dpdk.git
4 years agocrypto/armv8: enable meson build
Dharmik Thakkar [Thu, 23 Jan 2020 03:45:54 +0000 (11:45 +0800)]
crypto/armv8: enable meson build

Add new meson.build file for crypto/armv8

Suggested-by: Thomas Monjalon <thomas@monjalon.net>
Signed-off-by: Dharmik Thakkar <dharmik.thakkar@arm.com>
Reviewed-by: Honnappa Nagarahalli <honnappa.nagarahalli@arm.com>
Acked-by: Akhil Goyal <akhil.goyal@nxp.com>
Tested-by: Ruifeng Wang <ruifeng.wang@arm.com>
4 years agocrypto/armv8: link to library hosted by Arm
Ruifeng Wang [Thu, 23 Jan 2020 03:45:53 +0000 (11:45 +0800)]
crypto/armv8: link to library hosted by Arm

Armv8 crypto PMD linked to armv8_crypto library created by Marvell.
Maintenance of armv8_crypto library will be discontinued.
Change Armv8 PMD to link to AArch64 crypto library hosted by Arm.

Signed-off-by: Ruifeng Wang <ruifeng.wang@arm.com>
Reviewed-by: Honnappa Nagarahalli <honnappa.nagarahalli@arm.com>
Reviewed-by: Gavin Hu <gavin.hu@arm.com>
Reviewed-by: Ola Liljedahl <ola.liljedahl@arm.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
Acked-by: Akhil Goyal <akhil.goyal@nxp.com>
4 years agotest/crypto: refactor unit tests into one combined array
Adam Dybkowski [Mon, 20 Jan 2020 13:11:46 +0000 (14:11 +0100)]
test/crypto: refactor unit tests into one combined array

This patch refactors most of unit tests to be contained in one
combined array, and run depending on the PMD capabilities instead of
providing multiple array with tests for individual PMDs.
Only a subset of unit tests was merged into one array - it combines
all tests originally meant to be run on these PMDs:
null, aesni_mb, aesni_gcm, openssl, qat, sw_snow3g, sw_kasumi, sw_zuc.

Signed-off-by: Adam Dybkowski <adamx.dybkowski@intel.com>
Acked-by: Fiona Trahe <fiona.trahe@intel.com>
Acked-by: Akhil Goyal <akhil.goyal@nxp.com>
Tested-by: Akhil Goyal <akhil.goyal@nxp.com>
4 years agotest/crypto: add capability checks
Adam Dybkowski [Mon, 20 Jan 2020 13:11:45 +0000 (14:11 +0100)]
test/crypto: add capability checks

This patch adds capability checks to many tests meant to be run
in the future on various PMDs. This way the code is prepared for
more thorough refactoring in order to create one big central
unit tests array.

Signed-off-by: Adam Dybkowski <adamx.dybkowski@intel.com>
Acked-by: Fiona Trahe <fiona.trahe@intel.com>
Acked-by: Akhil Goyal <akhil.goyal@nxp.com>
Tested-by: Akhil Goyal <akhil.goyal@nxp.com>
4 years agotest/crypto: refactor unit tests
Adam Dybkowski [Mon, 20 Jan 2020 13:11:44 +0000 (14:11 +0100)]
test/crypto: refactor unit tests

This patch gets rid of individual functions that all call
test_blockcipher_all_tests separately for every PMD and instead
provides just one set universal for all PMDs that's basing on the
driver id from the global variable gbl_driver_id.

Signed-off-by: Adam Dybkowski <adamx.dybkowski@intel.com>
Acked-by: Akhil Goyal <akhil.goyal@nxp.com>
Tested-by: Akhil Goyal <akhil.goyal@nxp.com>
4 years agocrypto/snow3g: use IPsec library
Pablo de Lara [Mon, 20 Jan 2020 11:47:56 +0000 (11:47 +0000)]
crypto/snow3g: use IPsec library

Link against Intel IPsec Multi-buffer library, which
added support for SNOW3G-UEA2 and SNOW3G-UIA2 from version v0.53,
moving from libSSO SNOW3G library.

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: Akhil Goyal <akhil.goyal@nxp.com>
4 years agocrypto/kasumi: use IPsec library
Pablo de Lara [Mon, 20 Jan 2020 11:47:55 +0000 (11:47 +0000)]
crypto/kasumi: use IPsec library

Link against Intel IPsec Multi-buffer library, which
added support for KASUMI-F8 and KASUMI-F9 from version v0.53,
moving from libSSO KASUMI library.

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: Akhil Goyal <akhil.goyal@nxp.com>
4 years agocrypto/zuc: use IPsec library
Pablo de Lara [Mon, 20 Jan 2020 11:47:54 +0000 (11:47 +0000)]
crypto/zuc: use IPsec library

Link against Intel IPsec Multi-buffer library, which
added support for ZUC-EEA3 and ZUC-EIA3 from version v0.53,
moving from libSSO ZUC library.

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: Akhil Goyal <akhil.goyal@nxp.com>
4 years agovhost: catch overflow causing mmap of size 0
Maxime Coquelin [Thu, 16 Jan 2020 10:44:27 +0000 (11:44 +0100)]
vhost: catch overflow causing mmap of size 0

This patch catches an overflow that could happen if an
invalid region size or page alignment is provided by the
guest via the VHOST_USER_SET_MEM_TABLE request.

If the sum of the size to mmap and the alignment overflows
uint64_t, then RTE_ALIGN_CEIL(mmap_size, alignment) macro
will return 0. This value was passed as is as size argument
to mmap().

While kernel handling of mmap() syscall returns an error
if size is 0, it is better to catch it earlier and provide
a meaningful error log.

Fixes: ec09c280b839 ("vhost: fix mmap not aligned with hugepage size")
Cc: stable@dpdk.org
Reported-by: Ilja Van Sprundel <ivansprundel@ioactive.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Reviewed-by: Tiwei Bie <tiwei.bie@intel.com>
4 years agovhost: fix packed virtqueue ready condition
Adrian Moreno [Sat, 25 Jan 2020 08:52:16 +0000 (09:52 +0100)]
vhost: fix packed virtqueue ready condition

Consider a virtqueue ready when, apart from the descriptor area,
both event suppression areas have been mapped.

Fixes: 2d1541e2b6b3 ("vhost: add vring address setup for packed queues")
Cc: stable@dpdk.org
Signed-off-by: Adrian Moreno <amorenoz@redhat.com>
Reviewed-by: Tiwei Bie <tiwei.bie@intel.com>
4 years agovhost/crypto: fix fetch size
Fan Zhang [Wed, 29 Jan 2020 10:24:22 +0000 (10:24 +0000)]
vhost/crypto: fix fetch size

This patch fixes the incorrect rte_vhost_crypto_fetch_requests
return value.

Coverity issue: 343401
Fixes: 3bb595ecd682 ("vhost/crypto: add request handler")
Cc: stable@dpdk.org
Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
4 years agovhost: flush shadow Tx if no more packets
Eugenio Pérez [Wed, 29 Jan 2020 19:33:10 +0000 (20:33 +0100)]
vhost: flush shadow Tx if no more packets

The current implementation of vhost_net in packed vring tries to fill
the shadow vector before send any actual changes to the guest. While
this can be beneficial for the throughput, it conflicts with some
bufferfloats methods like the linux kernel napi, that stops
transmitting packets if there are too much bytes/buffers in the
driver.

To solve it, we flush the shadow packets at the end of
virtio_dev_tx_packed if we have starved the vring, i.e. the next
buffer is not available for the device.

Since this last check can be expensive because of the atomic, we only
check it if we have not obtained the expected "count" packets. If it
happens to obtain "count" packets and there is no more available
packets the caller needs to keep call virtio_dev_tx_packed again.

Fixes: 31d6c6a5b820 ("vhost: optimize packed ring dequeue")
Cc: stable@dpdk.org
Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
4 years agovhost: do not treat empty socket message as error
Vitaliy Mysak [Thu, 30 Jan 2020 08:05:39 +0000 (09:05 +0100)]
vhost: do not treat empty socket message as error

According to recvmsg() specification, 0 is a valid
return code when client is disconnecting.
Therefore, it should not be reported as error, unless there
are other dependencies that require message to not be empty.
But there are none, since the next immediate caller of recvmsg()
reports "vhost peer closed" info (not error) when message is empty.

This patch changes return code check for recvmsg() so that
misleading error message is not printed when the code is 0.

Fixes: 8f972312b8f4 ("vhost: support vhost-user")
Cc: stable@dpdk.org
Signed-off-by: Vitaliy Mysak <vitaliy.mysak@intel.com>
Reviewed-by: Tiwei Bie <tiwei.bie@intel.com>
4 years agovhost: fix crash on port deletion
Zhike Wang [Thu, 16 Jan 2020 02:07:37 +0000 (10:07 +0800)]
vhost: fix crash on port deletion

The vhost_user_read_cb() and rte_vhost_driver_unregister()
can be called at the same time by 2 threads. Eg thread1
calls vhost_user_read_cb() and removes the vsocket from
conn_list, then thread2 calls rte_vhost_driver_unregister()
and frees the vsocket since it is NOT in the conn_list.
So thread1 will access invalid memory when trying to
reconnect.

The fix is to move the "removing of vsocket from conn_list"
to end of the vhost_user_read_cb(), then avoid the race
condition.

The core trace is:
Program terminated with signal 11, Segmentation fault.

Fixes: af1475918124 ("vhost: introduce API to start a specific driver")
Cc: stable@dpdk.org
Signed-off-by: Zhike Wang <wangzhike@jd.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
4 years agonet/virtio-user: do not reset virtqueues for split ring
Tiwei Bie [Wed, 5 Feb 2020 02:11:08 +0000 (10:11 +0800)]
net/virtio-user: do not reset virtqueues for split ring

Add missing braces to avoid resetting virtqueues unconditionally
during reconnection.

Fixes: 6ebbf4109f35 ("net/virtio-user: fix packed ring server mode")
Cc: stable@dpdk.org
Signed-off-by: Tiwei Bie <tiwei.bie@intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
4 years agonet/mlx5: fix dirty array of actions
Dekel Peled [Wed, 5 Feb 2020 06:42:02 +0000 (08:42 +0200)]
net/mlx5: fix dirty array of actions

Previous patch changed the format of struct
mlx5_flow_dv_modify_hdr_resource, to use a flexible array for
modification actions.
In __flow_dv_translate() a union was defined with item of this struct,
and an array of maximal possible size.
Array elements are filled in several functions.
In function flow_dv_convert_action_set_reg(), array element is filled
partially, while the other fields of this array element are left
uninitialized.
This may cause failure of flow_dv_modify_hdr_resource_register()
when calling driver function with the 'dirty' array.

This patch updates flow_dv_convert_action_set_reg(), setting the
selected array element fields while clearing the other fields.
Other functions that fill the same array elements are also updated
for clarity and proofing future use.

Fixes: 024e95759c16 ("net/mlx5: fix modify actions support limitation")
Cc: stable@dpdk.org
Signed-off-by: Dekel Peled <dekelp@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agonet/mlx5: fix memory regions release deadlock
Michael Baum [Tue, 4 Feb 2020 13:36:09 +0000 (15:36 +0200)]
net/mlx5: fix memory regions release deadlock

The mpx5 PMD maintains the list of devices for those the memory
operation callback routines must be invoked to keep the device MRs (MR
is the entity backing the hardware DMA transactions) consistent with the
mapped memory.
Each device context in the list is protected with dedicated lock on per
device basis, which might be taken inside the callback routine.

When device is closing the PMD frees all MRs by calling
mlx5_mr_release(), that might call rte_free() under the taken device
lock.  If this rte_free call triggers the entire memory segment freeing
it, in its turn, invokes the callback routine and attempt to take the
lock inside this one causes the deadlock.

The patch proposes the remove the device from the callback list first
and then call mlx5_mr_release() and free the remaining device MRs
explicitly.

Fixes: 0e3d0525b2f2 ("net/mlx5: fix memory event callback list")
Cc: stable@dpdk.org
Signed-off-by: Michael Baum <michaelba@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>
4 years agonet/failsafe: fix reported hash key size in device info
Raslan Darawsheh [Tue, 4 Feb 2020 15:10:51 +0000 (17:10 +0200)]
net/failsafe: fix reported hash key size in device info

Hash key size is missing from reported device info.
This fills the hash key size in device info.

Fixes: 4586be3743d4 ("net/failsafe: fix reported device info")
Cc: stable@dpdk.org
Signed-off-by: Raslan Darawsheh <rasland@mellanox.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
4 years agoapp/testpmd: fix uninitialized members when setting PFC
Wei Hu (Xavier) [Tue, 21 Jan 2020 11:44:33 +0000 (19:44 +0800)]
app/testpmd: fix uninitialized members when setting PFC

Only a part of members in the local structure variable named pfc_conf
are initialized in the function named cmd_priority_flow_ctrl_set_parsed
when typing "set pfc_ctrl..." command, and others are random values.
However, those uninitialized members may cause failure.

This patch adds clearing zero operation before calling the API named
rte_eth_dev_priority_flow_ctrl_set API with pfc_conf as the input
parameter.

Fixes: 9b53e542e9e1 ("app/testpmd: add priority flow control")
Cc: stable@dpdk.org
Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Xuan Li <lixuan47@hisilicon.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
4 years agoapp/testpmd: fix initial value when setting PFC
Wei Hu (Xavier) [Tue, 21 Jan 2020 11:44:32 +0000 (19:44 +0800)]
app/testpmd: fix initial value when setting PFC

Currently, the initial values of the local structure variable named
rx_tx_onoff_2_lfc_mode and rx_tx_onoff_2_pfc_mode are different in the
similar part of these two following functions:
cmd_link_flow_ctrl_set_parsed
cmd_priority_flow_ctrl_set_parsed
1) The code snippset in cmd_link_flow_ctrl_set_parsed function:
static enum rte_eth_fc_mode rx_tx_onoff_2_lfc_mode[2][2] = {
    {RTE_FC_NONE, RTE_FC_TX_PAUSE}, {RTE_FC_RX_PAUSE, RTE_FC_FULL}
};

if (!cmd || cmd == &cmd_link_flow_control_set_rx)
rx_fc_en = (!strcmp(res->rx_lfc_mode, "on")) ? 1 : 0;
if (!cmd || cmd == &cmd_link_flow_control_set_tx)
tx_fc_en = (!strcmp(res->tx_lfc_mode, "on")) ? 1 : 0;

fc_conf.mode = rx_tx_onoff_2_lfc_mode[rx_fc_en][tx_fc_en];
<...>
ret = rte_eth_dev_flow_ctrl_set(res->port_id, &fc_conf);
<...>
2) The code snippset in cmd_priority_flow_ctrl_set_parsed function:
static enum rte_eth_fc_mode rx_tx_onoff_2_pfc_mode[2][2] = {
    {RTE_FC_NONE, RTE_FC_RX_PAUSE}, {RTE_FC_TX_PAUSE, RTE_FC_FULL}
};

rx_fc_enable = (!strncmp(res->rx_pfc_mode, "on",2)) ? 1 : 0;
tx_fc_enable = (!strncmp(res->tx_pfc_mode, "on",2)) ? 1 : 0;
pfc_conf.fc.mode =
 rx_tx_onoff_2_pfc_mode[rx_fc_enable][tx_fc_enable];
<...>
ret = rte_eth_dev_priority_flow_ctrl_set(res->port_id, &pfc_conf);
<...>
The initial value of rx_tx_onoff_2_pfc_mode is wrong, it should be the
same as rx_tx_onoff_2_lfc_mode.

Fixes: 9b53e542e9e1 ("app/testpmd: add priority flow control")
Cc: stable@dpdk.org
Signed-off-by: Huisong Li <lihuisong@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Xuan Li <lixuan47@hisilicon.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
4 years agoapp/testpmd: fix copy of dynamic flag name
Ori Kam [Tue, 4 Feb 2020 13:39:46 +0000 (13:39 +0000)]
app/testpmd: fix copy of dynamic flag name

When working with testpmd and setting the dynflag name, we copy the
name given by the cmd to the dynflag name.

The issue is that the size of the dynflag name is smaller then the
string used by testpmd.

This commit solves this issue by checking that the length of the requested
flag name is not too long.

Coverity issue: 353610
Fixes: b57b66a97ebf ("app/testpmd: support mbuf dynamic flag")

Signed-off-by: Ori Kam <orika@mellanox.com>
Acked-by: Bernard Iremonger <bernard.iremonger@intel.com>
4 years agonet/memif: add link info
Július Milan [Tue, 4 Feb 2020 11:30:24 +0000 (12:30 +0100)]
net/memif: add link info

This information is useful or needed for user applications as t-rex.

Signed-off-by: Július Milan <jmilan.dev@gmail.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
4 years agoapp/testpmd: fix uninitialized members of MPLS
Ori Kam [Thu, 30 Jan 2020 16:59:35 +0000 (16:59 +0000)]
app/testpmd: fix uninitialized members of MPLS

Some of the members of the MPLS struct are not initialized.
This commit init the uninitialized members.

Coverity issue: 325735
Fixes: 3e77031be855 ("app/testpmd: add MPLSoGRE encapsulation")
Cc: stable@dpdk.org
Signed-off-by: Ori Kam <orika@mellanox.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
4 years agovdpa/mlx5: disable RoCE
Matan Azrad [Sun, 2 Feb 2020 16:03:53 +0000 (16:03 +0000)]
vdpa/mlx5: disable RoCE

In order to support virtio queue creation by the FW, RoCE mode
should be disabled in the device.

Do it by netlink which is like the devlink tool commands:
1. devlink dev param set pci/[pci] name enable_roce value false
   cmode driverinit
2. devlink dev reload pci/[pci]
Or by sysfs which is like:
echo 0 >  /sys/bus/pci/devices/[pci]/roce_enable

The IB device is matched again after ROCE disabling.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Acked-by: Maxime Coquelin <maxime.coquelin@redhat.com>
4 years agovdpa/mlx5: support close and config operations
Matan Azrad [Sun, 2 Feb 2020 16:03:52 +0000 (16:03 +0000)]
vdpa/mlx5: support close and config operations

Support dev_conf and dev_conf operations.
These operations allow vdpa traffic.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Acked-by: Maxime Coquelin <maxime.coquelin@redhat.com>
4 years agovdpa/mlx5: support live migration
Matan Azrad [Sun, 2 Feb 2020 16:03:51 +0000 (16:03 +0000)]
vdpa/mlx5: support live migration

Add support for live migration feature by the HW:
Create a single Mkey that maps the memory address space of the
VHOST live migration log file.
Modify VIRTIO_NET_Q object and provide vhost_log_page,
dirty_bitmap_mkey, dirty_bitmap_size, dirty_bitmap_addr
and dirty_bitmap_dump_enable.
Modify VIRTIO_NET_Q object and move state to SUSPEND.
Query VIRTIO_NET_Q and get hw_available_idx and hw_used_idx.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Acked-by: Maxime Coquelin <maxime.coquelin@redhat.com>
4 years agovdpa/mlx5: map doorbell
Matan Azrad [Sun, 2 Feb 2020 16:03:50 +0000 (16:03 +0000)]
vdpa/mlx5: map doorbell

The HW supports only 4 bytes doorbell writing detection.
The virtio device set only 2 bytes when it rings the doorbell.

Map the virtio doorbell detected by the virtio queue kickfd to the HW
VAR space when it expects to get the virtio emulation doorbell.

Use the EAL interrupt mechanism to get notification when a new event
appears in kickfd by the guest and write 4 bytes to the HW doorbell space
in the notification callback.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Acked-by: Maxime Coquelin <maxime.coquelin@redhat.com>
4 years agovdpa/mlx5: support queue state operation
Matan Azrad [Sun, 2 Feb 2020 16:03:49 +0000 (16:03 +0000)]
vdpa/mlx5: support queue state operation

Add support for set_vring_state operation.

Using DevX API the virtq state can be changed as described in PRM:
enable - move to ready state.
disable - move to suspend state.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Acked-by: Maxime Coquelin <maxime.coquelin@redhat.com>
4 years agovdpa/mlx5: add basic steering configurations
Matan Azrad [Sun, 2 Feb 2020 16:03:48 +0000 (16:03 +0000)]
vdpa/mlx5: add basic steering configurations

Add a steering object to be managed by a new file mlx5_vdpa_steer.c.

Allow promiscuous flow to scatter the device Rx packets to the virtio
queues using RSS action.

In order to allow correct RSS in L3 and L4, split the flow to 7 flows
as required by the device.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Acked-by: Maxime Coquelin <maxime.coquelin@redhat.com>
4 years agovdpa/mlx5: support stateless offloads
Matan Azrad [Sun, 2 Feb 2020 16:03:47 +0000 (16:03 +0000)]
vdpa/mlx5: support stateless offloads

Add support for the next features in virtq configuration:
VIRTIO_F_RING_PACKED,
VIRTIO_NET_F_HOST_TSO4,
VIRTIO_NET_F_HOST_TSO6,
VIRTIO_NET_F_CSUM,
VIRTIO_NET_F_GUEST_CSUM,
VIRTIO_F_VERSION_1,

These features support depends in the DevX capabilities reported by the
device.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
4 years agovdpa/mlx5: prepare virtio queues
Matan Azrad [Sun, 2 Feb 2020 16:03:46 +0000 (16:03 +0000)]
vdpa/mlx5: prepare virtio queues

The HW virtq object represents an emulated context for a VIRTIO_NET
virtqueue which was created and managed by a VIRTIO_NET driver as
defined in VIRTIO Specification.

Add support to prepare and release all the basic HW resources needed
the user virtqs emulation according to the rte_vhost configurations.

This patch prepares the basic configurations needed by DevX commands to
create a virtq.

Add new file mlx5_vdpa_virtq.c to manage virtq operations.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agovdpa/mlx5: prepare HW queues
Matan Azrad [Sun, 2 Feb 2020 16:03:45 +0000 (16:03 +0000)]
vdpa/mlx5: prepare HW queues

As an arrangement to the vitrio queues creation, a 2 QPs and CQ may be
created for the virtio queue.

The design is to trigger an event for the guest and for the vdpa driver
when a new CQE is posted by the HW after the packet transition.

This patch add the basic operations to create and destroy the above HW
objects  and to trigger the CQE events when a new CQE is posted.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agovdpa/mlx5: prepare memory regions
Matan Azrad [Sun, 2 Feb 2020 16:03:44 +0000 (16:03 +0000)]
vdpa/mlx5: prepare memory regions

In order to map the guest physical addresses used by the virtio device
guest side to the host physical addresses used by the HW as the host
side, memory regions are created.

By this way, for example, the HW can translate the addresses of the
packets posted by the guest and to take the packets from the correct
place.

The design is to work with single MR which will be configured to the
virtio queues in the HW, hence a lot of direct MRs are grouped to single
indirect MR.

Create functions to prepare and release MRs with all the related
resources that are required for it.

Create a new file mlx5_vdpa_mem.c to manage all the MR related code
in the driver.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Acked-by: Maxime Coquelin <maxime.coquelin@redhat.com>
4 years agovdpa/mlx5: support features get operations
Matan Azrad [Sun, 2 Feb 2020 16:03:43 +0000 (16:03 +0000)]
vdpa/mlx5: support features get operations

Add support for get_features and get_protocol_features operations.

Part of the features are reported by the DevX capabilities.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
4 years agovdpa/mlx5: support queues number operation
Matan Azrad [Sun, 2 Feb 2020 16:03:42 +0000 (16:03 +0000)]
vdpa/mlx5: support queues number operation

Support get_queue_num operation to get the maximum number of queues
supported by the device.

This number comes from the DevX capabilities.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
4 years agovdpa/mlx5: introduce Mellanox vDPA driver
Matan Azrad [Sun, 2 Feb 2020 16:03:41 +0000 (16:03 +0000)]
vdpa/mlx5: introduce Mellanox vDPA driver

Add a new driver to support vDPA operations by Mellanox devices.

The first Mellanox devices which support vDPA operations are
ConnectX-6 Dx and Bluefield1 HCA for their PF ports and VF ports.

This driver is depending on rdma-core like the mlx5 PMD, also it is
going to use mlx5 DevX to create HW objects directly by the FW.
Hence, the common/mlx5 library is linked to the mlx5_vdpa driver.

This driver will not be compiled by default due to the above
dependencies.

Register a new log type for this driver.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
4 years agonet/mlx5: add Rx/Tx burst mode info
Alexander Kozyrev [Thu, 30 Jan 2020 16:28:01 +0000 (18:28 +0200)]
net/mlx5: add Rx/Tx burst mode info

Get a burst mode information for Rx/Tx queues in mlx5.
Provide callback functions to show this information in
a "show rxq info" and "show txq info" output.

Signed-off-by: Alexander Kozyrev <akozyrev@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agonet/octeontx: fix user supplied MAC address index
Sunil Kumar Kori [Wed, 29 Jan 2020 06:42:42 +0000 (12:12 +0530)]
net/octeontx: fix user supplied MAC address index

Earlier after a successful mac_addr_add operation, index was returned
by underlying layer which was unused but same as provided by DPDK API.

So API is enhanced to use application provided index location to add
MAC address entry.

Fixes: e4373bf1b3f5 ("net/octeontx: add unicast MAC filter")

Signed-off-by: Sunil Kumar Kori <skori@marvell.com>
Acked-by: Harman Kalra <hkalra@marvell.com>
4 years agonet/octeontx: fix memory leak of MAC address table
Sunil Kumar Kori [Wed, 29 Jan 2020 09:17:04 +0000 (14:47 +0530)]
net/octeontx: fix memory leak of MAC address table

MAC address table is allocated during octeontx device create and
same is used to maintain list of MAC address associated to port.
This table is not getting freed niether in case of error nor during
graceful shutdown of port.

Patch fixes memory required memory for both the cases as mentioned.

Fixes: f18b146c498d ("net/octeontx: create ethdev ports")
Cc: stable@dpdk.org
Signed-off-by: Sunil Kumar Kori <skori@marvell.com>
Acked-by: Harman Kalra <hkalra@marvell.com>
4 years agonet/octeontx2: fix Tx flow control for HIGIG
Kiran Kumar K [Thu, 30 Jan 2020 16:23:24 +0000 (21:53 +0530)]
net/octeontx2: fix Tx flow control for HIGIG

Tx flow controlled is disabled in the Ax silicon version due to an errata.
This errata is not applicable for HIGIG Tx flow control, therefore
not enabling in HIGIG case.

Fixes: 602009ee2dfb ("net/octeontx2: support HIGIG2")
Cc: stable@dpdk.org
Signed-off-by: Kiran Kumar K <kirankumark@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
4 years agonet/i40e: fix multi-queue Rx interrupt for VF
Lunyuan Cui [Fri, 17 Jan 2020 08:22:58 +0000 (08:22 +0000)]
net/i40e: fix multi-queue Rx interrupt for VF

The interrupt vector which bind to queues should not be larger than
the max available vector. It will cause port start failed. This patch
changed the judgement condition of the limited vector id. It can
effectively avoid vector id out of range.

Fixes: 6a6cf5f88b4a ("net/i40e: enable multi-queue Rx interrupt for VF")

Signed-off-by: Lunyuan Cui <lunyuanx.cui@intel.com>
Reviewed-by: Qiming Yang <qiming.yang@intel.com>
4 years agonet/ice: fix GTP-U rule conflict
Qi Zhang [Wed, 22 Jan 2020 05:28:24 +0000 (13:28 +0800)]
net/ice: fix GTP-U rule conflict

The patch distinguishes fdir rules for GTPU with or without
extend header, so flow to match below patterns can be created
correctly.

1. eth / ipv4 / udp / gtpu teid is 10 / ...
2. eth / ipv4 / udp / gtpu teid is 10 / gtp_psc / ...
3. eth / ipv4 / udp / gtpu / gtp_psc qfi is 10 / ...
4. eth / ipv4 / udp / gtpu teid is 10 / gtp_psc is 10 / ...

Fixes: efc16c621415 ("net/ice: support flow director GTPU tunnel")
Cc: stable@dpdk.org
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Xiaolong Ye <xiaolong.ye@intel.com>
4 years agonet/e1000: use macro for PCI log format
Yunjian Wang [Mon, 13 Jan 2020 12:20:28 +0000 (20:20 +0800)]
net/e1000: use macro for PCI log format

Use PCI_PRI_FMT instead of "%04d:%02d:%02d:%d" print format.

Signed-off-by: Yunjian Wang <wangyunjian@huawei.com>
Acked-by: Xiaolong Ye <xiaolong.ye@intel.com>
4 years agonet/mlx5: fix VXLAN-GPE item translation
Raslan Darawsheh [Mon, 27 Jan 2020 15:34:12 +0000 (17:34 +0200)]
net/mlx5: fix VXLAN-GPE item translation

Currently, when using VXLAN-GPE or VXLAN item in the flow
both are being treated the same with flags 0x8 in VXLAN
header. Which mean the matching of the item VXLAN-GPE
will match any VXLAN packet.

This fixes the translation of VXLAN GPE item into PMD flow
item. Which will by default set the flags to VXLAN-GPE
to be 0xc.

Fixes: 3d69434113d1 ("net/mlx5: add Direct Verbs validation function")
Cc: stable@dpdk.org
Signed-off-by: Raslan Darawsheh <rasland@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>
4 years agodoc: add secondary VF details in thunderx guide
Krzysztof Kanas [Tue, 28 Jan 2020 13:31:47 +0000 (14:31 +0100)]
doc: add secondary VF details in thunderx guide

thunderx-nic uses secondary VF's to provide more queues to DPDK.
Current instructions explain the concept but don't show an easy way to
find which PCI id is primary and which is secondary VF's.

This patch extending the documentation of secondary VF w.r.t
the enumeration details.

Signed-off-by: Krzysztof Kanas <kkanas@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
4 years agonet/octeontx2: change default RSS hash calculation
Jerin Jacob [Wed, 29 Jan 2020 17:24:12 +0000 (22:54 +0530)]
net/octeontx2: change default RSS hash calculation

Before C0 HW revision, The RSS adder was computed based the following
static formula.

rss_adder<7:0> = flow_tag<7:0> ^ flow_tag<15:8> ^
flow_tag<23:16> ^ flow_tag<31:24>

The above scheme has the following drawbacks:
1) It is not in line with other standard NIC behavior.
2) There can be an SW use case where SW can compute the hash
upfront using Toeplitz function and predict the queue selection
to optimize some packet lookup function. The nonstandard
way of doing XOR makes the consumer to not predict the queue selection.

C0 HW revision onward, The HW can configure the
rss_adder<7:0> as flow_tag<7:0> to align with standard NICs.

This patch adds an option to select legacy RSS adder mode
using tag_as_xor=1 devargs option while keeping the standard NIC
behavior as default.

Signed-off-by: Jerin Jacob <jerinj@marvell.com>
4 years agonet/ionic: fix packet type mask
Alfredo Cardigliano [Thu, 30 Jan 2020 17:26:49 +0000 (18:26 +0100)]
net/ionic: fix packet type mask

Fix the IONIC_RXQ_COMP_PKT_TYPE_MASK define.

Coverity issue: 353608
Fixes: 01489e5d79a7 ("net/ionic: add hardware structures definitions")

Signed-off-by: Alfredo Cardigliano <cardigliano@ntop.org>
4 years agonet/bnxt: remove spurious warning in Rx handler
Somnath Kotur [Fri, 31 Jan 2020 06:16:58 +0000 (11:46 +0530)]
net/bnxt: remove spurious warning in Rx handler

HW seems to populate the cfa code in the Rx descriptor even
if an explicit flow rule is not configured via application as
there might be a default rule configured in HW even for promisc
mode.

Fixes: 94eb699bc82e ("net/bnxt: support flow mark action")

Signed-off-by: Somnath Kotur <somnath.kotur@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
4 years agocommon/mlx5: improve assert control
Alexander Kozyrev [Thu, 30 Jan 2020 16:14:40 +0000 (18:14 +0200)]
common/mlx5: improve assert control

Use the MLX5_ASSERT macros instead of the standard assert clause.
Depends on the RTE_LIBRTE_MLX5_DEBUG configuration option to define it.
If RTE_LIBRTE_MLX5_DEBUG is enabled MLX5_ASSERT is equal to RTE_VERIFY
to bypass the global CONFIG_RTE_ENABLE_ASSERT option.
If RTE_LIBRTE_MLX5_DEBUG is disabled, the global CONFIG_RTE_ENABLE_ASSERT
can still make this assert active by calling RTE_VERIFY inside RTE_ASSERT.

Signed-off-by: Alexander Kozyrev <akozyrev@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agocommon/mlx5: remove NDEBUG
Alexander Kozyrev [Thu, 30 Jan 2020 16:14:39 +0000 (18:14 +0200)]
common/mlx5: remove NDEBUG

Use the RTE_LIBRTE_MLX5_DEBUG configuration flag to get rid of dependency
on the NDEBUG definition. This is a preparation step to switch
from standard assert clauses to DPDK RTE_ASSERT ones in MLX5 driver.

Signed-off-by: Alexander Kozyrev <akozyrev@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agonet/mlx4: improve assert control
Alexander Kozyrev [Thu, 30 Jan 2020 16:14:38 +0000 (18:14 +0200)]
net/mlx4: improve assert control

Use the MLX4_ASSERT macros instead of the standard assert clause.
Depends on the RTE_LIBRTE_MLX4_DEBUG configuration option to define it.
If RTE_LIBRTE_MLX4_DEBUG is enabled MLX4_ASSERT is equal to RTE_VERIFY
to bypass the global CONFIG_RTE_ENABLE_ASSERT option.
If RTE_LIBRTE_MLX4_DEBUG is disabled, the global CONFIG_RTE_ENABLE_ASSERT
can still make this assert active by calling RTE_VERIFY inside RTE_ASSERT.

Signed-off-by: Alexander Kozyrev <akozyrev@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agonet/mlx4: remove NDEBUG flag
Alexander Kozyrev [Thu, 30 Jan 2020 16:14:37 +0000 (18:14 +0200)]
net/mlx4: remove NDEBUG flag

Use the RTE_LIBRTE_MLX4_DEBUG compilation flag to get rid of dependency
on the NDEBUG definition. This is a preparation step to switch
from standard assert clauses to DPDK RTE_ASSERT ones in MLX4 driver.

Signed-off-by: Alexander Kozyrev <akozyrev@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agomk: remove promotion of icc warnings as errors
Alexander Kozyrev [Thu, 30 Jan 2020 16:14:36 +0000 (18:14 +0200)]
mk: remove promotion of icc warnings as errors

Remove -Werror-all flag in ICC configuration file to stop treating ICC
warnings as errors in DPDK due to many false positives. We are using
GCC and Clang as a benchmark for warnings anyway for simplification.

Suggested-by: Thomas Monjalon <thomas@monjalon.net>
Signed-off-by: Alexander Kozyrev <akozyrev@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Acked-by: Thomas Monjalon <thomas@monjalon.net>
4 years agonet/mlx5: add no-inline Tx flag
Viacheslav Ovsiienko [Wed, 29 Jan 2020 12:21:07 +0000 (12:21 +0000)]
net/mlx5: add no-inline Tx flag

This patch adds support for dynamic flag that hints transmit
datapath do not copy data to the descriptors. This flag is
useful when data are located in the memory of another (not NIC)
physical device and copying to the host memory is undesirable.

This hint flag is per mbuf for multi-segment packets.

This hint flag might be partially ignored if:

- hardware requires minimal data header to be inline into
  descriptor, it depends on the hardware type and its configuration.
  In this case PMD copies the minimal required number of bytes to
  the descriptor, ignoring the no inline hint flag, the rest of data
  is not copied.

- VLAN tag insertion offload is requested and hardware does not
  support this options. In this case the VLAN tag is inserted by
  software means and at least 18B are copied to descriptor.

Signed-off-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>
4 years agonet/mlx5: support fine grain dynamic flag
Ori Kam [Wed, 29 Jan 2020 12:21:06 +0000 (12:21 +0000)]
net/mlx5: support fine grain dynamic flag

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>
4 years agonet/mlx5: make FDB default rule optional
Matan Azrad [Tue, 28 Jan 2020 17:06:43 +0000 (17:06 +0000)]
net/mlx5: make FDB default rule optional

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 b67b4ecbde22 ("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>
4 years agocommon/mlx5: support ROCE disable through Netlink
Matan Azrad [Wed, 29 Jan 2020 12:38:50 +0000 (12:38 +0000)]
common/mlx5: support ROCE disable through Netlink

Add new 4 Netlink commands to support enable/disable ROCE:
        1. mlx5_nl_devlink_family_id_get to get the Devlink family ID of
           Netlink general command.
        2. mlx5_nl_enable_roce_get to get the ROCE current status.
        3. mlx5_nl_driver_reload - to reload the device kernel driver.
        4. mlx5_nl_enable_roce_set - to set the ROCE status.

When the user changes the ROCE status, the IB device may disappear and
appear again, so DPDK driver should wait for it and to restart itself.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agocommon/mlx5: share Netlink commands
Matan Azrad [Wed, 29 Jan 2020 12:38:49 +0000 (12:38 +0000)]
common/mlx5: share Netlink commands

Move Netlink mechanism and its dependencies from net/mlx5 to
common/mlx5 in order to be ready to use by other mlx5 drivers.

The dependencies are BITFIELD defines, the ppc64 compilation workaround
for bool type and the function mlx5_translate_port_name.

Update build mechanism accordingly.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agonet/mlx5: reduce Netlink commands dependencies
Matan Azrad [Wed, 29 Jan 2020 12:38:48 +0000 (12:38 +0000)]
net/mlx5: reduce Netlink commands dependencies

As an arrangment for Netlink command moving to the common library,
reduce the net/mlx5 dependencies.

Replace ethdev class command parameters.

Improve Netlink sequence number mechanism to be controlled by the
mlx5 Netlink mechanism.

Move mlx5_nl_check_switch_info to mlx5_nl.c since it is the only one
which uses it.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agonet/mlx5: separate Netlink command interface
Matan Azrad [Wed, 29 Jan 2020 12:38:47 +0000 (12:38 +0000)]
net/mlx5: separate Netlink command interface

The Netlink commands interfaces is included in the mlx5.h file with a
lot of other PMD interfaces.

As an arrangement to make the Netlink commands shared with different
PMDs, this patch moves the Netlink interface to a new file called
mlx5_nl.h.

Move non Netlink pure vlan commands from mlx5_nl.c to the
mlx5_vlan.c.

Rename Netlink commands and structures to use prefix mlx5_nl.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agonet/mlx5: select driver by class device argument
Matan Azrad [Wed, 29 Jan 2020 12:38:46 +0000 (12:38 +0000)]
net/mlx5: select driver by class device argument

There might be a case that one Mellanox device can be probed by
multiple mlx5 drivers.

One case is that any mlx5 vDPA device can be probed by both net/mlx5
and vdpa/mlx5.

Add a new mlx5 common API to get the requested driver by devargs:
class=[net/vdpa].

Skip net/mlx5 PMD probing while the device is selected to be probed by
the vDPA driver.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agocommon/mlx5: get DevX capability for max RQT size
Matan Azrad [Wed, 29 Jan 2020 12:38:45 +0000 (12:38 +0000)]
common/mlx5: get DevX capability for max RQT size

In order to allow RQT size configuration which is limited to the
correct maximum value, add log_max_rqt_size for DevX capability
structure.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agocommon/mlx5: add DevX command to modify RQT
Matan Azrad [Wed, 29 Jan 2020 12:38:44 +0000 (12:38 +0000)]
common/mlx5: add DevX command to modify RQT

RQ table can be changed to support different list of queues.

Add DevX command to modify DevX RQT object to point on new RQ list.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agocommon/mlx5: add TIR field constants
Matan Azrad [Wed, 29 Jan 2020 12:38:43 +0000 (12:38 +0000)]
common/mlx5: add TIR field constants

The DevX TIR object configuration should get L3 and L4 protocols
expected to be forwarded by the TIR.

Add the PRM constant values needed to configure the L3 and L4 protocols.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agocommon/mlx5: allow type configuration for DevX RQT
Matan Azrad [Wed, 29 Jan 2020 12:38:42 +0000 (12:38 +0000)]
common/mlx5: allow type configuration for DevX RQT

Allow virtio queue type configuration in the RQ table.
The needed fields and configuration was added.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agocommon/mlx5: support DevX QP operations
Matan Azrad [Wed, 29 Jan 2020 12:38:41 +0000 (12:38 +0000)]
common/mlx5: support DevX QP operations

QP creation is needed for vDPA virtq support.

Add 2 DevX commands to create QP and to modify QP state.

The support is for RC QP only in force loopback address mode.

By this way, the packets can be sent to other inernal destinations in
the nic. For example: other QPs or virtqs.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agocommon/mlx5: add DevX virtq commands
Matan Azrad [Wed, 29 Jan 2020 12:38:40 +0000 (12:38 +0000)]
common/mlx5: add DevX virtq commands

Virtio emulation offload allows SW to offload the I/O operations of a
virtio virtqueue, using the device, allowing an improved performance
for its users.
While supplying all the relevant Virtqueue information (type, size,
memory location, doorbell information, etc.). The device can then
offload the I/O operation of this queue, according to its device type
characteristics.

Some of the virtio features can be supported according to the device
capability, for example, TSO and checksum.

Add virtio queue create, modify and query DevX commands.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agocommon/mlx5: glue VAR allocation
Matan Azrad [Wed, 29 Jan 2020 12:38:39 +0000 (12:38 +0000)]
common/mlx5: glue VAR allocation

Virtio access region(VAR) is the UAR that allocated for virtio emulation
access.

Add rdma-core operations to allocate and free VAR.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agocommon/mlx5: add DevX command to create CQ
Matan Azrad [Wed, 29 Jan 2020 12:38:38 +0000 (12:38 +0000)]
common/mlx5: add DevX command to create CQ

HW implements completion queues(CQ) used to post completion reports upon
completion of work request.

Used for Rx and Tx datapath.

Add DevX command to create a CQ.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agocommon/mlx5: glue UAR allocation
Matan Azrad [Wed, 29 Jan 2020 12:38:37 +0000 (12:38 +0000)]
common/mlx5: glue UAR allocation

The isolated, protected and independent direct access to the HW by
multiple processes is implemented via User Access Region (UAR)
mechanism.
The UAR is part of PCI address space that is mapped for direct access to
the HW from the CPU.
UAR is comprised of multiple pages, each page containing registers that
control the HW operation.
UAR mechanism is used to post execution or control requests to the HW.
It is used by the HW to enforce protection and isolation between
different processes.

Add a glue command to allocate and free an UAR.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agocommon/mlx5: glue event interrupt commands
Matan Azrad [Wed, 29 Jan 2020 12:38:36 +0000 (12:38 +0000)]
common/mlx5: glue event interrupt commands

Add the next commands to glue in order to support interrupt event
channel operations associated to events in the EQ:
devx_create_event_channel,
devx_destroy_event_channel,
devx_subscribe_devx_event,
devx_subscribe_devx_event_fd,
devx_get_event.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agocommon/mlx5: glue event queue query
Matan Azrad [Wed, 29 Jan 2020 12:38:35 +0000 (12:38 +0000)]
common/mlx5: glue event queue query

The event queue is managed only by the kernel.

Add the rdma-core command in glue to query the kernel event queue
details.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agocommon/mlx5: support DevX indirect mkey creation
Matan Azrad [Wed, 29 Jan 2020 12:38:34 +0000 (12:38 +0000)]
common/mlx5: support DevX indirect mkey creation

Add option to create an indirect mkey by the current
mlx5_devx_cmd_mkey_create command.

Indirect mkey points to set of direct mkeys.
By this way, the HW\SW can reference fragmented memory by one object.

Align the net/mlx5 driver usage in the above command.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agocommon/mlx5: glue null memory region allocation
Matan Azrad [Wed, 29 Jan 2020 12:38:33 +0000 (12:38 +0000)]
common/mlx5: glue null memory region allocation

Add support for rdma-core API to allocate NULL MR.

When the device HW get a NULL MR address, it will do nothing with the
address, no read and no write.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agocommon/mlx5: query vDPA DevX capabilities
Matan Azrad [Wed, 29 Jan 2020 12:38:32 +0000 (12:38 +0000)]
common/mlx5: query vDPA DevX capabilities

Add the DevX capabilities for vDPA configuration and information of
Mellanox devices.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agocommon/mlx5: share CQ entry check
Matan Azrad [Wed, 29 Jan 2020 12:38:31 +0000 (12:38 +0000)]
common/mlx5: share CQ entry check

The CQE has owner bit to indicate if it is in SW control or HW.

Share a CQE check for all the mlx5 drivers.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agocommon/mlx5: share devices information
Matan Azrad [Wed, 29 Jan 2020 12:38:30 +0000 (12:38 +0000)]
common/mlx5: share devices information

Move the vendor information, vendor ID and device IDs from net/mlx5 PMD
to the common mlx5 file.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agocommon/mlx5: share PCI device detection
Matan Azrad [Wed, 29 Jan 2020 12:38:29 +0000 (12:38 +0000)]
common/mlx5: share PCI device detection

Move PCI detection by IB device from mlx5 PMD to the common code.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agocommon/mlx5: introduce common library
Matan Azrad [Wed, 29 Jan 2020 12:38:27 +0000 (12:38 +0000)]
common/mlx5: introduce common library

A new Mellanox vdpa PMD will be added to support vdpa operations by
Mellanox adapters.

This vdpa PMD design includes mlx5_glue and mlx5_devx operations and
large parts of them are shared with the net/mlx5 PMD.

Create a new common library in drivers/common for mlx5 PMDs.
Move mlx5_glue, mlx5_devx_cmds and their dependencies to the new mlx5
common library in drivers/common.

The files mlx5_devx_cmds.c, mlx5_devx_cmds.h, mlx5_glue.c,
mlx5_glue.h and mlx5_prm.h are moved as is from drivers/net/mlx5 to
drivers/common/mlx5.

Share the log mechanism macros.
Separate also the log mechanism to allow different log level control to
the common library.

Build files and version files are adjusted accordingly.
Include lines are adjusted accordingly.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agonet/mlx5: separate DevX commands interface
Matan Azrad [Wed, 29 Jan 2020 12:38:26 +0000 (12:38 +0000)]
net/mlx5: separate DevX commands interface

The DevX commands interface is included in the mlx5.h file with a lot
of other PMD interfaces.

As an arrangement to make the DevX commands shared with different PMDs,
this patch moves the DevX interface to a new file called mlx5_devx_cmds.h.

Also remove shared device structure dependency on DevX commands.

Replace the DevX commands log mechanism from the mlx5 driver log
mechanism to the EAL log mechanism.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agonet/bnx2x: fix VLAN stripped flag
Rasesh Mody [Wed, 29 Jan 2020 01:43:13 +0000 (17:43 -0800)]
net/bnx2x: fix VLAN stripped flag

For VLAN packet the tci is saved in rx_mb->vlan_tci, however the
STRIPPED offload flag is not set along with PKT_RX_VLAN flag.
Set the PKT_RX_VLAN_STRIPPED flag as well.

Fixes: 380a7aab1ae2 ("mbuf: rename deprecated VLAN flags")
Fixes: b37b528d957c ("mbuf: add new Rx flags for stripped VLAN")
Cc: stable@dpdk.org
Signed-off-by: Rasesh Mody <rmody@marvell.com>
4 years agonet/bnxt: fix return code handling in VLAN config
Kalesh AP [Tue, 28 Jan 2020 07:29:23 +0000 (12:59 +0530)]
net/bnxt: fix return code handling in VLAN config

return value stored in "ret" but it has been overwritten before use.

Coverity issue: 353621
Fixes: 7fe5668d2ea3 ("net/bnxt: support VLAN filter and strip")
Fixes: df6cd7c1f73a ("net/bnxt: handle reset notify async event from FW")
Cc: stable@dpdk.org
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Somnath Kotur <somnath.kotur@broadcom.com>
4 years agonet/bnxt: fix supporting zero mark ID with RSS action
Somnath Kotur [Tue, 28 Jan 2020 07:29:22 +0000 (12:59 +0530)]
net/bnxt: fix supporting zero mark ID with RSS action

Certain applications(Ex: OVS-DPDK) can issue rte_flow_create with mark
id set to 0. The mark table entry creation in the driver was assuming
non-zero mark ids. Fix it to have a valid flag for each entry.
Also, it is possible that both RSS flags and cfa_code/ mark id are set
as part of the Rx packet completion descriptor if the 'action' for a
flow is MARK + RSS.

Fix Rx completion processing to look for both fields as opposed to only
either.

Fixes: 94eb699bc82e ("net/bnxt: support flow mark action")

Signed-off-by: Somnath Kotur <somnath.kotur@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
4 years agonet/bnxt: remove a redundant variable
Kalesh AP [Tue, 28 Jan 2020 07:29:21 +0000 (12:59 +0530)]
net/bnxt: remove a redundant variable

Use "dev->data->dev_started" state, instead of local "dev_stopped"
to check whether port has been started or not.

Fixes: 316e412299fd ("net/bnxt: fix crash when closing")
Cc: stable@dpdk.org
Reviewed-by: Santoshkumar Karanappa Rastapur <santosh.rastapur@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Somnath Kotur <somnath.kotur@broadcom.com>
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
4 years agonet/bnxt: remove unnecessary structure variable
Kalesh AP [Tue, 28 Jan 2020 07:29:20 +0000 (12:59 +0530)]
net/bnxt: remove unnecessary structure variable

This change could help in reducing the size of bnxt PMD private
data structure by converting a uint8_t variable to use bit map flag.

Fixes: 5cd0e2889c43 ("net/bnxt: support NIC Partitioning")
Cc: stable@dpdk.org
Reviewed-by: Santoshkumar Karanappa Rastapur <santosh.rastapur@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Somnath Kotur <somnath.kotur@broadcom.com>
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
4 years agonet/bnxt: remove redundant macro
Kalesh AP [Tue, 28 Jan 2020 07:29:19 +0000 (12:59 +0530)]
net/bnxt: remove redundant macro

Use "dev->data->dev_started" state, instead of local BNXT_FLAG_INIT_DONE
to check whether device has been initialised or not.

Fixes: ed2ced6fe927 ("net/bnxt: check initialization before accessing stats")
Cc: stable@dpdk.org
Reviewed-by: Santoshkumar Karanappa Rastapur <santosh.rastapur@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Somnath Kotur <somnath.kotur@broadcom.com>
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
4 years agonet/bnxt: remove redundant if statement
Kalesh AP [Tue, 28 Jan 2020 07:29:18 +0000 (12:59 +0530)]
net/bnxt: remove redundant if statement

Since "eth_dev->data->dev_started" has been assigned to 0 at the
beginning of bnxt_dev_stop_op() routine, the code inside the if()
condition is redundant. Remove it.

Anyways "eth_dev->data->dev_link.link_status" will be set to 0 in
bnxt_dev_set_link_down_op() later in the routine.

Fixes: 316e412299fd ("net/bnxt: fix crash when closing")
Cc: stable@dpdk.org
Signed-off-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Signed-off-by: Somnath Kotur <somnath.kotur@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
4 years agonet/bnxt: allow group ID 0 for RSS action
Somnath Kotur [Tue, 28 Jan 2020 07:29:17 +0000 (12:59 +0530)]
net/bnxt: allow group ID 0 for RSS action

Allow RSS action with group ID 0. The RSS match check will ensure if
requested RSS action configuration parameters should be allowed as per
the VNIC's RSS configuration or not and if it does match as it is
by design in OVS-DPDK use case, the default vnic can be used to create
the filter. As part of this ensure that rx_queue_cnt for the default
vnic is setup during init as this field was being set only for non-zero
vnics while handling RSS action for flow create.
Check for out of bounds error while accessing the vnic array.

Fixes: adc0f81c6552 ("net/bnxt: support RSS action")
Cc: stable@dpdk.org
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Somnath Kotur <somnath.kotur@broadcom.com>
4 years agonet/bnxt: fix bumping of L2 filter reference count
Somnath Kotur [Tue, 28 Jan 2020 07:29:16 +0000 (12:59 +0530)]
net/bnxt: fix bumping of L2 filter reference count

Now that the L2 filter reference count is bumped up in all cases
including bnxt_alloc_filter() which is issued in init, just move this
ref count bump inside the routine issuing the HWRM cmd so that it is
bumped up only if the cmd is successful.

Fixes: 5c1171c97216 ("net/bnxt: refactor filter/flow")
Cc: stable@dpdk.org
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Signed-off-by: Somnath Kotur <somnath.kotur@broadcom.com>
Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
4 years agonet/bnxt: fix alloc filter to use a common routine
Somnath Kotur [Tue, 28 Jan 2020 07:29:15 +0000 (12:59 +0530)]
net/bnxt: fix alloc filter to use a common routine

Invoke bnxt_get_unused_filter() inside bnxt_alloc_filter() so that
all filters are allocated from one common routine.

Fixes: f92735db1e4c ("net/bnxt: add L2 filter alloc/init/free")
Cc: stable@dpdk.org
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Signed-off-by: Somnath Kotur <somnath.kotur@broadcom.com>
Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
4 years agonet/octeontx2: disable HIGIG on port stop
Kiran Kumar K [Fri, 24 Jan 2020 13:54:26 +0000 (19:24 +0530)]
net/octeontx2: disable HIGIG on port stop

If HIGIG mode is enabled on configure, This needs to be disabled
on port stop. Adding support to send mbox message on port stop
to configure the port to default.

Signed-off-by: Kiran Kumar K <kirankumark@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
4 years agonet/octeontx2: extend RSS supported offload types
Kiran Kumar K [Fri, 24 Jan 2020 03:46:02 +0000 (09:16 +0530)]
net/octeontx2: extend RSS supported offload types

Extend RSS offload types for octeontx2. Add support to select
L3 SRC, L3 DST, L4 SRC and L4 DST for RSS calculation.

Add support to select L3 SRC or DST only, L4 SRC or DST only for RSS
calculation.

With this requirement there will be following combinations,
IPV[4,6]_SRC_ONLY, IPV[4,6]_DST_ONLY, [TCP,UDP,SCTP]_SRC_ONLY,
[TCP,UDP,SCTP]_DST_ONLY. So, instead of creating a bit for each
combination, we are using upper 4 bits (31:28) in the flow_key_cfg
to represent the SRC, DST selection. 31 => L3_SRC, 30 => L3_DST,
29 => L4_SRC, 28 => L4_DST. These won't be part of flow_cfg, so that
we don't need to change the existing ABI.

Signed-off-by: Kiran Kumar K <kirankumark@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
4 years agonet/bnx2x: fix to sync fastpath Rx queue access
Rasesh Mody [Sun, 26 Jan 2020 22:54:19 +0000 (14:54 -0800)]
net/bnx2x: fix to sync fastpath Rx queue access

PMD handles fast path completions in the Rx handler and control path
completions in the interrupt handler. They both are processing
completions from the same fastpath completion queue. There is a
potential for race condition when these two paths are processing
the completions from the same queue and trying to updating Rx Producer.

Add a fastpath Rx lock between these two paths to close this race.

Fixes: 540a211084a7 ("bnx2x: driver core")
Cc: stable@dpdk.org
Signed-off-by: Rasesh Mody <rmody@marvell.com>
4 years agonet/bnx2x: fix reset of scan FP flag
Rasesh Mody [Sun, 26 Jan 2020 22:54:18 +0000 (14:54 -0800)]
net/bnx2x: fix reset of scan FP flag

The fastpath task queue handler resets the fastpath scan flag
unconditionally, this patch changes that to reset the flag
only if it was set.

Fixes: 08a6e472c3d7 ("net/bnx2x: fix packet drop")
Cc: stable@dpdk.org
Signed-off-by: Rasesh Mody <rmody@marvell.com>
4 years agoapp/testpmd: fix RFC addresses for Tx only
Stephen Hemminger [Mon, 27 Jan 2020 03:34:58 +0000 (19:34 -0800)]
app/testpmd: fix RFC addresses for Tx only

The previous patch to change default IP addresses for tx only
mode got the wrong values (typo).

Fixes: bf5b2126bf44 ("app/testpmd: add ability to set Tx IP and UDP parameters")
Cc: stable@dpdk.org
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
4 years agonet/mlx5: fix register usage in meter
Suanming Mou [Thu, 23 Jan 2020 06:01:02 +0000 (08:01 +0200)]
net/mlx5: fix register usage in meter

Flow with meter will split to three subflows, the prefix subflow with
meter action do the color, the meter subflow  filter the packets, the
suffix subflow do all the left actions for packets pass the filter.
Both the color and the subflow match between prefix and suffix use the
register to store the tag.

For some of the NICs with meter color register share capability, it
only uses 8 LSB of the register for color, the left 24 MSB can be used
for flow id match between meter prefix subflow and suffix subflow.

Currently, one entire register is allocated for flow matching which
causes the NICs with limited registers don't have enough register for
other matching.

Add the meter color share capability checking to fix lacking of
registers issue.

Fixes: 9ea9b049a960 ("net/mlx5: split meter flow")
Cc: stable@dpdk.org
Signed-off-by: Suanming Mou <suanmingm@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agonet/mlx5: support maximum flow id allocation
Suanming Mou [Thu, 23 Jan 2020 06:01:01 +0000 (08:01 +0200)]
net/mlx5: support maximum flow id allocation

The id allocated is for the register unique id match. Some registers may
not use the full 32 bits. Add the maximum id to avoid allocate id over
the register restriction.

Signed-off-by: Suanming Mou <suanmingm@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agonet/mlx5: fix ICMPv6 header rewrite actions
Shiri Kuzin [Wed, 22 Jan 2020 14:37:54 +0000 (14:37 +0000)]
net/mlx5: fix ICMPv6 header rewrite actions

The ConnectX-5 HW cannot calculate the checksum for ICMPv6,
therefore flows with pattern 'ipv6 proto is 58' with actions that change
the header should be rejected. the actions that change the header
in this type of flow are 'set_ipv6_src' and 'set_ipv6_dst'.

Fixes: 4bb14c83df95 ("net/mlx5: support modify header using Direct Verbs")
Cc: stable@dpdk.org
Signed-off-by: Shiri Kuzin <shirik@mellanox.com>
Acked-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
4 years agonet/mlx5: fix setting of port ID for egress rules
Dekel Peled [Wed, 22 Jan 2020 12:33:06 +0000 (14:33 +0200)]
net/mlx5: fix setting of port ID for egress rules

Previous patch fixed the setting of port-id for eswitch rules, which
are ingress only.
This patch expands the fix, to support NIC rules as well, which can
be ingress or egress.

Fixes: ce777b147bf8 ("net/mlx5: fix E-Switch flow without port item")
Cc: stable@dpdk.org
Signed-off-by: Dekel Peled <dekelp@mellanox.com>
Acked-by: Ori Kam <orika@mellanox.com>
4 years agonet/mlx5: fix zero out UDP checksum in encap data
Eli Britstein [Wed, 22 Jan 2020 15:03:58 +0000 (15:03 +0000)]
net/mlx5: fix zero out UDP checksum in encap data

The cited commit zeroed the UDP checksum for raw-encap case.
Add the same handling for vxlan-encap case.

Fixes: bf1d7d9a033a ("net/mlx5: zero out UDP checksum in encapsulation")
Cc: stable@dpdk.org
Signed-off-by: Eli Britstein <elibr@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>