Added support for RTE_FLOW_ITEM_TYPE_MARK. This item type
can be used to create ingress flow rules to match packets
from CPT's second pass packets.
Signed-off-by: Satheesh Paul <psatheesh@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Whether the RSS is enabled depends on RTE_ETH_MQ_RX_RSS_FLAG and packet
tuple are enabled. So the RSS switch is unnecessary.
Fixes: 5e782bc2570c ("net/hns3: fix configuring RSS hash when rules are flushed")
Fixes: fd8196838763 ("net/hns3: fix configuring device with RSS enabled")
Cc: stable@dpdk.org
Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
The 'rss_tuple_fields' in struct struct hns3_rss_conf::rss_tuple_sets is
redundant. Because the enabled RSS tuple in PMD is already managed by
the 'types' in struct hns3_rss_conf::conf. This patch removes this
redundant variable.
Fixes: c37ca66f2b27 ("net/hns3: support RSS")
Cc: stable@dpdk.org
Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
The RSS tuple isn't restored when RSS key length is invalid or setting
algo key failed. This patch fixes it.
Fixes: c37ca66f2b27 ("net/hns3: support RSS")
Cc: stable@dpdk.org
Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
Currently, hns3 PMD disable RSS by resetting redirection table when user
set rss_hf to 0 so as to all packets go to queue 0. The implementation
may cause following problems:
1) the same type packet may go to different queue on the case of
disabling all tuples and partial tuples. The problem is determined by
hardware design.
2) affect the configuration of redirection table and user experience.
For hns3 hardware, the packets with RSS disabled are always go to the
queue corresponding to first entry of the redirection table. Generally,
disable RSS should be implemented by disabling all tuples, This patch
fix the implementation.
Fixes: c37ca66f2b27 ("net/hns3: support RSS")
Cc: stable@dpdk.org
Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
Currently, the hns3 PMD may free more mbufs than free_cnt parameter,
this is an incorrect implementation. This patch fixes it.
Fixes: 0b77e8f3d364 ("net/hns3: optimize Tx performance")
Cc: stable@dpdk.org
Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
Some fields in the end of 'struct hns3_rx_queue' and
'struct hns3_tx_queue' are not accessed in the I/O path.
But these fields may be accessed in other threads, which may lead to the
problem of cache pseudo-sharing of IO threads. This patch add a
cacheline alignment to avoid it.
Fixes: 9261fd3caf1f ("net/hns3: improve IO path data cache usage")
Cc: stable@dpdk.org
Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
The MAC and queues statistics are 32-bit registers in hardware. If
hardware statistics are not obtained for a long time, these statistics
will be overflow.
So PF and VF driver have to periodically obtain and save these
statistics. Since the periodical task and the stats API are in different
threads, we introduce a statistics lock to protect the statistics.
Fixes: 8839c5e202f3 ("net/hns3: support device stats")
Cc: stable@dpdk.org
Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
Clearing imissed registers in PF hardware depends on the
'drop_stats_mode' in struct hns3_hw. The variable is initialized after
the "hns3_get_configuration". But, in current code, the clearing
operation runs before the function.
So this patch fixes this order. In addition, this patch extracts a
public function to initialize and uninitialize statistics to improve the
maintainability of these codes.
Fixes: 3e9f3042d7c8 ("net/hns3: add imissed packet stats")
Cc: stable@dpdk.org
Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
rte_pmd_tun/tap_probe() allocates pmd->intr_handle in eth_dev_tap_create()
and it should not be freed until rte_pmd_tap_remove() is called.
Inspection of tap_rx_intr_vec_set() shows that the call to
tap_tx_intr_vec_uninstall() was calling rte_intr_instance_free() but
tap_tx_intr_vec_install() can then be immediately called, and this then
uses pmd->intr_handle without it being reallocated.
Move rte_intr_instance_free() call from tap_tx_intr_vec_uninstall()
to rte_pmd_tap_remove().
Fixes: d61138d4f0e2 ("drivers: remove direct access to interrupt handle")
Cc: stable@dpdk.org
Signed-off-by: Quentin Armitage <quentin@armitage.org.uk>
Reviewed-by: David Marchand <david.marchand@redhat.com>
All slaves will be stopped and removed when closing a bonded port.
But the while loop can not end if both rte_eth_dev_stop and
rte_eth_bond_slave_remove fails, runs infinitely.
This is because the skipped slave port counted in both function failures
but it should be counted only one.
Fixing by not continue to process in the loop after first failure.
Fixes: fb0379bc5db3 ("net/bonding: check stop call status")
Cc: stable@dpdk.org
Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
When stopping a bonded port, all slaves should be stopped. But only
active slaves are stopped.
So fix by stopping all slave ports and later do "deactivate_slave()" for
active slaves.
Fixes: 0911d4ec0183 ("net/bonding: fix crash when stopping mode 4 port")
Cc: stable@dpdk.org
Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
Implement support for the power management API by implementing a
'get_monitor_addr' function that will return an address of an RX ring's
status bit.
Signed-off-by: Kevin Liu <kevinx.liu@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
Support configuration of unicast and multicast promisc on dcf.
Signed-off-by: Alvin Zhang <alvinx.zhang@intel.com>
Signed-off-by: Kevin Liu <kevinx.liu@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
Add API "dev_supported_ptypes_get" to DCF, that DCF pmd can get
ptypes through the new API.
Signed-off-by: Jie Wang <jie1x.wang@intel.com>
Signed-off-by: Kevin Liu <kevinx.liu@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
add API "mtu_set" to dcf, and it can configure the port mtu through
cmdline.
Signed-off-by: Jie Wang <jie1x.wang@intel.com>
Signed-off-by: Kevin Liu <kevinx.liu@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
Add support for ops rte_eth_tx_done_cleanup in dcf
Signed-off-by: Robin Zhang <robinx.zhang@intel.com>
Signed-off-by: Kevin Liu <kevinx.liu@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
RSS HASH should be updated and queried by application,
Add related ops ('.rss_hash_update', '.rss_hash_conf_get') for DCF.
Because DCF doesn't support configure RSS HASH, only HASH key can be
updated within ops '.rss_hash_update'.
Signed-off-by: Steve Yang <stevex.yang@intel.com>
Signed-off-by: Kevin Liu <kevinx.liu@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
RSS RETA should be updated and queried by application,
Add related ops ('.reta_update', '.reta_query') for DCF.
Signed-off-by: Steve Yang <stevex.yang@intel.com>
Signed-off-by: Kevin Liu <kevinx.liu@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
In this patch, We use CPU ticks instead of HW register
to determine whether low 32 bits timestamp has turned
over. It can avoid requesting register value frequently
and improve receiving performance.
Signed-off-by: Wenjun Wu <wenjun1.wu@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
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>
Remove the limitation that max queue pair number must be 2^n.
With this patch, even on a 8 ports device, the max queue pair
number increased from 128 to 254.
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Wenjun Wu <wenjun1.wu@intel.com>
The 'max_rx_pktlen' value was previously read from hardware, which was
set by the running firmware. This caused confusion due to different
meanings of 'MAX_MTU'. This patch updates the 'max_rx_pktlen' to the
maximum value that the NFP NIC can support. The 'max_mtu' value that is
read from hardware, is assigned to the 'dev_info->max_mtu' variable.
If more layer 2 metadata must be used, the firmware can be updated to
report a smaller 'max_mtu' value.
The constant defined for NFP_FRAME_SIZE_MAX is derived for the maximum
supported buffer size of 10240, minus 136 bytes that is reserved by the
hardware and another 56 bytes reserved for expansion in firmware. This
results in a usable maximum packet length of 10048 bytes.
Signed-off-by: Walter Heymans <walter.heymans@corigine.com>
Signed-off-by: Niklas Söderlund <niklas.soderlund@corigine.com>
Reviewed-by: Louis Peens <louis.peens@corigine.com>
Reviewed-by: Chaoyong He <chaoyong.he@corigine.com>
Reviewed-by: Richard Donkin <richard.donkin@corigine.com>
Move message handler description and callbacks into a single array and
remove unneeded VHOST_USER_MAX and VHOST_SLAVE_MAX enums.
Signed-off-by: David Marchand <david.marchand@redhat.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
By default, TSO feature should be disabled because it requires
application's support to be functional as mentioned in the
documentation.
However, if "tso" devarg was not specified, the feature did
not get disabled.
This patch fixes this issue, so that TSO is disabled, even if
"tso=0" is not passed as devarg.
Fixes: e289400669d5 ("net/vhost: support TSO disabling")
Cc: stable@dpdk.org
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Reviewed-by: David Marchand <david.marchand@redhat.com>
Those optimisations were only enabled with make builds, fix the meson
part.
Fixes: 77d66da83834 ("net/virtio: add vectorized packed ring Rx")
Cc: stable@dpdk.org
Signed-off-by: David Marchand <david.marchand@redhat.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
This patch fixes heap-use-after-free reported by ASan.
It is possible for the rte_vhost_dequeue_burst() to access the vq
is freed when numa_realloc() gets called in the device running state.
The control plane will set the vq->access_lock to protected the vq
from the data plane. Unfortunately the lock will fail at the moment
the vq is freed, allowing the rte_vhost_dequeue_burst() to access
the fields of the vq, which will trigger a heap-use-after-free error.
In the case of multiple queues, the vhost pmd can access other queues
that are not ready when the first queue is ready, which makes no sense
and also allows numa_realloc() and rte_vhost_dequeue_burst() access to
vq to happen at the same time. By controlling vq->allow_queuing we can make
the pmd access only the queues that are ready.
Fixes: 1ce3c7fe149 ("net/vhost: emulate device start/stop behavior")
Signed-off-by: Yuan Wang <yuanx.wang@intel.com>
Tested-by: Wei Ling <weix.ling@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
In patch [1], NAPI has been supported in kernel tun driver to accelerate
packet processing received from vhost-net. This will greatly improve the
throughput of the tap device in the vhost-net backend.
[1]: https://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next.git/commit/?id=fb3f903769e8
Signed-off-by: Harold Huang <baymaxhuang@gmail.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Match the closest supported Rx payload buffer size with the mempool
data size and program it for the Rx queue. This removes unnecessary
need for handling additional padding, packing, and alignment, when
posting Rx buffers to hardware.
Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
When trying to coalesce mbufs with chain on Tx side, it is possible
to get stuck during queue wrap around. When coalescing this mbuf
chain fails, the Tx path returns EBUSY and when the same packet
is retried again, it couldn't get coalesced again, and the loop
repeats. Fix by pushing the packet through the normal Tx path.
Also use FW_ETH_TX_PKTS_WR to handle mbufs with chain for FW
to optimize.
Fixes: 6c2809628cd5 ("net/cxgbe: improve latency for slow traffic")
Cc: stable@dpdk.org
Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Fill the correct DPDK ethdev port_id, instead of local adapter
physical port_id in mbufs allocated for Rx.
Fixes: 78fc1a716ae8 ("cxgbe: improve Rx performance")
Cc: stable@dpdk.org
Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Rx/Tx queue info dump and pvid info dump is both supported
in PF and VF. This patch fixed it.
Fixes: 1a03c659cb9d ("net/hns3: dump device basic info")
Cc: stable@dpdk.org
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
Shouldn't dump TM info when TM is not supported by the NIC.
This patch fixed it.
Fixes: e4cfe6bb9114 ("net/hns3: dump TM configuration info")
Cc: stable@dpdk.org
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
This patch refactors queue info dump.
Fixes: 6038c8a3f63c ("net/hns3: dump queue info")
Cc: stable@dpdk.org
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
This patch refactors adapter state dump.
Fixes: 1a03c659cb9d ("net/hns3: dump device basic info")
Cc: stable@dpdk.org
Signed-off-by: Min Hu (Connor) <humin29@huawei.com>
When creating a bonding device, if the slave device's
RSS key length = standard_rss_key length + extended_hash_key length,
then bonding device will be same as slave,
in function bond_ethdev_configure(), the default_rss_key length is 40,
it is not matched, so it should calculate a new key for bonding device
if the default key could not be used.
Fixes: 6b1a001ec546 ("net/bonding: fix RSS key length")
Cc: stable@dpdk.org
Signed-off-by: Ke Zhang <ke1x.zhang@intel.com>
Acked-by: Min Hu (Connor) <humin29@huawei.com>
Looking at this driver history, there was never a need for including
execinfo.h.
Fixes: bdbf45c24b07 ("net/nfp: fix build with musl libc")
Cc: stable@dpdk.org
Signed-off-by: David Marchand <david.marchand@redhat.com>
Reviewed-by: Niklas Söderlund <niklas.soderlund@corigine.com>
This patch fixes two issues with hot removing/adding a VF PCI device:
1. The original device argument is lost when it's hot added
2. If there are multiple VFs hot adding at the same time, some of the
VFs may not get added successfully because only one single VF status
is stored in the netvsc.
Fix these by storing the original device arguments and maintain a list
of hot add contexts to deal with multiple VF devices.
Fixes: a2a23a794b ("net/netvsc: support VF device hot add/remove")
Cc: stable@dpdk.org
Signed-off-by: Long Li <longli@microsoft.com>
Driver is creating a fle pool with a fixed number of
buffers for all queue pairs of a DPSECI object.
These fle buffers are equivalent to the number of descriptors.
In this patch, creating the fle pool for each queue pair
so that user can control the number of descriptors of a
queue pair using API rte_cryptodev_queue_pair_setup().
Signed-off-by: Gagandeep Singh <g.singh@nxp.com>
Acked-by: Akhil Goyal <gakhil@marvell.com>
The driver wrongly set the LRO configurations to the TIR of the DevX
drop queue even when LRO is not supported.
Actually, the LRO configuration is not relevant to the drop queue at
all.
This causes failure in the initialization of the device, which doesn't
support LRO where the drop queue is created.
Probably, the drop queue creation by DevX missed the fact that LRO is
set by default in the TIR creation function and didn't unset it in the
drop queue case like other cases that unset LRO.
Move the default LRO configuration to unset it and set it only in the
case of all the TIR queues configured with LRO.
Fixes: bc5bee028ebc ("net/mlx5: create drop queue using DevX")
Cc: stable@dpdk.org
Signed-off-by: Michael Baum <michaelba@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
