On systems where the required Netlink commands are not supported but
Mellanox OFED is installed, representors information must be retrieved
through sysfs.
net/mlx5: add port ID pattern item to switch flow rules
This enables flow rules to match traffic coming from a different DPDK port
ID associated with the device (PORT_ID pattern item), mainly for the
convenience of applications that want to deal with a single port ID for all
flow rules associated with some physical device.
Testpmd example:
- Creating a flow rule on port ID 1 to consume all traffic from port ID 0
and direct it to port ID 2:
flow create 1 ingress transfer pattern port_id id is 0 / end actions
port_id id 2 / end
net/mlx5: add VLAN item and actions to switch flow rules
This enables flow rules to explicitly match VLAN traffic (VLAN pattern
item) and perform various operations on VLAN headers at the switch level
(OF_POP_VLAN, OF_PUSH_VLAN, OF_SET_VLAN_VID and OF_SET_VLAN_PCP actions).
Testpmd examples:
- Directing all VLAN traffic received on port ID 1 to port ID 0:
flow create 1 ingress transfer pattern eth / vlan / end actions
port_id id 0 / end
- Adding a VLAN header to IPv6 traffic received on port ID 1 and directing
it to port ID 0:
flow create 1 ingress transfer pattern eth / ipv6 / end actions
of_push_vlan ethertype 0x8100 / of_set_vlan_vid vlan_vid 42 /
port_id id 0 / end
This patch enables creation of rte_flow rules that direct matching traffic
to a different port (e.g. another VF representor) or drop it directly at
the switch level (PORT_ID and DROP actions).
Testpmd examples:
- Directing all traffic to port ID 0:
flow create 1 ingress transfer pattern end actions port_id id 0 / end
- Dropping all traffic normally received by port ID 1:
flow create 1 ingress transfer pattern end actions drop / end
Note the presence of the transfer attribute, which requests them to be
applied at the switch level. All traffic is matched due to empty pattern.
Because mlx5 switch flow rules are configured through Netlink (TC
interface) and have little in common with Verbs, this patch adds a separate
parser function to handle them.
- mlx5_nl_flow_transpose() converts a rte_flow rule to its TC equivalent
and stores the result in a buffer.
- mlx5_nl_flow_brand() gives a unique handle to a flow rule buffer.
- mlx5_nl_flow_create() instantiates a flow rule on the device based on
such a buffer.
- mlx5_nl_flow_destroy() performs the reverse operation.
These functions are called by the existing implementation when encountering
flow rules which must be offloaded to the switch (currently relying on the
transfer attribute).
With mlx5, unlike normal flow rules implemented through Verbs for traffic
emitted and received by the application, those targeting different logical
ports of the device (VF representors for instance) are offloaded at the
switch level and must be configured through Netlink (TC interface).
This patch adds preliminary support to manage such flow rules through the
flow API (rte_flow).
Instead of rewriting tons of Netlink helpers and as previously suggested by
Stephen [1], this patch introduces a new dependency to libmnl [2]
(LGPL-2.1) when compiling mlx5.
While deleting the elements from the linked list, TAILQ_FOREACH causes
read from the freed pointer. Fixes the issue by using TAILQ_FOREACH_SAFE
instead.
Radu Nicolau [Wed, 18 Jul 2018 11:12:01 +0000 (12:12 +0100)]
net/bonding: fix invalid port id
Setting up the bonding options before calling rte_eth_dev_probing_finish
triggers an invalid port id error because of port state not set, or set
unused
Fixes: fbe90cdd776c ("ethdev: add probing finish function") Cc: stable@dpdk.org Signed-off-by: Radu Nicolau <radu.nicolau@intel.com> Acked-by: Chas Williams <chas3@att.com>
Andrew Rybchenko [Wed, 18 Jul 2018 07:40:06 +0000 (08:40 +0100)]
net/sfc: fix assert in set multicast address list
Return value equal to zero means success and really expected.
Fixes: 0fa0070e4391 ("net/sfc: support multicast addresses list controls") Cc: stable@dpdk.org Signed-off-by: Andrew Rybchenko <arybchenko@solarflare.com>
This feature is used to offload stripping of VLAN header from received
packets and update vlan_tci field in mbuf when
DEV_RX_OFFLOAD_VLAN_STRIP & ETH_VLAN_STRIP_MASK flag is set.
Set the rx and tx queue state appropriately when the queues or device are
started or stopped. This enables usage of the ethdev rx/tx queue start/stop
functions with the PCAP PMD.
Fixes: fd68b4739d2c ("vhost: use buffer vectors in dequeue path") Reported-by: Yinan Wang <yinan.wang@intel.com> Signed-off-by: Tiwei Bie <tiwei.bie@intel.com> Acked-by: Zhihong Wang <zhihong.wang@intel.com> Tested-by: Yinan Wang <yinan.wang@intel.com>
Andrew Rybchenko [Sun, 15 Jul 2018 09:56:14 +0000 (10:56 +0100)]
net/sfc: fix Rx queue offloads reporting in queue info
CRC_STRIP offload is always enabled, but lost here.
It is more robust to use device level offloads directly to
compose returned value.
Fixes: ff6a1197c3b1 ("net/sfc: convert to new Rx offload API") Cc: stable@dpdk.org Signed-off-by: Andrew Rybchenko <arybchenko@solarflare.com> Reviewed-by: Ivan Malov <ivan.malov@oktetlabs.ru>
Andrew Rybchenko [Sun, 15 Jul 2018 09:56:13 +0000 (10:56 +0100)]
net/sfc: move Rx checksum offload check to device level
Rx checksum offloads are on device level and should be checked and
enforced on device level.
Avoid logging of Rx checksum offloads enforced for each Rx queue.
Fixes: ff6a1197c3b1 ("net/sfc: convert to new Rx offload API") Cc: stable@dpdk.org Signed-off-by: Andrew Rybchenko <arybchenko@solarflare.com> Reviewed-by: Ivan Malov <ivan.malov@oktetlabs.ru>
Igor Romanov [Sat, 14 Jul 2018 07:38:59 +0000 (08:38 +0100)]
net/sfc: fallback to filter with zero vid
Fallback to filter with VLAN=0 if match without VLAN is not supported
Strictly speaking it is not 100% equivalent, but good tradeoff -
untagged and priority only tagged frames will match.
Signed-off-by: Igor Romanov <igor.romanov@oktetlabs.ru> Signed-off-by: Andrew Rybchenko <arybchenko@solarflare.com>
Igor Romanov [Sat, 14 Jul 2018 07:38:23 +0000 (08:38 +0100)]
net/sfc: fix filter exceptions logic
Now exception logic handles these cases:
When FW variant does not support filters with transport ports, but
IP protocol filters are supported, TCP/UDP protocol filters may be
used. When FW variant does not support filters with IPv4/6 addresses
or IP protocol, but filters with EtherType are supported, IPv4 and
IPv6 EtherTypes may be used
Fixes: 096dba799b4a ("net/sfc: avoid creation of ineffective flow rules") Cc: stable@dpdk.org Signed-off-by: Igor Romanov <igor.romanov@oktetlabs.ru> Signed-off-by: Andrew Rybchenko <arybchenko@solarflare.com>
We were using LICENSE.qede_pmd to reference inclusion of SPDX licensing
tag from all the source file. Remove the LICENSE.qede_pmd file and
directly include SPDX tags in source files.
Make a few updates in preparation for 18.08.
- Use SPDX
- Add 1400 series VIC adapters to supported models
- Describe the VXLAN port number
- Expand the description for ig-vlan-rewrite
- Add inner RSS and checksum to the features
Signed-off-by: Hyong Youb Kim <hyonkim@cisco.com> Reviewed-by: John Daley <johndale@cisco.com>
Xiaoyun Li [Thu, 12 Jul 2018 07:55:46 +0000 (15:55 +0800)]
app/testpmd: fix little performance drop
There is about 3% perf drop. And it is because of a bitrate
calculation in the datapath. So improve it by maintaining an array
of port indexes in testpmd, which is updated with ethdev events.
Fixes: 8728ccf37615 ("fix ethdev ports enumeration") Cc: stable@dpdk.org Signed-off-by: Xiaoyun Li <xiaoyun.li@intel.com> Acked-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Due to the complex NVGRE_ENCAP flow action and based on the fact testpmd
does not allocate memory, this patch adds a new command in testpmd to
initialise a global structure containing the necessary information to
make the outer layer of the packet. This same global structure will
then be used by the flow command line in testpmd when the action
nvgre_encap will be parsed, at this point, the conversion into such
action becomes trivial.
This global structure is only used for the encap action.
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com> Acked-by: Ori Kam <orika@mellanox.com> Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com> Acked-by: Mohammad Abdul Awal <mohammad.abdul.awal@intel.com> Tested-by: Mohammad Abdul Awal <mohammad.abdul.awal@intel.com>
Due to the complex VXLAN_ENCAP flow action and based on the fact testpmd
does not allocate memory, this patch adds a new command in testpmd to
initialise a global structure containing the necessary information to
make the outer layer of the packet. This same global structure will
then be used by the flow command line in testpmd when the action
vxlan_encap will be parsed, at this point, the conversion into such
action becomes trivial.
This global structure is only used for the encap action.
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com> Acked-by: Ori Kam <orika@mellanox.com> Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com> Acked-by: Mohammad Abdul Awal <mohammad.abdul.awal@intel.com> Tested-by: Mohammad Abdul Awal <mohammad.abdul.awal@intel.com>
net/i40e: fix check of flow director programming status
In i40e FDIR PMD code for checking programming status
function i40e_check_fdir_programming_status(), the initial value
of return value ret should be set to -1 not 0, because if DD bit of
I40E_RX_DESC_STATUS_DD is not write back, this function will return
0 to upper function, this give an error info to upper function, the
fact for this is it is time out for DD write back and it should return
-1.
Xiaoyun Li [Tue, 10 Jul 2018 09:30:54 +0000 (17:30 +0800)]
net/i40e: fix link speed
When link needs to go up, I40E_AQ_PHY_AN_ENABLED is always be set in DPDK.
So all speeds are always set. This causes speed config never works.
This patch fixes this issue and only allows to set available speeds. If
link needs to go up and speed setting is not supported, it will print
warning and set default available speeds. And when link needs to go down,
link speed field should be set to non-zero to avoid link down issue when
binding back to kernel driver.
Fixes: ca7e599d4506 ("net/i40e: fix link management") Fixes: 1bb8f661168d ("net/i40e: fix link down and negotiation") Cc: stable@dpdk.org Signed-off-by: Xiaoyun Li <xiaoyun.li@intel.com> Acked-by: Qi Zhang <qi.z.zhang@intel.com>
Shaopeng He [Tue, 10 Jul 2018 10:37:27 +0000 (06:37 -0400)]
net/i40e: fix Tx queue setup after stop
Currently, i40e_dev_tx_queue_setup_runtime checks simple tx and treats
mbuf fast free offloading as No-simple, which is classified as simple tx
in i40e_set_tx_function_flag. This inconsistent behavior causes tx queue
setup fail after queue was stopped. This patch fixes this bug.
Fixes: 399421100e08 ("net/i40e: fix missing mbuf fast free offload") Cc: stable@dpdk.org Signed-off-by: Shaopeng He <shaopeng.he@intel.com> Acked-by: Qi Zhang <qi.z.zhang@intel.com>
TPID can be set by set_switch_config AdminQ command on
new FW release. But find fail to set 0x88A8 on some NICs.
According to the datasheet, Switch Tag value should not
be identical to either the First Tag or Second Tag values.
So set something other than common Ethertype for internal
switching.
Fixes: 73cd7d6dc8e1 ("net/i40e: use set switch AQ instead of register setting") Cc: stable@dpdk.org Signed-off-by: Beilei Xing <beilei.xing@intel.com> Acked-by: Qi Zhang <qi.z.zhang@intel.com>
To parse packet type correctly, profile needs to be
loaded again to update tables in SW even if profile
has exited. But previously packet type parser will
be false after loading profile when profile already
exists.
This patch fixes the issue.
Currently, VFIO will try to map around MSI-X table in the BARs. When
MSI-X table (page-aligned) size is equal to (page-aligned) size of BAR,
VFIO will just skip the BAR.
Recent kernel versions will allow VFIO to map the entire BAR containing
MSI-X tables (*), so instead of trying to map around the MSI-X vector
or skipping the BAR entirely if it's not possible, we can now try
mapping the entire BAR first. If mapping the entire BAR doesn't
succeed, fall back to the old behavior of mapping around MSI-X table or
skipping the BAR.
A few regressions with virtio/vhost have been discovered, due to the
strong dependency of virtio/vhost on the underlying memory layout.
Specifically, virtio/vhost share all memory pages starting from the
beginning of the segment, while the patch below made it so that the
memory is always allocated from the top of VA space, not from the
bottom.
Fixes: 179f916e88e4 ("mem: allocate in reverse to reduce fragmentation") Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
Bruce Richardson [Fri, 20 Jul 2018 14:13:10 +0000 (15:13 +0100)]
eal: fix error message for unsupported platforms
When building with meson on e.g. cygwin, the error message about an
unsupported platform referenced an unknown variable since
"host_machine" was missing an "_".
Fixes: 844514c73569 ("eal: build with meson") Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Bruce Richardson [Thu, 19 Jul 2018 14:26:48 +0000 (15:26 +0100)]
kvargs: fix building with meson outside source tree
When building DPDK with meson with the <builddir> being outside the source
directory, an error was generated as the path to the EAL headers was not
found. The path specified for the includes backed out unnecessarily far and
so broke when the build directory was not in the expected location of
just inside the source dir. Changing the path to be shorter, just going
down one level and then into EAL builds for both cases - where builddir is
in source directory, and also when it's outside it.
Fixes: 092ee516496d ("kvargs: build before EAL") Reported-by: Anatoly Burakov <anatoly.burakov@intel.com> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Tested-by: Anatoly Burakov <anatoly.burakov@intel.com>
Bruce Richardson [Thu, 19 Jul 2018 14:22:06 +0000 (15:22 +0100)]
build: remove duplicate checks for cflags
There is no need to check for each library, driver and example whether
certain cflags are supported. Instead of checking inside the loop, do
so outside and reuse the value.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Bruce Richardson [Fri, 20 Jul 2018 13:28:25 +0000 (14:28 +0100)]
build: improve error message for missing dependency
When building a driver or library and a required internal dependency
is missing, the error message printed was:
Tried to get unknown variable "shared_rte_<libname>".
However, a better way to handle this is to catch the missing variable
earlier and convert it into a proper message, informing the user
that the required dependency is missing, and most importantly, what
library or driver wanted that dependency.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Bruce Richardson [Fri, 20 Jul 2018 13:53:36 +0000 (14:53 +0100)]
build: fix meson build on FreeBSD
A number of drivers have dependencies on libraries which
are only built on Linux, and so they need to be disabled
on FreeBSD. Rather than basing the disabling on OS, in
each case we base the building of the library on the
presence of the required dependency.
Octentx zipvf PMD provides hardware acceleration for
deflate and lzs compression and decompression operations
using Octeontx zip co-processor, which provide 8
virtualized zip devices.
This patch add basic initialization routine to register zip VFs
to compressdev library.
Add initial PMD setup routines in compressdev
framework. ZLIB PMD appears as virtual compression
device. User would need to install zlib prior to
enabling this PMD.
Pablo de Lara [Thu, 19 Jul 2018 09:05:56 +0000 (10:05 +0100)]
crypto/virtio: fix memory leak
Put session private data back to mempool when clearing
a crypto session, which is expected to be done in the PMD.
Fixes: b7fa78c7d3b0 ("crypto/virtio: support session related ops") Cc: stable@dpdk.org Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Reviewed-by: Jay Zhou <jianjay.zhou@huawei.com>
Added extra case to support 8 byte key size
for 3DES CBC. Also changed capabilities to reflect
the change.
Signed-off-by: Marko Kovacevic <marko.kovacevic@intel.com> Signed-off-by: Fiona Trahe <fiona.trahe@intel.com> Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Marko Kovacevic [Fri, 8 Jun 2018 09:10:01 +0000 (10:10 +0100)]
crypto/aesni_mb: support 3DES
Added support for 3DES cipher algorithm which
will support 8, 16 and 24 byte keys, which also has been
added in the v0.50 of the IPSec Multi-buffer lib.
Signed-off-by: Marko Kovacevic <marko.kovacevic@intel.com> Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Pablo de Lara [Tue, 3 Jul 2018 08:22:34 +0000 (09:22 +0100)]
crypto/aesni_mb: support IPsec Multi-buffer lib v0.50
Adds support for the v0.50 of the IPsec Multi-buffer lib.
The library now exposes its version, with the idea
of maintaining backwards compatibility in the future,
avoiding breaking the compilation of the PMD every time
there is a new version available.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Pablo de Lara [Tue, 3 Jul 2018 08:22:22 +0000 (09:22 +0100)]
crypto/aesni_gcm: support IPsec Multi-buffer lib v0.50
Adds support for the v0.50 of the IPsec Multi-buffer lib.
The library now exposes its version, with the idea
of maintaining backwards compatibility in the future,
avoiding breaking the compilation of the PMD every time
there is a new version available.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Pablo de Lara [Tue, 3 Jul 2018 05:58:01 +0000 (06:58 +0100)]
crypto/aesni_mb: call buffer manager allocation
Instead of having a static field for the buffer manager
MB_MGR in the queue pair structure, use the provided API
that allocates memory for it and store a pointer to it.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Some PMDs may have a limitation on the number of segments
in an SGL mbuf that they can process.
Thefore, an upper limit is set to avoid having
an indeterminate number of segments.
Signed-off-by: Fiona Trahe <fiona.trahe@intel.com> Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Now that device capabilities are checked separately,
before setting the xform parameters, it is not required
to do the check again, leaving only the xform setting
with the device configuration.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Fiona Trahe <fiona.trahe@intel.com>
Pablo de Lara [Thu, 19 Jul 2018 08:39:58 +0000 (09:39 +0100)]
examples/l2fwd-crypto: fix session mempool size
The session mempool size for this application depends
on the number of crypto devices that are capable
of performing the operation given by the parameters on the app.
However, previously this calculation was done before all devices
were checked, resulting in an incorrect number of sessions
required.
Now the calculation of the devices to be used is done first
(checking the capabilities of the enabled devices),
followed by the creation of the session pool, resulting
in a correct number of objects needed for the sessions
to be created.
Fixes: e3bcb99a5e13 ("examples/l2fwd-crypto: limit number of sessions") Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Fiona Trahe <fiona.trahe@intel.com>
Pablo de Lara [Thu, 19 Jul 2018 08:39:56 +0000 (09:39 +0100)]
examples/l2fwd-crypto: skip device not supporting operation
When a crypto device does not support an algorithm, it is skipped
and not used. However, when it does support it, but not the rest
of the parameters (IV, key, AAD sizes...), application stops.
Instead, the device should be skipped and the search of a suitable
device should continue.
Fixes: a061e50a0d97 ("examples/l2fwd-crypto: fix ambiguous input key size") Cc: stable@dpdk.org Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Fiona Trahe <fiona.trahe@intel.com>
Pablo de Lara [Mon, 16 Jul 2018 08:26:16 +0000 (09:26 +0100)]
examples/l2fwd-crypto: fix digest with AEAD algo
When performing authentication verification (both for AEAD algorithms,
such as AES-GCM, or for authentication algorithms, such as SHA1-HMAC),
the digest address is calculated based on the packet size and the
algorithm used (substracting digest size and IP header to the packet size).
However, for AEAD algorithms, this was not calculated correctly,
since the digest size was not being substracted.
Radu Nicolau [Mon, 16 Jul 2018 11:14:16 +0000 (12:14 +0100)]
app/crypto-perf: fix mempool creation
Using a small number of sessions results in rte_mempool_create call
with cache_size > n, which fails. There is no need to cache the elements,
as there is no performance impact.
Fixes: 501c0a3b14c3 ("app/crypto-perf: limit number of sessions") Signed-off-by: Radu Nicolau <radu.nicolau@intel.com> Acked-by: Akhil Goyal <akhil.goyal@nxp.com>
Pablo de Lara [Wed, 11 Jul 2018 06:38:46 +0000 (07:38 +0100)]
compress/isal: fix memory leak
Processed operations ring is created for each queue pair,
but it was not being freed when the queue pair was released.
Fixes: b0e23c458a6f ("compress/isal: add queue pair related ops") Cc: stable@dpdk.org Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Lee Daly <lee.daly@intel.com>
Pablo de Lara [Wed, 11 Jul 2018 06:38:45 +0000 (07:38 +0100)]
compress/isal: set null pointer after freeing
Fixes: b0e23c458a6f ("compress/isal: add queue pair related ops") Cc: stable@dpdk.org Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Lee Daly <lee.daly@intel.com>
Pablo de Lara [Wed, 11 Jul 2018 06:38:44 +0000 (07:38 +0100)]
compress/isal: fix log type name
There is a naming convention for logtypes of PMDs:
"pmd.driverType.driverName".
Therefore, the logtype for ISA-L PMD should be "pmd.compress.isal".
Fixes: 490e725b95b2 ("compress/isal: add device init and de-init") Cc: stable@dpdk.org Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com> Acked-by: Lee Daly <lee.daly@intel.com>
Lee Daly [Wed, 11 Jul 2018 14:21:45 +0000 (15:21 +0100)]
compress/isal: fix offset usage
This patch allows the ISA-L compression PMD,
to be used with offsets in the mbuf.
Offsets can now be used for source and destination buffers,
during compression or decompression.
Fixes: 7bf4f0630af6 ("compress/isal: add ISA-L decomp functionality") Fixes: dc49e6aa4879 ("compress/isal: add ISA-L compression functionality") Cc: stable@dpdk.org Signed-off-by: Lee Daly <lee.daly@intel.com> Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
QAT driver checks byte alignment for KASUMI/SNOW 3G/ZUC algorithms using
cipher/auth_param, which are not initialized at this moment yet. Use
operation params instead.
Fixes: 39e0bee48e81 ("crypto/qat: rework request builder for performance") Cc: stable@dpdk.org Reported-by: Dmitry Eremin-Solenikov <dmitry.ereminsolenikov@linaro.org> Signed-off-by: Fiona Trahe <fiona.trahe@intel.com> Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
For ESP transport and BYPASS mode the app might generate output
packets with invalid IPv4 header checksum.
At least such behavior was observed on few Intel NICs.
The reason is that the app didn't set ipv4 header checksum to zero
before passing it to the HW.
Fixes: 906257e965b7 ("examples/ipsec-secgw: support IPv6") Cc: stable@dpdk.org Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com> Acked-by: Radu Nicolau <radu.nicolau@intel.com>
Extend QAT guide to cover crypto and compression and common
information, particularly about kernel driver dependency.
Update release note.
Update compression feature list for qat.
git.droids-corp.org / dpdk.git / log