This patch separates Tx function declarations to different header file
in preparation for removing their implementation from the source file
and as an optional preparation for Tx cleanup.
Signed-off-by: Michael Baum <michaelba@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
The mlx5_rxtx.c file contains a lot of Tx burst functions, each of those
is performance-optimized for the specific set of requested offloads.
These ones are generated on the basis of the template function and it
takes significant time to compile, just due to a large number of giant
functions generated in the same file and this compilation is not being
done in parallel with using multithreading.
Therefore we can split the mlx5_rxtx.c file into several separate files
to allow different functions to be compiled simultaneously.
In this patch, we separate Rx function declarations to different header
file in preparation for removing them from the source file and as an
optional preparation step for further consolidation of Rx burst
functions.
Signed-off-by: Michael Baum <michaelba@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
The rte_ethdev_driver.h, rte_ethdev_vdev.h and rte_ethdev_pci.h files are
for drivers only and should be a private to DPDK and not installed.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Acked-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Steven Webster <steven.webster@windriver.com>
IPv6 broadcast flow creation is unsupported in Windows.
do not fail on IPv6 broadcast flow creation on this mast
to avoid entire default rules creation failure.
Signed-off-by: Tal Shnaiderman <talshn@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
Replace Linux API usleep() and nanosleep() with rte_delay_us_sleep().
The replacement occurs in shared files compiled under different
operating systems.
Signed-off-by: Ophir Munk <ophirmu@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
In the implementation of mlx5_hairpin_unbind, a copy-paste error was
inside. If a single peer Rx port needed to be unbound, it would be
bound again by mistake.
All the hardware resources were released when stopping the device and
no mess of the configuration was introduced. But when trying to unbind
the ports again, the issue would appear.
The typo of the function call is fixed. If there is no hairpin queue
bound between two ports, the unbinding process should be considered
successful.
Fixes: 37cd4501e8 ("net/mlx5: support two ports hairpin mode")
Signed-off-by: Bing Zhao <bingz@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
The Tx queue stop API doesn't call the PMD callback when the state of
the queue is stopped.
The drivers should update the state to be stopped when the queue stop
callback is done successfully or when the port is stopped.
The drivers should update the state to be started when the queue start
callback is done successfully or when the port is started.
The driver wrongly didn't update the state as started when the port
start callback was done which kept the state as stopped.
Following call to a queue stop API was not completed by ethdev layer
because the state is already stopped.
Move the state update from the Tx queue setup to the port start
callback.
Fixes: 161d103b23 ("net/mlx5: add queue start and stop")
Cc: stable@dpdk.org
Signed-off-by: Matan Azrad <matan@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
In single port hairpin mode, after the queues are configured during
start up. The binding process will be enabled automatically in the
port start phase and the default control flow for egress will be
created.
When switching to two ports hairpin mode, the auto binding process
should be skipped if there is no TX queue with the peer RX queue on
the same device, and it should be skipped also if the queues are
configured with manual bind attribute.
If the explicit TX flow rule mode is configured or hairpin is
between two ports, the default control flows for TX queues should
not be created.
Signed-off-by: Bing Zhao <bingz@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
In real-life business, one device could be attached and detached
dynamically. The hairpin configuration of this port to/from all the
other ports should be enabled and disabled accordingly.
The RTE ethdev lib and PMD should provide this ability to get the
peer ports list in case that the application doesn't save it. It is
recommended that the size of the array to save the port IDs is as
large as the "RTE_MAX_ETHPORTS" to have the maximal capacity.
The order of the peer port IDs may be different from that during
hairpin queues set in the initialization stage. The peer port ID
could be the same as the current device port ID when the hairpin
peer ports contain itself - the single port hairpin.
The application should check the ports' status and decide if the
peer port should be bound / unbound when starting / stopping the
current device.
Signed-off-by: Bing Zhao <bingz@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
In order to support hairpin between two ports, mlx5 PMD needs to
implement the functions and provide them as the function pointers.
The bind and unbind functions are executed per port pairs. All the
hairpin queues between the two ports should have the same attributes
during queues setup. Different configurations among queue pairs from
the same ports are not supported. It is allowed that two ports only
have one direction hairpin.
In order to set up the connection between two queues, peer Rx queue
HW information must be fetched via the internal RTE API and the queue
information could be used to modify the SQ object. Then the RQ object
will be modified with the Tx queue HW information. The reverse
operation is not supported right now.
When disconnecting the queues pair, SQ and RQ object should be reset
without any peer HW information. The unbinding operation will try to
disconnect all Tx queues from the port from the Rx queues of the peer
port.
Tx explicit mode attribute will be saved and used when creating a
hairpin flow.
Signed-off-by: Bing Zhao <bingz@nvidia.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
The split feature for receiving packets was added to the mlx5
PMD, now Rx queue can receive the data to the buffers belonging
to the different pools and the memory of all the involved pool
must be registered for DMA operations in order to allow hardware
to store the data.
Signed-off-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
The functions rte_mbuf_dynfield_lookup() and rte_mbuf_dynflag_lookup()
can return an offset starting with 0 or a negative error code.
In reality the first offsets are probably reserved forever,
but for the sake of strict API compliance,
the checks which considered 0 as an error are fixed.
Fixes: efa79e68c8 ("net/mlx5: support fine grain dynamic flag")
Fixes: 3172c471b8 ("net/mlx5: prepare Tx queue structures to support timestamp")
Fixes: 0febfcce36 ("net/mlx5: prepare Tx to support scheduling")
Cc: stable@dpdk.org
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
Acked-by: David Marchand <david.marchand@redhat.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
Change eth_dev_stop_t return value from void to int.
Make eth_dev_stop_t implementations across all drivers to return
negative errno values if case of error conditions.
Signed-off-by: Ivan Ilchenko <ivan.ilchenko@oktetlabs.ru>
Signed-off-by: Andrew Rybchenko <arybchenko@solarflare.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
Move Tx object similar resources allocations and debug logs from DevX
and Verbs modules to a shared location.
Signed-off-by: Michael Baum <michaelba@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
As an arrangement to Windows OS support, the Verbs operations should be
separated to another file.
By this way, the build can easily cut the unsupported Verbs APIs from
the compilation process.
Define operation structure and DevX module in addition to the existing
Linux Verbs module.
Separate Tx object creation into the Verbs/DevX modules and update the
operation structure according to the OS support and the user
configuration.
Signed-off-by: Michael Baum <michaelba@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
Separate Rx queue drop creation into both Verbs and DevX modules.
Signed-off-by: Michael Baum <michaelba@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
Move Rx object similar resources allocations and debug logs from DevX
and Verbs modules to a shared location.
Signed-off-by: Michael Baum <michaelba@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
As an arrangement to Windows OS support, the Verbs operations should be
separated to another file.
By this way, the build can easily cut the unsupported Verbs APIs from
the compilation process.
Define operation structure and DevX module in addition to the existing
linux Verbs module.
Separate Rx object creation into the Verbs/DevX modules and update the
operation structure according to the OS support and the user
configuration.
Signed-off-by: Michael Baum <michaelba@nvidia.com>
Acked-by: Matan Azrad <matan@nvidia.com>
1. The shared data communication between the primary and the secondary
processes is implemented using Linux API. Move the Linux API code under
linux directory (file linux/mlx5_os.c).
2. File net/mlx5/mlx5_mp.c handles requests to the primary and secondary
processes (e.g. start_rxtx, stop_rxtx). It is Linux based so it is moved
under linux (new file linux/mlx5_mp_os.c).
Signed-off-by: Ophir Munk <ophirmu@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>
The ConnectX-6DX supports the timestamps in various formats,
the new realtime format is introduced - the upper 32-bit word
of timestamp contains the UTC seconds and the lower 32-bit word
contains the nanoseconds. This patch detects what format is
configured in the NIC and performs the conversion accordingly.
Signed-off-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>
The fields to support send scheduling on dynamic timestamp
field are introduced and initialized on device start.
Signed-off-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>
To provide the packet send schedule on mbuf timestamp the Tx
queue must be attached to the same UAR as Clock Queue is.
UAR is special hardware related resource mapped to the host
memory and provides doorbell registers, the assigning UAR
to the queue being created is provided via DevX API only.
Signed-off-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>
This patch creates the special completion queue providing
reference completions to schedule packet send from
other transmitting queues.
Signed-off-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>
The new devarg will control the steering of the lacp traffic.
When setting dv_lacp_by_user = 0 the lacp traffic will be
steered to kernel and managed there.
When setting dv_lacp_by_user = 1 the lacp traffic will
not be steered and the user will need to manage it.
Signed-off-by: Shiri Kuzin <shirik@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>
mlx5 statistics are calculated by several methods:
1. In software when packets go through datapath.
2. Calling ioctl with ETHTOOL command (Linux specific).
3. Reading counters from SYSFS device path (Linux specific).
The Linux related functions are moved to file linux/mlx5_os.c.
Signed-off-by: Ophir Munk <ophirmu@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>
Currently, the DevX counter query works asynchronously with Devx
interrupt handler return the query result. When port closes, the
interrupt handler will be uninstalled and the Devx comp obj will
also be destroyed. Meanwhile the query is still not cancelled.
In this case, counter query may use the invalid Devx comp which
has been destroyed, and query failure with invalid FD will be
reported.
Adjust the shared interrupt install and uninstall timing to make
the counter asynchronous query stop before interrupt uninstall.
Fixes: f15db67df0 ("net/mlx5: accelerate DV flow counter query")
Cc: stable@dpdk.org
Signed-off-by: Suanming Mou <suanmingm@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>
Recent patch exposed a minor performance issue,
so it is reverted.
Fixes: d237d22fbe ("net/mlx5: prefer DevX API to create Rx objects")
Signed-off-by: Dekel Peled <dekelp@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>
Currently, DevX API is used to create Rx objects (RQ, RQT, TIR) only
if LRO or hairpin features are enabled on this RQ.
This patch uses DevX API by default, if DevX is supported and can be
used. Otherwise, Verbs API is used.
Signed-off-by: Dekel Peled <dekelp@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>
Using a global mbuf dynamic field for metadata incurs some
performance penalty on a datapath. Store this information in
the Rx queue descriptor for a better cache locality.
Fixes: a18ac61133 ("net/mlx5: add metadata support to Rx datapath")
Cc: stable@dpdk.org
Signed-off-by: Alexander Kozyrev <akozyrev@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Refactor common memory btree and cache management to common driver.
Replace some input parameters of MR APIs to more common data structure
like PD, port_id, share_cache,... so that multiple PMD drivers can
use those MR APIs.
Modify mlx5 net pmd driver to use MR management APIs from common driver.
Signed-off-by: Vu Pham <vuhuong@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Only the members of flow handle structure will be used when trying
to destroy a flow. Other members of mlx5 device flow resource will
only be used for flow creating, and they could be reused for different
flows.
So only the device flow handle structure needs to be saved for further
usage. This could be separated from the whole mlx5 device flow and
stored with a list for each rte flow.
Other members will be pre-allocated with an array, and an index will
be used to help to apply each device flow to the hardware.
The flow handle sizes of Verbs and DV mode will be different, and
some calculation could be done before allocating a verbs handle.
Then the total memory consumption will less for Verbs when there is
no inbox driver being used.
Signed-off-by: Bing Zhao <bingz@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>
When stopping a mlx5 device, all the flows inserted will be flushed
since they are with non-cached mode. And no more action will be done
for these flows in the device closing stage.
If the device restarts after stopped, no flow with non-cached mode
will be re-inserted.
The flush operation through rte interface will remain the same, and
all the flows will be flushed actively.
Signed-off-by: Bing Zhao <bingz@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>
When running mlx5_dev_start in mlx5_ethdev the function calls
mlx5_dev_configure_rss_reta in order to configure the rxq's.
Before mlx5_dev_configure_rss_reta there isn't a check whether
there are rxq's and if rxq's are 0 the function fails.
For example, this command:
/build/app/test-pmd/testpmd -n 4 -w 0000:08:00.0,rx_vec_en=0
-- --burst=64 --mbcache=512 -i --nb-cores=27 --txd=2048 --rxd=2048
--vxlan-gpe-port=6081 --mp-alloc=xbuf --rxq 0 --forward-mode=txonly
would fail.
In order to fix this issue, we should call mlx5_dev_configure_rss_reta
only if we have rxq's.
Fixes: 63bd16292c ("net/mlx5: support RSS on hairpin")
Cc: stable@dpdk.org
Reported-by: Sylvain Rodon <srn@nexatech.fr>
Signed-off-by: Shiri Kuzin <shirik@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>
The inline feature is designed to save PCI bandwidth by copying some
of the data to the wqe. This feature if enabled works for all packets.
In some cases when using external memory, the PCI bandwidth is not
relevant since the memory can be accessed by other means.
This commit introduce the ability to control the inline with mbuf
granularity.
In order to use this feature the application should register the field
name, and restart the port.
Signed-off-by: Ori Kam <orika@mellanox.com>
Signed-off-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>
There are RDMA-CORE versions which are not supported multi-table for
some Mellanox mlx5 devices.
Hence, the optimization added in commit [1] which forwards all the FDB
traffic to table 1 cannot be configured.
Make the above optimization optional:
Do not fail when either table 1 cannot be created or the jump rule
(all =>jump to table 1) is not configured successfully.
In this case, all the flows will be configured to table 0.
[1] commit b67b4ecbde ("net/mlx5: skip table zero to improve
insertion rate")
Cc: stable@dpdk.org
Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
If configuring the number of tx/rx queue with rte_eth_dev_configure
to nr_queues + hairpin_nr_queues, and setting tx/rx queues to
nr_queues with rte_eth_tx/rx_queue_setup. But not configuring the
hairpin queues via rte_eth_tx/rx_hairpin_queue_setup.
When starting the netdev, there is a crash because of NULL accessing.
Fixes: cf5516696d ("ethdev: add hairpin queue")
Cc: stable@dpdk.org
Signed-off-by: Tonghao Zhang <xiangxia.m.yue@gmail.com>
Acked-by: Ori Kam <orika@mellanox.com>
This reverts commit 304ffe576f.
It tried to optimize the amount of the default flow
rules and created it only once on top of the PF representor.
For each FDB rule, the default port ID to match on is like of the port
that triggers the flow.
Hence, the single default rule will not be matched on the VF traffic.
As a result, all the traffic of the VFs will not match to the jump flow
in the root table and will always be forwarded to the representor port.
Revert the commit.
Fixes: 304ffe576f ("net/mlx5: fix condition to create default rule")
Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Previous patch added creation of a default flow rule on port start.
Rule is created under the condition that device is in eswitch mode,
and is not a VF, to make sure rule is created only once.
In Bluefield, where PF representor is used, this condition is not
sufficient. Rule is created twice, causing loss of traffic.
This patch updates this condition, adding check that device is also
not a representor.
Fixes: b67b4ecbde ("net/mlx5: skip table zero to improve insertion rate")
Signed-off-by: Dekel Peled <dekelp@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
When using hairpin all traffic from TX hairpin queues should jump
to dedecated table where matching can be done using regesters.
Signed-off-by: Ori Kam <orika@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
When starting the port, in addition to creating the queues
we need to bind the hairpin queues.
Signed-off-by: Ori Kam <orika@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
This commit adds the support for creating Tx hairpin queues.
Hairpin queue is a queue that is created using DevX and only used
by the HW.
Signed-off-by: Ori Kam <orika@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Currently all Tx queues are created using Verbs.
This commit modify the naming so it will not include verbs,
since in next commit a new type will be introduce (hairpin)
Signed-off-by: Ori Kam <orika@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
This commit adds the support for creating Rx hairpin queues.
Hairpin queue is a queue that is created using DevX and only used
by the HW. This results in that all the data part of the RQ is not being
used.
Signed-off-by: Ori Kam <orika@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
E-switch tables one and above provide higher insertion rate
than table zero, as well as enhanced functionality.
This patch adds a mechanism to utilize these advantages, by creating
a default rule on port start, which directs all packets from e-switch
table zero to table one.
Other flow rules, requested for group n, will be created in
e-switch table n+1.
Jump action to e-switch group n will be created to group n+1.
Utility function mlx5_flow_group_to_table() is added to translate the
rte_flow group value to HW table value, and is called by PMD flow
engine on flow rule validation and creation.
Signed-off-by: Dekel Peled <dekelp@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Enabling LRO offload per queue makes sense because the user will
probably want to allocate different mempool for LRO queues - the LRO
mempool mbuf size may be bigger than non LRO mempool.
Change the LRO offload to be per queue instead of per port.
If one of the queues is with LRO enabled, all the queues will be
configured via DevX.
If RSS flows direct TCP packets to queues with different LRO enabling,
these flows will not be offloaded with LRO.
Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Implement LRO support using a single RQ object per DPDK RxQ.
Signed-off-by: Dekel Peled <dekelp@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Function mlx5_rxq_obj_new(), previously called mlx5_rxq_ibv_new(),
supports creating Rx queue objects using verbs.
This patch expands the relevant functions, to support creating
verbs or DevX Rx queue objects:
Function mlx5_rxq_obj_new() updated to create RQ object using DevX.
Function mlx5_ind_table_obj_new() updated to create RQT object using DevX.
Function mlx5_hrxq_new() updated to create TIR object using DevX.
New utility functions added to perform specific operations:
mlx5_devx_rq_new(), mlx5_devx_wq_attr_fill(),
mlx5_devx_create_rq_attr_fill().
Signed-off-by: Dekel Peled <dekelp@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
