Remove RTE_LIBRTE_SFC_EFX_TSO config option since it is not
required any more:
- unreasonable limit on number of Tx queues when TSO is not
actually required should be solved using per-device parameter
- performance difference with and without TSO compiled in is small
Signed-off-by: Andrew Rybchenko <arybchenko@solarflare.com>
Alarms are not supported on the FreeBSD.
Application must poll link status periodically itself using
rte_eth_link_get_nowait() to avoid management event queue overflow.
Fixes: 2de39f4e13 ("net/sfc: periodic management EVQ polling using alarm")
Signed-off-by: Andrew Rybchenko <arybchenko@solarflare.com>
Reviewed-by: Andrew Lee <alee@solarflare.com>
Reviewed-by: Andy Moreton <amoreton@solarflare.com>
When the WQ is wrapped around, it wrongly checks the condition when
resetting the pointer. It should be compared against the end of the queue,
not the beginning of the queue. And this isn't even needed when the length
of the copying data crosses the boundary.
Fixes: fdcb0f5305 ("net/mlx5: use work queue buffer as a raw buffer")
Cc: stable@dpdk.org
Signed-off-by: Yongseok Koh <yskoh@mellanox.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
The size of Rx RSS indirection table was limited by 256, but it is not
required anymore for all Mellanox NICs. However, the librte_ether still
limits the size by 512.
Signed-off-by: Yongseok Koh <yskoh@mellanox.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
On receiving a compressed session of Rx completion, prefetch every entries
to be invalidated. Also, invalidate consumed completions per every 8
mini-completions, not to wait until the last entry is consumed. This helps
to reduce jitter in rx_burst.
Signed-off-by: Yongseok Koh <yskoh@mellanox.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
When no error reported in Rx descriptor, we should set CKSUM_GOOD flag
before return.
Fixes: b704f9071b ("net/i40e: implement new Rx checksum flag")
Cc: stable@dpdk.org
Signed-off-by: Xiao Wang <xiao.w.wang@intel.com>
While handling link status change (LSC) interrupt, all interrupts are
blocked until delayed interrupt handler finishes.
The wait duration is at least one second and this may cause timeouts in
VF to PF mailbox.
Make sure only LSC interrupt is blocked while waiting for delayed
interrupt handler to finish.
Fixes: 0a45657a67 ("pci: rework interrupt handling")
Cc: stable@dpdk.org
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
efx_phy_adv_cap_set() sets all advertised phy capabilities including
pause capabilities which are also configured using efx_mac_fcntl_set().
If we set speed and autonegotiation capabilities only, we should
preserve already configured pause capabilities.
Fixes: d23f3a89ab ("net/sfc: support link speed and duplex settings")
Signed-off-by: Andrew Rybchenko <arybchenko@solarflare.com>
Reviewed-by: Andrew Lee <alee@solarflare.com>
Fixes: 886f8d8a05 ("net/sfc: retrieve link info")
Signed-off-by: Andrew Rybchenko <arybchenko@solarflare.com>
Reviewed-by: Andrew Lee <alee@solarflare.com>
Fixes: 886f8d8a05 ("net/sfc: retrieve link info")
Signed-off-by: Andrew Rybchenko <arybchenko@solarflare.com>
Reviewed-by: Andrew Lee <alee@solarflare.com>
In fact efx_port_poll() always initializes it, but it isn't
explicitly documented feature of the API. Moreover, the API
annocation suggests that return code should be checked.
Fixes: 886f8d8a05 ("net/sfc: retrieve link info")
Signed-off-by: Andrew Rybchenko <arybchenko@solarflare.com>
Reviewed-by: Andrew Lee <alee@solarflare.com>
When any layout is used, the header is stored in the head room of mbuf.
mbuf is allocated and filled by user, means there is no gurateen the
header is all zero for non TSO case. Therefore, we have to do the reset
by ourself:
memest(hdr, 0, head_size);
The memset has two impacts on performance:
- memset could not be inlined, which is a bit costly.
- more importantly, it touches the mbuf, which could introduce severe
cache issues as described by former patch.
Similiary, we could do the same trick: reset just when necessary, when
the corresponding field is already 0, which is likely true for a simple
l2 forward case. It could boost the performance up to 20+% in micro
benchmarking.
Cc: stable@dpdk.org
Cc: Maxime Coquelin <maxime.coquelin@redhat.com>
Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
TSO is now enabled, but it's not actually being used by default in a
simple L2 forward mode. In such case, we have to zero the virtio net
headers, to inform the vhost backend that no offload is being used:
hdr->csum_start = 0;
hdr->csum_offset = 0;
hdr->flags = 0;
hdr->gso_type = 0;
hdr->gso_size = 0;
hdr->hdr_len = 0;
Such writes could be very costly; it introduces severe cache issues:
The above operations introduce cache write for each packet, which
stalls the read operation from the vhost backend.
The fact that virtio net header is initiated to zero in PMD driver
init stage means that these costly writes are unnecessary and could
be avoided:
if (hdr->csum_start != 0)
hdr->csum_start = 0;
And that's what the macro ASSIGN_UNLESS_EQUAL does. With this, the
performance drop introduced by TSO enabling is recovered: it could
be up to 20% in micro benchmarking.
Fixes: 58169a9c81 ("net/virtio: support Tx checksum offload")
Fixes: 696573046e ("net/virtio: support TSO")
Cc: stable@dpdk.org
Cc: Olivier Matz <olivier.matz@6wind.com>
Cc: Maxime Coquelin <maxime.coquelin@redhat.com>
Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Reviewed-by: Olivier Matz <olivier.matz@6wind.com>
The commit aed0b12930 ("net/vhost: fix socket file deleted on stop")
moves rte_vhost_driver_register and rte_vhost_driver_unregister from
dev_start() and dev_stop() into driver's probe() and remove().
Apps, like testpmd, using vhost pmd in server mode, usually calls
dev_stop() and dev_close() as quitting, instead of driver-specific
remove(). Then those unix socket files have no chance to get removed.
Semantically, device-specific things should be put into device-specific
APIs. Fix this issue by moving rte_vhost_driver_unregister, plus other
structure free into dev_close().
Fixes: aed0b12930 ("net/vhost: fix socket file deleted on stop")
Cc: stable@dpdk.org
Reported-by: Lei Yao <lei.a.yao@intel.com>
Signed-off-by: Jianfeng Tan <jianfeng.tan@intel.com>
Acked-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Found with clang static analysis:
lib/librte_vhost/vhost_user.c:996:3: warning:
Value stored to 'ret' is never read
ret = vhost_user_get_vring_base(dev, &msg.payload.state);
^ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Signed-off-by: Emmanuel Roullit <emmanuel.roullit@gmail.com>
Acked-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Found with clang static analysis:
lib/librte_vhost/virtio_net.c:723:17: warning:
Access to field 'data_off' results in a dereference of a null pointer
(loaded from variable 'tcp_hdr')
m->l4_len = (tcp_hdr->data_off & 0xf0) >> 2;
^~~~~~~~~~~~~~~~~
Fixes: d0cf91303d ("vhost: add Tx offload capabilities")
Cc: stable@dpdk.org
Signed-off-by: Emmanuel Roullit <emmanuel.roullit@gmail.com>
Acked-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Value returned from malloc is not checked for errors before being used.
This patch fixes following coverity issue.
static struct vhost_memory_kernel *
prepare_vhost_memory_kernel(void)
{
...
vm = malloc(sizeof(struct vhost_memory_kernel) +
max_regions *
sizeof(struct vhost_memory_region));
...
>>> CID 140744: (NULL_RETURNS)
>>> Dereferencing a null pointer "vm".
mr = &vm->regions[k++];
Coverity issue: 140744
Fixes: e3b434818b ("net/virtio-user: support kernel vhost")
Signed-off-by: Jianfeng Tan <jianfeng.tan@intel.com>
Acked-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
The vtpci_ops assignment needs the 'hw->port_id' as an input parameter.
That said, we should set 'hw->port_id' firstly, then do the vtpci_ops
assignment, while the code does reversely. That would result to a crash
when more than one virtio devices are used, because we keep assigning
proper vtpci_ops to virtio_hw_internal[0]->vtpci_ops, leaving the pointer
for other ports being NULL.
Reverse the order fixes this issue.
Fixes: 9470427c88 ("net/virtio: do not store PCI device pointer at shared memory")
Cc: stable@dpdk.org
Reported-by: Lei Yao <lei.a.yao@intel.com>
Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Setting up the mapping from GPA (guest physical address) to HPA (guest
physical address) could be very time consuming when the guest memory is
backened with small pages (4K). The bigger the guest memory, the longer
it takes. This could lead a very long vhost-user negotiation.
Since the mapping is only needed in zero copy mode so far, we could
avoid such time consuming settup when zero copy is turned off (which is
the default case).
It's actually a workaround, a right fix might be to start a new thread,
and hide the big latency there.
Fixes: e246896178 ("vhost: get guest/host physical address mappings")
Cc: stable@dpdk.org
Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
If a malicious guest forges a dead loop desc chain (let desc->next point
to itself) and desc->len is zero, this could lead to a dead loop in
copy_mbuf_to_desc(following is a simplified code to show this issue
clearly):
while (mbuf_is_not_totally_consumed) {
if (desc_avail == 0) {
desc = &descs[desc->next];
desc_avail = desc->len;
}
COPY(desc, mbuf, desc_avail);
}
I have actually fixed a same issue before: commit a436f53ebf ("vhost:
avoid dead loop chain"); it fixes the dequeue path though, leaving the
enqueue path still vulnerable.
The fix is the same. Add a var nr_desc to avoid the dead loop.
Fixes: f1a519ad98 ("vhost: fix enqueue/dequeue to handle chained vring descriptors")
Cc: stable@dpdk.org
Reported-by: Xieming Katty <katty.xieming@huawei.com>
Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Add documentation to describe using the new performance test application.
Signed-off-by: Slawomir Mrozowicz <slawomirx.mrozowicz@intel.com>
Signed-off-by: Piotr Azarewicz <piotrx.t.azarewicz@intel.com>
This patchset introduce new application which allows measuring
performance parameters of PMDs available in crypto tree. The goal of
this application is to replace existing performance tests in app/test.
Parameters available are: throughput (--ptest throughput) and latency
(--ptest latency). User can use multiply cores to run tests on but only
one type of crypto PMD can be measured during single application
execution. Cipher parameters, type of device, type of operation and
chain mode have to be specified in the command line as application
parameters. These parameters are checked using device capabilities
structure.
Couple of new library functions in librte_cryptodev are introduced for
application use.
To build the application a CONFIG_RTE_APP_CRYPTO_PERF flag has to be set
(it is set by default).
Example of usage: -c 0xc0 --vdev crypto_aesni_mb_pmd -w 0000:00:00.0 --
--ptest throughput --devtype crypto_aesni_mb --optype cipher-then-auth
--cipher-algo aes-cbc --cipher-op encrypt --cipher-key-sz 16 --auth-algo
sha1-hmac --auth-op generate --auth-key-sz 64 --auth-digest-sz 12
--total-ops 10000000 --burst-sz 32 --buffer-sz 64
Signed-off-by: Declan Doherty <declan.doherty@intel.com>
Signed-off-by: Slawomir Mrozowicz <slawomirx.mrozowicz@intel.com>
Signed-off-by: Piotr Azarewicz <piotrx.t.azarewicz@intel.com>
Signed-off-by: Marcin Kerlin <marcinx.kerlin@intel.com>
Signed-off-by: Michal Kobylinski <michalx.kobylinski@intel.com>
This patch adds helper functions for new performance application which
provide identifiers and number of crypto device and
provide and check capabilities available for defined device and algorithm.
The performance application can be used to measure throughput and latency
of cryptography operation performed by crypto device.
Signed-off-by: Declan Doherty <declan.doherty@intel.com>
Signed-off-by: Slawomir Mrozowicz <slawomirx.mrozowicz@intel.com>
Signed-off-by: Marcin Kerlin <marcinx.kerlin@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Adds the description of the cryptodev scheduler PMD overview,
limitations, build, instructions, modes, etc.
Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Same as other cryptodev PMDs, it is necessary to carry out the unit
test for scheduler PMD. Currently the test is designed to attach 2
AESNI-MB cryptodev PMDs as slaves, sets the scheduling mode as round-
robin, and runs almost all AESNI-MB test items (except for sessionless
tests). In the end, the slaves are detached.
Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Adds Makefile for scheduler cryptodev PMD, and updates existing
Makefiles. Different than other cryptodev PMDs, scheduler PMD
is required to be built as shared libraries.
Adds scheduler PMD enable and debug flags to config/common_base.
Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Implements all standard operations required for cryptodev,
and register them to cryptodev operation function pointer table.
Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Adds crypto scheduler's PMD's probe and remove function and the device's
enqueue and dequeue burst functions. A cryptodev scheduler PMD is
then registered in the end.
Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Implements round-robin scheduling mode and register into cryptodev
scheduler ops structure. This mode enqueues a burst of operation
to one of its slaves, and iterates the next burst to the other
slave. Same procedure is done on dequeueing operations.
Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
Signed-off-by: Declan Doherty <declan.doherty@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Adds the implementations of the APIs for scheduler cryptodev PMD.
Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Adds a number of internal structures for the cryptodev scheduler PMD. The
structures include the scheduler context, slave, queue pair context,
and session.
Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
Signed-off-by: Declan Doherty <declan.doherty@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Adds APIs and function prototypes for the scheduler PMD to perform extra
operations other than standard cryptodev APIs.
Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
Signed-off-by: Declan Doherty <declan.doherty@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
This patch adds the cryptodev scheduler PMD name and type identifier to
librte_cryptodev.
Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
This makes struct rte_cryptodev independent of struct rte_pci_device by
replacing it with a pointer to the generic struct rte_device.
This is inline with the recent changes in ethdev
Signed-off-by: Hemant Agrawal <hemant.agrawal@nxp.com>
Acked-by: John Griffin <john.griffin@intel.com>
Reviewed-by: Shreyansh Jain <shreyansh.jain@nxp.com>
rte_cryptodev_pmd_get_dev, rte_cryptodev_pmd_get_named_dev,
rte_cryptodev_pmd_is_valid_dev were incorrectly marked as inline and
therefore not useable from crypto PMDs when built as shared
libraries as they accessed the global rte_cryptodev_globals device
structure.
Fixes: d11b0f30 ("cryptodev: introduce API and framework for crypto devices")
Signed-off-by: Declan Doherty <declan.doherty@intel.com>
Acked-by: Fan Zhang <roy.fan.zhang@intel.com>
make config dependency resolving was always running serial,
parallelize it for better performance.
$ time make T=x86_64-native-linuxapp-gcc config
real 0m12.633s
$ time make -j8 T=x86_64-native-linuxapp-gcc config
real 0m1.826s
When config creation done under a single make target, using a for loop,
make has no control on the action, and it needs to run as implemented in
the rule. But if for loop converted into multiple targets, make can
detect independent targets and run them parallel based on -j parameter.
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
When adding crypto devices, the "Active" and "Ssh_if" attributes of
existing network devices were reset. This causes the following issues:
- Network interfaces aren't marked as "*Active*" in the --status output.
- Active network interfaces can be unbound without the --force option,
causing loss of network connection.
The reset was caused by the call to devices[d].update in
get_crypto_details.
This patch prevents the update on non-crypto devices.
Fixes: cb4a1d14bf ("tools: bind crypto devices")
Cc: stable@dpdk.org
Signed-off-by: Yoni Gilad <yonig@radcom.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
To avoid confusion with distributor app, this commit
renames the flow-distributor sample app to server_node_efd,
since it shows how to use the EFD library and it is based
on a server/nodes model.
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
When RTE_EFD_VALUE_NUM_BITS is 32, there was a compilation issue
because of an overflow:
app/test/test_efd.c:157:55: error: overflow in expression;
result is 2147483647 with type 'int' [-Werror,-Winteger-overflow]
data[0] = mrand48() & ((1 << RTE_EFD_VALUE_NUM_BITS) - 1);
This commit fixes the issue by using a setting a different
macro VALUE_BITMASK with a conditional
Fixes: 0e925aef27 ("app/test: add EFD functional and perf tests")
Reported-by: Yong Liu <yong.liu@intel.com>
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Fix GCC 4.8.2 20140120 (Red Hat 4.8.2-16) (RHEL 7.0) false warning
when build with EXTRA_CFLAGS='--coverage'.
Fixes: 278f945402 ("pdump: add new library for packet capture")
Signed-off-by: Andrew Rybchenko <arybchenko@solarflare.com>
Add two new commands "show port cap <port>" and "show
port cap all" to display what offload capabilities supported
in ports. It will not only display all the capabilities of
the port, but also the enabling condition for each capability
in the running time.
Signed-off-by: Qiming Yang <qiming.yang@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
Acked-by: Beilei Xing <beilei.xing@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
This enables ACL matches to return 0 where the distinction
from no-match case is not needed.
Signed-off-by: Michał Mirosław <michal.miroslaw@atendesoftware.pl>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
