queue.
Signed-off-by: David Marchand <david.marchand@redhat.com> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Fix the tunnel port counting logic.
Currently we are incrementing the port count without checking
the if bnxt_hwrm_tunnel_dst_port_alloc would return success or failure.
Modify the logic to increment it only if the firmware returns success.
Kalesh AP [Fri, 24 Sep 2021 05:17:53 +0000 (10:47 +0530)]
net/bnxt: improve error recovery information messages
The error recovery async event messages are often mistaken
for errors. Improved the wording to clarify the meaning of
these events.
Also, take the first step towards more inclusive language.
The references to master will be changed to primary.
For example: "bnxt_is_master_func" will be renamed to
"bnxt_is_primary_func()".
The device cleanup logic was freeing most of the ring related memory,
but was not freeing up the memzone associated with the rings.
This patch fixes the issue.
Kalesh AP [Wed, 22 Sep 2021 08:30:44 +0000 (14:00 +0530)]
net/bnxt: fix function driver register/unregister
1. Fix to use correct fields in the request structure of
HWRM_FUNC_DRV_RGTR.
2. Remove the "flags" argument to bnxt_hwrm_func_driver_unregister()
as it is not needed.
Simei Su [Tue, 28 Sep 2021 06:27:53 +0000 (14:27 +0800)]
net/ice: support IEEE 1588 PTP
Add ice support for new ethdev APIs to enable/disable and read/write/adjust
IEEE1588 PTP timestamps. Currently, only scalar path supports 1588 PTP,
vector path doesn't.
The example command for running ptpclient is as below:
./build/examples/dpdk-ptpclient -c 1 -n 3 -- -T 0 -p 0x1
Signed-off-by: Simei Su <simei.su@intel.com> Acked-by: Qi Zhang <qi.z.zhang@intel.com>
In the iavf_init_arq function, if an exception occurs in the
iavf_config_arq_regs function, and the previously applied ARQ (Admin
Receive Queue) bufs resource is released. This patch maintains the same
modification as the iavf_init_asq function to roll back resources.
In the iavf_config_rx_queues_irqs function, the memory pointed to by the
intr_handle->intr_vec and qv_map addresses is not released in the
subsequent hook branch, resulting in resource leakage.
Fixes: f593944fc988 ("net/iavf: enable IRQ mapping configuration for large VF") Cc: stable@dpdk.org Signed-off-by: Qiming Chen <chenqiming_huawei@163.com> Acked-by: Qi Zhang <qi.z.zhang@intel.com>
Simei Su [Sun, 26 Sep 2021 14:04:45 +0000 (22:04 +0800)]
net/ice: enable Rx timestamp on flex descriptor
Use the dynamic mbuf to register timestamp field and flag.
The ice has the feature to dump Rx timestamp value into dynamic
mbuf field by flex descriptor. This feature is turned on by dev
config "enable-rx-timestamp". Currently, it's only supported
under scalar path.
Signed-off-by: Simei Su <simei.su@intel.com> Acked-by: Qi Zhang <qi.z.zhang@intel.com>
Robin Zhang [Fri, 24 Sep 2021 06:22:26 +0000 (06:22 +0000)]
net/i40e: remove i40evf
The default VF driver for Intel 700 Series Ethernet Controller already
switch to iavf in DPDK 21.05. And i40evf is no need to maintain now,
so remove i40evf related code.
Signed-off-by: Robin Zhang <robinx.zhang@intel.com> Acked-by: Qi Zhang <qi.z.zhang@intel.com>
There is currently a scenario test, which will continuously obtain port
statistics, causing the CPU usage to soar, which does not meet the
demand. After positioning analysis, it is found that the VF and PF
command interaction is completed through the iavf_execute_vf_cmd
function.
After the message is sent, it needs to wait for the interrupt thread to
obtain the response from the PF. For the data, the rte_delay_ms
interface is used here to wait, but the CPU will not be released during
the waiting period of this interface, which will cause the statistics to
keep occupying the CPU. This is also the root cause of the soaring CPU.
The command interaction should belong to the control plane, and there
will not be too high requirements for performance. It is recommended to
wait for the interface iavf_msec_delay to complete without taking up the
CPU time.
git.droids-corp.org / dpdk.git / log