dpdk.git
9 years agoi40e: refactor xstats per queue handling
Harry van Haaren [Fri, 6 Nov 2015 14:12:55 +0000 (14:12 +0000)]
i40e: refactor xstats per queue handling

This patch refactors the queue and priority statistic handling.
Generic queue stats are presented by rte_eth_xstats_get(), and the
i40e_xstats_get() exposes only the extra stats.

Signed-off-by: Harry van Haaren <harry.van.haaren@intel.com>
Acked-by: Maryam Tahhan <maryam.tahhan@intel.com>
9 years agoixgbe: refactor xstats per queue handling
Harry van Haaren [Fri, 6 Nov 2015 14:12:54 +0000 (14:12 +0000)]
ixgbe: refactor xstats per queue handling

This patch refactors the queue handling. Generic queue stats are
handled by rte_eth_xstats_get() and the ixgbe_xstats_get() exposes
only the extra stats.

Signed-off-by: Harry van Haaren <harry.van.haaren@intel.com>
Acked-by: Maryam Tahhan <maryam.tahhan@intel.com>
9 years agoethdev: format generic xstats per queue always
Harry van Haaren [Fri, 6 Nov 2015 14:12:53 +0000 (14:12 +0000)]
ethdev: format generic xstats per queue always

This patch refactors the generic queue stats to be exposed
by rte_ethdev_xstats_get().

Signed-off-by: Harry van Haaren <harry.van.haaren@intel.com>
Acked-by: Maryam Tahhan <maryam.tahhan@intel.com>
9 years agoeal: support driver loading from directory
Panu Matilainen [Tue, 10 Nov 2015 14:28:15 +0000 (16:28 +0200)]
eal: support driver loading from directory

Add support for directories as arguments to -d for loading all drivers
from a given directory. Additionally a default driver directory can be
set in build-time configuration, in which case it will be always be used
when EAL is initialized.

This simplifies usage in shared library configuration significantly over
manually loading individual drivers with -d, and allows distros to
establish a drop-in driver directory for seamless integration
with 3rd party drivers etc.

Suggested-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Suggested-by: David Marchand <david.marchand@6wind.com>
Signed-off-by: Panu Matilainen <pmatilai@redhat.com>
Acked-by: David Marchand <david.marchand@6wind.com>
9 years agoeal: move plugin loading from linuxapp to common
Panu Matilainen [Tue, 10 Nov 2015 14:28:14 +0000 (16:28 +0200)]
eal: move plugin loading from linuxapp to common

There's no good reason to limit plugins to Linux, make it available
on FreeBSD too. Refactor the plugin code from Linux EAL to common
helper functions, also check for and fail on errors during initialization.

Signed-off-by: Panu Matilainen <pmatilai@redhat.com>
Acked-by: David Marchand <david.marchand@6wind.com>
9 years agomaintainers: replace for reorder library
Reshma Pattan [Wed, 21 Oct 2015 10:50:43 +0000 (11:50 +0100)]
maintainers: replace for reorder library

Updated maintainers list for reorder library

Signed-off-by: Reshma Pattan <reshma.pattan@intel.com>
Acked-by: Sergio Gonzalez Monroy <sergio.gonzalez.monroy@intel.com>
9 years agoreorder: fix unit test for random starting point
Reshma Pattan [Wed, 21 Oct 2015 10:50:44 +0000 (11:50 +0100)]
reorder: fix unit test for random starting point

The reorder library unit test was performed under the assumption that
the start sequence number was always 0.
This is not the case anymore as the start sequence number is initialized
by the first packet inserted into the reorder buffer.

This patch updates the unit test to reflect the new behavior.

Fixes: 7e1fa1de8a53 ("reorder: allow random number as starting point")

Reported-by: Mukesh Dua <mukesh.dua81@gmail.com>
Signed-off-by: Reshma Pattan <reshma.pattan@intel.com>
Acked-by: Sergio Gonzalez Monroy <sergio.gonzalez.monroy@intel.com>
9 years agodoc: add enic Tx improvement in release notes
John Daley [Fri, 6 Nov 2015 23:08:35 +0000 (15:08 -0800)]
doc: add enic Tx improvement in release notes

Signed-off-by: John Daley <johndale@cisco.com>
9 years agomaintainers: add maintainers for enic
John Daley [Thu, 5 Nov 2015 19:14:25 +0000 (11:14 -0800)]
maintainers: add maintainers for enic

Signed-off-by: John Daley <johndale@cisco.com>
Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
9 years agomaintainers: claim to be reviewer of virtio and vhost
Yuanhan Liu [Thu, 12 Nov 2015 04:10:40 +0000 (12:10 +0800)]
maintainers: claim to be reviewer of virtio and vhost

Firstly, Chuangchun's email address's been invalid for a while.

Secondly, I'd like to take the responsibility to review patches
of virtio/vhost component.

Cc: Huawei Xie <huawei.xie@intel.com>
Cc: Thomas Monjalon <thomas.monjalon@6wind.com>
Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
9 years agovhost: reset device properly
Yuanhan Liu [Thu, 12 Nov 2015 04:10:41 +0000 (12:10 +0800)]
vhost: reset device properly

Currently, we reset all fields of a device to zero when reset
happens, which is wrong, since for some fields like device_fh,
ifname, and virt_qp_nb, they should be same and be kept after
reset until the device is removed. And this is what's the new
helper function reset_device() for.

And use rte_zmalloc() instead of rte_malloc, so that we could
avoid init_device(), which basically dose zero reset only so far.
Hence, init_device() is dropped in this patch.

This patch also removes a hack of using the offset a specific
field (which is virtqueue now) inside of `virtio_net' structure
to do reset, which could be broken easily if someone changed the
field order without caution.

Cc: Tetsuya Mukawa <mukawa@igel.co.jp>
Cc: Huawei Xie <huawei.xie@intel.com>
Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Acked-by: Rich Lane <rich.lane@bigswitch.com>
9 years agovhost: make destroy callback on reset owner message
Rich Lane [Tue, 10 Nov 2015 02:15:13 +0000 (18:15 -0800)]
vhost: make destroy callback on reset owner message

QEMU sends VHOST_RESET_OWNER first when shutting down.
There was previously no way for the dataplane to know that the
virtio_net instance had become unusable and it would segfault
when trying to do RX/TX.

Signed-off-by: Rich Lane <rich.lane@bigswitch.com>
Acked-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
9 years agoi40e: fix VSI allocation for VMDq
Helin Zhang [Thu, 12 Nov 2015 07:09:03 +0000 (15:09 +0800)]
i40e: fix VSI allocation for VMDq

It fixes the issue of trying to allocate more VSIs for VMDq than
hardware remaining. It adds a check of the hardware remaining
before allocating VSIs for VMDq.

Fixes: c80707a0fd9c ("i40e: fix VMDq pool limit")

Signed-off-by: Helin Zhang <helin.zhang@intel.com>
9 years agobonding: fix build with icc
Tomasz Kulasek [Thu, 5 Nov 2015 10:55:07 +0000 (11:55 +0100)]
bonding: fix build with icc

ICC complains about enumerated types being mixed in link bonding driver,
as ETH_MQ_RX_RSS is an enum type of mq_mode and not a bitmask as it was
being treated.

Fixes: 734ce47f71e0 ("bonding: support RSS dynamic configuration")

Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
9 years agoixgbe: fix Tx hang when RS distance exceeds HW limit
Konstantin Ananyev [Tue, 10 Nov 2015 13:48:21 +0000 (13:48 +0000)]
ixgbe: fix Tx hang when RS distance exceeds HW limit

One of the ways to reproduce the issue:

testpmd <EAL-OPTIONS> -- -i --txqflags=0
testpmd> set fwd txonly
testpmd> set txpkts 64,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4
testpmd> set txsplit rand
testpmd> start

After some time TX on ixgbe queue will hang,
and all packet transmission on that queue will stop.

This bug was first reported and investigated by
Vlad Zolotarov <vladz@cloudius-systems.com>:
"We can reproduce this issue when stressed the xmit path with a lot of highly
fragmented TCP frames (packets with up to 33 fragments with non-headers
fragments as small as 4 bytes) with all offload features enabled."

The root cause is that ixgbe_xmit_pkts() in some cases violates the HW rule
that the distance between TDs with RS bit set should not exceed 40 TDs.

>From the latest 82599 spec update:
"When WTHRESH is set to zero, the software device driver should set the RS bit
in the Tx descriptors with the EOP bit set and at least once in the 40
descriptors."

The fix is to make sure that the distance between TDs with RS bit set
would never exceed HW limit.
As part of that fix, tx_rs_thresh for ixgbe PMD is not allowed to be greater
then to 32 to comply with HW restrictions.

With that fix slight slowdown for the full-featured ixgbe TX path
might be observed (from our testing - up to 4%).

ixgbe simple TX path is unaffected by that patch.

Reported-by: Vlad Zolotarov <vladz@cloudius-systems.com>
Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
9 years agoapp/testpmd: add ability to split outgoing packets
Konstantin Ananyev [Tue, 10 Nov 2015 13:48:20 +0000 (13:48 +0000)]
app/testpmd: add ability to split outgoing packets

For CSUM forwarding mode add ability to copy & split outgoing packet
into the new mbuf that consists of multiple segments.
For TXONLY and CSUM forwarding modes add ability to make number of
segments in the outgoing packet to vary on a per packet basis.
Number of segments and size of each segment is controlled by
'set txpkts' command.
Split policy is controlled by 'set txsplit' command.
Possible values are: on | off | rand.
Tha allows to increase test coverage for TX PMD codepaths.

Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
9 years agoi40e: fix stats reset
Harry van Haaren [Mon, 9 Nov 2015 10:22:22 +0000 (10:22 +0000)]
i40e: fix stats reset

This patch fixes a bug where only some of the statistics were being
reset when calling rte_eth_stats_reset() or rte_eth_xstats_reset().
As both the stats reset and xstats reset do the same, refactor away
the duplicated function.

This patch marks the VSI to update its offset, causing the stats be
look like they are reset.

Fixes: 9aace75fc82e ("i40e: fix statistics")

Signed-off-by: Harry van Haaren <harry.van.haaren@intel.com>
Acked-by: Maryam Tahhan <maryam.tahhan@intel.com>
9 years agoi40e: fix memzone freeing
Helin Zhang [Wed, 11 Nov 2015 05:28:34 +0000 (13:28 +0800)]
i40e: fix memzone freeing

This fixes the issue of not freeing memzone in a call to free the
memory for adminq DMA.

Fixes: 4861cde46116 ("i40e: new poll mode driver")

Signed-off-by: Helin Zhang <helin.zhang@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
9 years agoi40e: fix VMDq pool limit
Helin Zhang [Wed, 11 Nov 2015 08:56:45 +0000 (16:56 +0800)]
i40e: fix VMDq pool limit

It fixes the issue of cannot using more than 1 poor for VMDq,
according to the queues left.

Fixes: 705b57f82054 ("i40e: enlarge the number of supported queues")

Signed-off-by: Helin Zhang <helin.zhang@intel.com>
9 years agoi40e: fix build with icc 2015
Pablo de Lara [Tue, 10 Nov 2015 10:19:40 +0000 (10:19 +0000)]
i40e: fix build with icc 2015

Fix for the following error on icc 2015, due to incorrect type:
drivers/net/i40e/i40e_fdir.c(1376): error #188:
enumerated type mixed with another type
case RTE_ETH_FDIR_FILTER_INPUT_SET_SELECT:

Fixes: 98f05570 ("i40e: configure input fields for RSS or flow director")

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: Helin Zhang <helin.zhang@intel.com>
9 years agoi40e: fix build with icc 13
Helin Zhang [Wed, 11 Nov 2015 06:11:15 +0000 (14:11 +0800)]
i40e: fix build with icc 13

It fixes compile issue on ICC 13.0.0.

Error logs:
i40e_ethdev.c(7943): error #188: enumerated type mixed with another type

Fixes: c8b9a3e3fe1b ("i40e: support DCB mode")

Signed-off-by: Helin Zhang <helin.zhang@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
9 years agoapp/testpmd: fix build with icc 13
Helin Zhang [Wed, 11 Nov 2015 06:11:16 +0000 (14:11 +0800)]
app/testpmd: fix build with icc 13

It fixes compile issue on ICC 13.0.0.

Error logs:
app/test-pmd/cmdline.c(8160): error #188: enumerated type mixed
with another type
    entry.input.flow.tunnel_flow.tunnel_type =
        str2fdir_tunneltype(res->tunnel_type);

Fixes: 53b2bb9b7ea7 ("app/testpmd: new flow director commands")

Signed-off-by: Helin Zhang <helin.zhang@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
9 years agoapp/testpmd: fix icmp echo Tx queues
David Marchand [Thu, 29 Oct 2015 08:47:03 +0000 (09:47 +0100)]
app/testpmd: fix icmp echo Tx queues

icmp echo forward engine relies on the core rank rather than the number of
queues available.
This fix is still wrong, because if we set a nb-rxq > nb-txq, then we will
use an invalid txq, but with this, we are aligned on other fwd engines.

Signed-off-by: David Marchand <david.marchand@6wind.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
9 years agoigbvf: fix vlan filtering
Julien Meunier [Tue, 3 Nov 2015 14:25:22 +0000 (15:25 +0100)]
igbvf: fix vlan filtering

Once posted through mailbox, we must check for nack from the PF
and report an error in this case.

Signed-off-by: Julien Meunier <julien.meunier@6wind.com>
Signed-off-by: David Marchand <david.marchand@6wind.com>
Acked-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
9 years agoenic: fix vlan filtering
Julien Meunier [Fri, 30 Oct 2015 16:13:23 +0000 (17:13 +0100)]
enic: fix vlan filtering

Report an error when something went wrong.

Signed-off-by: Julien Meunier <julien.meunier@6wind.com>
Signed-off-by: David Marchand <david.marchand@6wind.com>
Acked-by: Sujith Sankar <ssujith@cisco.com>
9 years agoethdev: fix error handling in PCI fields copy
Bernard Iremonger [Fri, 6 Nov 2015 17:20:01 +0000 (17:20 +0000)]
ethdev: fix error handling in PCI fields copy

add return statement to if branch

Dereferencing NULL pointer reported by Coverity, CID 119256.
Dereferencing NULL pointer reported by Coverity, CID 119257.

Fixes: 494adb7f63f2 ("ethdev: add device fields from PCI layer")

Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
9 years agoapp/testpmd: fix flow director help and doc
Wenzhuo Lu [Fri, 6 Nov 2015 07:49:30 +0000 (15:49 +0800)]
app/testpmd: fix flow director help and doc

After implementing the fdir new modes for x550, the CLIs are modified.
Forgot to update the related help info and doc.

Fixes: 53b2bb9b7ea7 ("app/testpmd: new flow director commands")

Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
9 years agoring: fix memory freeing on error
John McNamara [Fri, 6 Nov 2015 17:28:37 +0000 (17:28 +0000)]
ring: fix memory freeing on error

Fix minor memory free issue in error clean-up.

Reported-by Coverity (CID 119258)

Fixes: 651c505af862 ("ring: enhance device setup from rings")

Signed-off-by: John McNamara <john.mcnamara@intel.com>
9 years agoapp/test: increase memory for hash unit tests
Pablo de Lara [Fri, 6 Nov 2015 16:02:54 +0000 (16:02 +0000)]
app/test: increase memory for hash unit tests

Hash_autotest is in test group 2 which had only 32 MB
of memory, which is increased to 64 MB to make it run.

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
9 years agoapp/test: fix unit test for option -n
Pablo de Lara [Fri, 6 Nov 2015 15:15:13 +0000 (15:15 +0000)]
app/test: fix unit test for option -n

eal_flags_autotest was broken after commit
19bfa4dd ("eal: make the -n argument optional"),
since the unit test was checking that app would not run
if -n flag was missing, which now it is possible.

Also, subtest test_missing_n_flag() has been renamed
to test_invalid_n_flag(), as now -n flag is not compulsory.

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
9 years agoversion: 2.2.0-rc1
Thomas Monjalon [Wed, 4 Nov 2015 23:48:18 +0000 (00:48 +0100)]
version: 2.2.0-rc1

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
9 years agodoc: fix release notes
Thomas Monjalon [Wed, 4 Nov 2015 23:40:39 +0000 (00:40 +0100)]
doc: fix release notes

A fix in hash library was in the drivers section.
Some ABI changes have been done without removing the notices.

Fixes: 62dbd2ffea9a ("ethdev: add more flow director modes")
Fixes: 381316f6a225 ("vhost-user: support protocol features")
Fixes: 5915699153d7 ("hash: fix scaling by reducing contention")

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
9 years agoigb_uio: fix build with kernel 4.3
David Hunt [Fri, 30 Oct 2015 12:14:17 +0000 (12:14 +0000)]
igb_uio: fix build with kernel 4.3

Fix to take this change into account: https://lkml.org/lkml/2015/7/9/101
Has been applied to Kernel 4.3.0-rc6

Linux: 4a7cc831 ("genirq/MSI: Move msi_list from struct pci_dev to struct device")

Signed-off-by: David Hunt <david.hunt@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
9 years agobonding: fix freeing with no queue
Raslsn Darawsheh [Mon, 26 Oct 2015 07:07:57 +0000 (09:07 +0200)]
bonding: fix freeing with no queue

In case of creating bond device without add any slaves and
quit from testpmd, application crashed since rx/tx queues
are NULL.

Add checking of this parameters before trying to free.

Signed-off-by: Raslsn Darawsheh <rdarawsheh@asaltech.com>
Signed-off-by: Yaacov Hazan <yaacovh@mellanox.com>
Acked-by: Declan Doherty <declan.doherty@intel.com>
9 years agobnx2x: fix probing driver for VF devices
Chas Williams [Mon, 2 Nov 2015 18:32:40 +0000 (13:32 -0500)]
bnx2x: fix probing driver for VF devices

Signed-off-by: Chas Williams <3chas3@gmail.com>
Acked-by: Harish Patil <harish.patil@qlogic.com>
9 years agoeal: do not reset getopt lib
Tiwei Bie [Mon, 19 Oct 2015 13:13:10 +0000 (21:13 +0800)]
eal: do not reset getopt lib

Someone may need to call rte_eal_init() with a fake argc/argv array
in the middle of using getopt() to parse its own unrelated argc/argv
parameters. So getopt lib shouldn't be reset by rte_eal_init().

Now eal will always save optind, optarg and optopt (and optreset on
FreeBSD) at the beginning, initialize optind (and optreset on FreeBSD)
to 1 before calling getopt_long(), then restore all values after.

Suggested-by: Don Provan <dprovan@bivio.net>
Suggested-by: Bruce Richardson <bruce.richardson@intel.com>
Signed-off-by: Tiwei Bie <btw@mail.ustc.edu.cn>
Reviewed-by: Don Provan <dprovan@bivio.net>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: David Marchand <david.marchand@6wind.com>
9 years agoethdev: fix link status race condition
Tim Shearer [Tue, 27 Oct 2015 21:38:55 +0000 (17:38 -0400)]
ethdev: fix link status race condition

Calling the Ethernet driver's link_update function from
rte_eth_dev_start can result in a race condition if the NIC raises
the link interrupt at the same time.
Depending on the interrupt handler implementation, the race can cause
the it to think that it received two consecutive link up interrupts,
and it exits without calling the user callback. Appears to impact
E1000/IGB and virtio drivers only.

Signed-off-by: Tim Shearer <tim.shearer@overturenetworks.com>
9 years agoethdev: clarify rate limit parameter
Simon Kagstrom [Tue, 20 Oct 2015 13:20:55 +0000 (15:20 +0200)]
ethdev: clarify rate limit parameter

The tx_rate unit is Mbps.

Gleaned from the ixgbe implementation, the 82599 datasheet and the use
in test-pmd.

Signed-off-by: Simon Kagstrom <simon.kagstrom@netinsight.net>
9 years agoixgbe: check for supported flow director behaviors
Andrey Chilikin [Wed, 30 Sep 2015 13:43:03 +0000 (14:43 +0100)]
ixgbe: check for supported flow director behaviors

Handle only supported flow director behaviors.

Signed-off-by: Andrey Chilikin <andrey.chilikin@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
9 years agoi40e: add flow director passthru
Andrey Chilikin [Wed, 30 Sep 2015 13:43:02 +0000 (14:43 +0100)]
i40e: add flow director passthru

Add support for RTE_ETH_FDIR_PASSTHRU flow director behavior so
output queue is assigned by other filters.

Signed-off-by: Andrey Chilikin <andrey.chilikin@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
9 years agoconfig: turn off fm10k driver on IBM POWER
Chao Zhu [Wed, 4 Nov 2015 06:15:26 +0000 (14:15 +0800)]
config: turn off fm10k driver on IBM POWER

The fm10k vector driver is specific for x86 platform which can't compile
on IBM POWER for lacking of tmmintrin.h header file. This patch turns
off fm10k driver compilation on IBM POWER to prevent compile issue.

Signed-off-by: Chao Zhu <chaozhu@linux.vnet.ibm.com>
Acked-by: Michael Qiu <michael.qiu@intel.com>
9 years agoeal/bsd: fix build
Bruce Richardson [Wed, 4 Nov 2015 17:05:57 +0000 (17:05 +0000)]
eal/bsd: fix build

The i40e driver was using a #define value for the max number of rxtx interrupts
supported. This value was defined only for linux, giving an error when compiling
on FreeBSD.

  CC i40e_ethdev.o
/usr/home/bruce/dpdk.org/drivers/net/i40e/i40e_ethdev.c:3885:9: fatal error: use of undeclared
        identifier 'RTE_MAX_RXTX_INTR_VEC_ID'

Copying the necessary #define into the FreeBSD EAL header fixes the compile
error.

Fixes: d37641029ada ("eal/linux: add interrupt vectors")

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
9 years agoixgbe: fix Rx errors statistics for UDP checksum
Harry van Haaren [Wed, 4 Nov 2015 16:14:43 +0000 (16:14 +0000)]
ixgbe: fix Rx errors statistics for UDP checksum

This patch removes l3_l4_xsum_errors from rx errors.

The reason to remove it is that UDP packets have an optional checksum, and
when not calculated the checksum field should be set to zero. When the
checksum is not calculated (zero-ed out), the hardware still counts a valid
UDP packet as an l3_l4_xsum_error.

This hardware issue is documented in 82599 errata, titled:
"Integrity Error Reported for IPv4/UDP Packets with Zero Checksum"

The solution is to remove l3_l4_xsum_errors from rx_errors, as discussed on
http://thread.gmane.org/gmane.comp.networking.dpdk.devel/25590/

Fixes: f6bf669b9900 ("ixgbe: account more Rx errors")

Suggested-by: Martin Weiser <martin.weiser@allegro-packets.com>
Signed-off-by: Harry van Haaren <harry.van.haaren@intel.com>
9 years agoethdev: check queue state before starting or stopping
Pablo de Lara [Wed, 16 Sep 2015 21:51:24 +0000 (22:51 +0100)]
ethdev: check queue state before starting or stopping

Following the same approach taken with dev_started field
in rte_eth_dev_data structure, this patch adds two new fields
in it, rx_queue_state and tx_queue_state arrays, which track
which queues have been started and which not.

This is important to avoid trying to start/stop twice a queue,
which will result in undefined behaviour
(which may cause RX/TX disruption).

Mind that only the PMDs which have queue_start/stop functions
have been changed to update this field, as the functions will
check the queue state before switching it.

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
9 years agoi40e: fix interrupt disabling when stopping
Jingjing Wu [Wed, 4 Nov 2015 15:01:25 +0000 (23:01 +0800)]
i40e: fix interrupt disabling when stopping

When stop device, should disable interrupt of FDIR but not
enable it. This patch corrected it.

Signed-off-by: Jingjing Wu <jingjing.wu@intel.com>
9 years agoi40evf: support Rx interrupt
Cunming Liang [Wed, 4 Nov 2015 08:45:40 +0000 (16:45 +0800)]
i40evf: support Rx interrupt

The patch enables rx interrupt support on i40e VF and some necessary
change on PF IOV mode to support VF.
On PF side, running in IOV mode via uio won't allow rx interrupt
which is exclusive with mbox interrupt in single vector competition.
On VF side, one single vector is shared for all the rx queues.

Signed-off-by: Cunming Liang <cunming.liang@intel.com>
9 years agoi40e: support Rx interrupt
Cunming Liang [Wed, 4 Nov 2015 08:45:39 +0000 (16:45 +0800)]
i40e: support Rx interrupt

The patch enables rx interrupt support on i40e PF non-IOV mode.
Per queue rx interrupt works on vfio, however on uio, all rx queues
share one interrupt vector.

Signed-off-by: Cunming Liang <cunming.liang@intel.com>
9 years agoixgbevf: cleanup unnecessary interrupt handler
Cunming Liang [Wed, 4 Nov 2015 08:45:37 +0000 (16:45 +0800)]
ixgbevf: cleanup unnecessary interrupt handler

As ixgbe vf doesn't support lsc, the patch removes those unused code.
In addition, it does some tiny cleanup.

Signed-off-by: Cunming Liang <cunming.liang@intel.com>
9 years agoixgbe: remove useless interrupt freeing when closing
Cunming Liang [Wed, 4 Nov 2015 08:45:36 +0000 (16:45 +0800)]
ixgbe: remove useless interrupt freeing when closing

The intr_vec is free in dev_stop. It's not necessary to check in dev_close.

Signed-off-by: Cunming Liang <cunming.liang@intel.com>
9 years agoixgbe: fix VF start with PF stopped
Cunming Liang [Wed, 4 Nov 2015 08:45:35 +0000 (16:45 +0800)]
ixgbe: fix VF start with PF stopped

When ixgbe runs as a PF, mbox interrupt is prerequisite to make VF
start normally.
And PF sometimes won't 'dev_start', so the mbox interrupt register
during 'dev_init' is required.
The patch rolls back the interrupt register for mbox,lsc to the 'dev_init'.
As UIO doesn't support multiple vector, mbox has to occupy the only one.
It adds condition check on 'dev_start', rxq interrupt is not allowed
when PF running in IOV mode via UIO.

Signed-off-by: Cunming Liang <cunming.liang@intel.com>
9 years agoigb: fix VF start with PF stopped
Cunming Liang [Wed, 4 Nov 2015 08:45:38 +0000 (16:45 +0800)]
igb: fix VF start with PF stopped

When igb runs as a PF, mbox interrupt is prerequisite to make VF
start normally.
And PF sometimes won't 'dev_start', so the mbox interrupt register
during 'dev_init' is required.
The patch rolls back the interrupt register for mbox,lsc to the 'dev_init'.
As UIO doesn't support multiple vector, mbox has to occupy the only one.
It adds condition check on 'dev_start', rxq interrupt is not allowed
when PF running in IOV mode via UIO.

Signed-off-by: Cunming Liang <cunming.liang@intel.com>
9 years agoeal: query multi-vector interrupt capability
Cunming Liang [Wed, 4 Nov 2015 08:45:34 +0000 (16:45 +0800)]
eal: query multi-vector interrupt capability

VFIO allows multiple MSI-X vector, others doesn't, but maybe will
allow it in the future.
Device drivers need to be aware of the capability.
It's better to avoid condition check on interrupt type (VFIO) everywhere,
instead a capability api is more flexible for the condition change.

Signed-off-by: Cunming Liang <cunming.liang@intel.com>
Acked-by: David Marchand <david.marchand@6wind.com>
9 years agoixgbe: avoid enabling interrupt with zero vector
Cunming Liang [Wed, 4 Nov 2015 08:45:32 +0000 (16:45 +0800)]
ixgbe: avoid enabling interrupt with zero vector

Signed-off-by: Cunming Liang <cunming.liang@intel.com>
9 years agoigb: avoid enabling interrupt with zero vector
Cunming Liang [Wed, 4 Nov 2015 08:45:33 +0000 (16:45 +0800)]
igb: avoid enabling interrupt with zero vector

Signed-off-by: Cunming Liang <cunming.liang@intel.com>
9 years agoeal/linux: forbid enabling zero interrupt vector
Cunming Liang [Wed, 4 Nov 2015 08:45:31 +0000 (16:45 +0800)]
eal/linux: forbid enabling zero interrupt vector

The patch adds condition check to avoid enable nothing.
In disable state, both max_intr and nb_efd are zero.

Signed-off-by: Cunming Liang <cunming.liang@intel.com>
Acked-by: David Marchand <david.marchand@6wind.com>
9 years agoixgbe: reserve VFIO vector zero for misc interrupt
Cunming Liang [Wed, 4 Nov 2015 08:45:29 +0000 (16:45 +0800)]
ixgbe: reserve VFIO vector zero for misc interrupt

According to the VFIO interrupt mapping, the interrupt vector id
for rxq starts from RX_VEC_START.
It doesn't impact the UIO cases.

Signed-off-by: Cunming Liang <cunming.liang@intel.com>
9 years agoigb: reserve VFIO vector zero for misc interrupt
Cunming Liang [Wed, 4 Nov 2015 08:45:30 +0000 (16:45 +0800)]
igb: reserve VFIO vector zero for misc interrupt

According to the VFIO interrupt mapping, the interrupt vector id
for rxq starts from RX_VEC_START.
It doesn't impact the UIO cases.

Signed-off-by: Cunming Liang <cunming.liang@intel.com>
9 years agoeal: reserve VFIO vector zero for misc interrupt
Cunming Liang [Wed, 4 Nov 2015 08:45:28 +0000 (16:45 +0800)]
eal: reserve VFIO vector zero for misc interrupt

During VFIO_DEVICE_SET_IRQS, the previous order is
{Q0_fd, ... Qn_fd, misc_fd}.
The vector number of misc is indeterminable which is
ugly to some NIC (e.g. i40e, fm10k).
The patch adjusts the order in {misc_fd, Q0_fd, ... Qn_fd},
always reserve the first vector to misc interrupt.

Signed-off-by: Cunming Liang <cunming.liang@intel.com>
Acked-by: David Marchand <david.marchand@6wind.com>
9 years agoapp/testpmd: extend commands for flow director in VF
Jingjing Wu [Wed, 4 Nov 2015 03:29:11 +0000 (11:29 +0800)]
app/testpmd: extend commands for flow director in VF

This patch extends commands to support filtering in VFs of flow director.

Signed-off-by: Jingjing Wu <jingjing.wu@intel.com>
9 years agoi40e: extend flow director to support VF
Jingjing Wu [Wed, 4 Nov 2015 03:29:10 +0000 (11:29 +0800)]
i40e: extend flow director to support VF

This patch extends flow director to filtering in VFs.

Signed-off-by: Jingjing Wu <jingjing.wu@intel.com>
9 years agoethdev: extend flow director to support VF
Jingjing Wu [Wed, 4 Nov 2015 03:29:09 +0000 (11:29 +0800)]
ethdev: extend flow director to support VF

This patch extends struct rte_eth_fdir_flow_ext to support flow
director in VFs.

Signed-off-by: Jingjing Wu <jingjing.wu@intel.com>
9 years agoi40e: fix statistics
Xutao Sun [Wed, 4 Nov 2015 09:20:48 +0000 (17:20 +0800)]
i40e: fix statistics

The old statistics on i40e only counted the packets on ports.
So the discarding packets on VSI were not counted.
This patch is to make statistics for packets both on ports and VSI.
Also update release notes.

Signed-off-by: Xutao Sun <xutao.sun@intel.com>
Acked-by: Harry van Haaren <harry.van.haaren@intel.com>
9 years agodoc: fix doxygen syntax of some comments
Thomas Monjalon [Wed, 4 Nov 2015 10:50:36 +0000 (11:50 +0100)]
doc: fix doxygen syntax of some comments

Some comments have a wrong space between /** and <.
Seen with
git grep '\*\* <'

Reported-by: David Marchand <david.marchand@6wind.com>
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
9 years agokni: fix igb build with kernel 4.2
Pablo de Lara [Mon, 12 Oct 2015 12:52:58 +0000 (13:52 +0100)]
kni: fix igb build with kernel 4.2

Kernel 4.2 has introduced two new parameters in ndo_bridge_getlink,
which breaks DPDK compilation.

Linux: 7d4f8d87 ("switchdev: ad VLAN support for ports bridge-getlink")

This patch adds the necessary checks to fix it.

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
9 years agokni: rename macro for igb nlflags
Pablo de Lara [Mon, 12 Oct 2015 12:52:57 +0000 (13:52 +0100)]
kni: rename macro for igb nlflags

Rename HAVE_NDO_BRIDGE_GETLINK_FILTER_MASK macro for
a more meaningful HAVE_NDO_BRIDGE_GETLINK_NLFLAGS,
as the macro is used to know if igb_ndo_bridge_getlink
function has nlflags parameter.

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
9 years agokni: allow per-net instances
Dex Chen [Tue, 27 Oct 2015 02:56:37 +0000 (10:56 +0800)]
kni: allow per-net instances

There is a global variable 'device_in_use' which is used to make sure
only one instance is using /dev/kni device. If you were using LXC, you
will find there is only one instance of KNI example could be run even
different namespaces were created.

In order to have /dev/kni used simultaneously in different namespaces,
making all of global variables as per network namespace variables.

With regard to single kernel thread mode, there will be one kernel
thread for each of network namespace.

Signed-off-by: Dex Chen <dex.chen@ruckuswireless.com>
Acked-by: Helin Zhang <helin.zhang@intel.com>
9 years agoeal: add option to delete hugepage backing files
Shesha Sreenivasamurthy [Wed, 28 Oct 2015 22:04:25 +0000 (15:04 -0700)]
eal: add option to delete hugepage backing files

When an application using huge-pages crash or exists, the hugetlbfs
backing files are not cleaned up. This is a patch to clean those files.
There are multi-process DPDK applications that may be benefited by those
backing files. Therefore, I have made that configurable so that the
application that does not need those backing files can remove them, thus
not changing the current default behavior. The application itself can
clean it up, however the rationale behind DPDK cleaning it up is, DPDK
created it and therefore, it is better it unlinks it.

Signed-off-by: Shesha Sreenivasamurthy <shesha@cisco.com>
Acked-by: Sergio Gonzalez Monroy <sergio.gonzalez.monroy@intel.com>
9 years agolpm: fix incorrect reuse of already allocated tbl8
Na Na [Tue, 3 Nov 2015 02:17:41 +0000 (10:17 +0800)]
lpm: fix incorrect reuse of already allocated tbl8

Fixes an initialization issue of 'valid_group' in the delete_depth_small().

When adding an entry to a tbl8, the .valid_group field should always be set,
so that future adds do not accidently find and use this table, thinking it is
currently invalid, i.e. unused, and thereby overwrite existing entries.

Signed-off-by: Na Na <nana.nn@alibaba-inc.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
9 years agolpm: fix condition check in delete
Na Na [Tue, 3 Nov 2015 02:17:40 +0000 (10:17 +0800)]
lpm: fix condition check in delete

Fixes an issue of check logic in delete_depth_small function.

For a tbl24 entry, the 'ext_entry' field indicates whether we need
to use tbl8_gindex to read the next_hop from a tbl8 entry, or whether
it can be read directly from this entry.

If a route is deleted, the prefix of previous route is used to override
the deleted route.

When checking the depth of the previous route the conditional checks
both the ext_entry and the depth, but the "else" leg fails to take
account that the condition could fail for one of two possible reasons,
leading to an incorrect flow when 'ext_entry == 0' is true,
but 'lpm->tbl24[i].depth > depth' is false.
The fix here is to add a condition check to the else leg so that it
only executes when ext_entry is set.

Signed-off-by: Na Na <nana.nn@alibaba-inc.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
9 years agohash: fix incorrect lookup if key is all zero
Pablo de Lara [Thu, 17 Sep 2015 10:30:48 +0000 (11:30 +0100)]
hash: fix incorrect lookup if key is all zero

If user has not added an all zero key in the hash table,
and tries to look it up, it results in an incorrect hit,
as dummy slot in the key table has all zero as well.

Fixes: 48a399119619 ("hash: replace with cuckoo hash implementation")

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
9 years agohash: fix scaling by reducing contention
Pablo de Lara [Fri, 30 Oct 2015 14:37:28 +0000 (14:37 +0000)]
hash: fix scaling by reducing contention

If using multiple cores on a system with hardware transactional
memory support, thread scaling does not work, as there was a single
point in the hash library which is a bottleneck for all threads,
which is the "free_slots" ring, which stores all the indices of
the free slots in the table.

This patch fixes the problem, by creating a local cache per logical core,
which stores locally indices of free slots,
so most times, writer threads will not interfere each other.

Fixes: 48a399119619 ("hash: replace with cuckoo hash implementation")

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
9 years agohash: free internal ring when freeing hash
Pablo de Lara [Fri, 2 Oct 2015 16:07:13 +0000 (17:07 +0100)]
hash: free internal ring when freeing hash

Since freeing a ring is now possible, then when freeing
a hash table, its internal ring can be freed as well.
Therefore when a new table, with the same name as a previously
freed table, is created, there is no need to look up
the already allocated ring.

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
9 years agoring: support freeing
Pablo de Lara [Fri, 2 Oct 2015 15:53:44 +0000 (16:53 +0100)]
ring: support freeing

When creating a ring, a memzone is created to allocate it in memory,
but the ring could not be freed, as memzones could not be.

Since memzones can be freed now, then rings can be as well,
taking into account if they were initialized using pre-allocated memory
(in which case, memory should be freed externally) or using rte_memzone_reserve
(with rte_ring_create), freeing the memory with rte_memzone_free.

Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
9 years agoapp/testpmd: add command to select filtering GRE key size
Helin Zhang [Tue, 3 Nov 2015 16:28:36 +0000 (00:28 +0800)]
app/testpmd: add command to select filtering GRE key size

Test commands are added to support selecting different length of GRE key.

Signed-off-by: Helin Zhang <helin.zhang@intel.com>
Signed-off-by: Andrey Chilikin <andrey.chilikin@intel.com>
9 years agoi40e: select GRE key length for filtering
Helin Zhang [Tue, 3 Nov 2015 16:28:35 +0000 (00:28 +0800)]
i40e: select GRE key length for filtering

By default, only first 3 bytes of GRE key will be used for hash or
FD calculation. With these changes, it can select 3 or 4 bytes of
GRE key for hash or FD calculation.

Signed-off-by: Helin Zhang <helin.zhang@intel.com>
Signed-off-by: Andrey Chilikin <andrey.chilikin@intel.com>
9 years agoapp/testpmd: add commands for input fields of RSS and flow director
Helin Zhang [Tue, 3 Nov 2015 16:07:55 +0000 (00:07 +0800)]
app/testpmd: add commands for input fields of RSS and flow director

Test commands are added to support selecting input set, or adding
new input set per different pctype.

Signed-off-by: Helin Zhang <helin.zhang@intel.com>
Signed-off-by: Andrey Chilikin <andrey.chilikin@intel.com>
9 years agoi40e: configure input fields for RSS or flow director
Helin Zhang [Tue, 3 Nov 2015 16:07:54 +0000 (00:07 +0800)]
i40e: configure input fields for RSS or flow director

The default input set of fields of a received packet are loaded from
firmware, which cannot be modified even users want to use different
fields for RSS or flow director. Here adds more flexibilities of
selecting packet fields for hash calculation or flow director for
users.

Signed-off-by: Helin Zhang <helin.zhang@intel.com>
Signed-off-by: Andrey Chilikin <andrey.chilikin@intel.com>
9 years agoi40e: enlarge the number of supported queues
Helin Zhang [Tue, 3 Nov 2015 15:40:28 +0000 (23:40 +0800)]
i40e: enlarge the number of supported queues

It enlarges the number of supported queues to hardware allowed
maximum. There was a software limitation of 64 per physical port
which is not reasonable.

Signed-off-by: Helin Zhang <helin.zhang@intel.com>
9 years agoi40e: adjust the number of queues for RSS
Helin Zhang [Tue, 3 Nov 2015 15:40:27 +0000 (23:40 +0800)]
i40e: adjust the number of queues for RSS

It adjusts the number of queues for RSS from power of 2 to any as
long as it does not exceeds the hardware allowed.

Signed-off-by: Helin Zhang <helin.zhang@intel.com>
9 years agomlx5: fix hash Rx queue type in RSS mode
Adrien Mazarguil [Tue, 3 Nov 2015 17:15:14 +0000 (18:15 +0100)]
mlx5: fix hash Rx queue type in RSS mode

Only seen since IPv6 RSS support was added, confusion about the purpose of
the hash_rxq_type_from_n() function has caused it to return invalid hash RX
queue types.

Refactor function for its intended purpose, rename it
hash_rxq_type_from_pos() and update comment with a better description.

Fixes: a76133214d88 ("mlx5: use separate indirection table for default hash Rx queue")

Reported-by: Olga Shern <olgas@mellanox.com>
Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
9 years agomlx5: fix build with debugging enabled
Adrien Mazarguil [Tue, 3 Nov 2015 17:15:13 +0000 (18:15 +0100)]
mlx5: fix build with debugging enabled

The following error occurs when CONFIG_RTE_LIBRTE_MLX5_DEBUG=y:

 drivers/net/mlx5/mlx5.c:381:4: error: ISO C forbids braced-groups within expressions

RTE_MIN() uses the non-standard ({ ... }) syntax to declare variables within
parentheses, which is rejected by -pedantic.

Since the RSS_INDIRECTION_TABLE_SIZE check is meant to go away as soon as
DPDK supports larger/variable indirection tables, put it in a separate
condition.

Fixes: 634efbc2c8c0 ("mlx5: support RETA query and update")

Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
9 years agovmxnet3: support port hotplug
Bernard Iremonger [Tue, 27 Oct 2015 17:12:26 +0000 (17:12 +0000)]
vmxnet3: support port hotplug

Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
Acked-by: Yong Wang <yongwang@vmware.com>
9 years agobonding: remove fake pci interface
Bernard Iremonger [Tue, 3 Nov 2015 13:02:05 +0000 (13:02 +0000)]
bonding: remove fake pci interface

remove pci_dev, pci_drv, rte_bond_pmd and pci_id_table.
handle numa_node for vdevs
handle RTE_ETH_DEV_INTR_LSC for vdevs
rename the function valid_bonded_device to check_for_bonded_device
remove branches on pci_dev

Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
Acked-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
9 years agompipe: remove fake pci interface
Bernard Iremonger [Tue, 3 Nov 2015 13:02:04 +0000 (13:02 +0000)]
mpipe: remove fake pci interface

Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
9 years agoxenvirt: remove fake pci interface
Bernard Iremonger [Tue, 3 Nov 2015 13:02:03 +0000 (13:02 +0000)]
xenvirt: remove fake pci interface

Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
9 years agoaf_packet: remove fake pci interface
Bernard Iremonger [Tue, 3 Nov 2015 13:02:02 +0000 (13:02 +0000)]
af_packet: remove fake pci interface

Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: John W. Linville <linville@tuxdriver.com>
9 years agopcap: remove fake pci interface
Bernard Iremonger [Tue, 3 Nov 2015 13:02:01 +0000 (13:02 +0000)]
pcap: remove fake pci interface

remove rte_pcap_pmd and pci_dev.

Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
9 years agoring: remove fake pci interface
Bernard Iremonger [Tue, 3 Nov 2015 13:02:00 +0000 (13:02 +0000)]
ring: remove fake pci interface

remove rte_ring_pmd and pci_dev.

Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
9 years agonull: remove fake pci interface
Bernard Iremonger [Tue, 3 Nov 2015 13:01:59 +0000 (13:01 +0000)]
null: remove fake pci interface

remove rte_null_pmd and pci_dev.

Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
9 years agoethdev: remove some PCI specific handling
Bernard Iremonger [Tue, 3 Nov 2015 13:01:58 +0000 (13:01 +0000)]
ethdev: remove some PCI specific handling

use dev_type to distinguish between vdev's and pdev's.
remove pci_dev branches.

Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
9 years agodrivers: copy fake PCI device info to ethdev data
Bernard Iremonger [Tue, 3 Nov 2015 13:01:57 +0000 (13:01 +0000)]
drivers: copy fake PCI device info to ethdev data

Initialise dev_flags, driver, kdrv, drv_name and numa_node fields
in eth_dev data.

for the following vdevs:
null
ring
pcap
af_packet
xenvirt
mpipe
bonding

Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
9 years agodrivers: copy PCI device info to ethdev data
Bernard Iremonger [Tue, 3 Nov 2015 13:01:56 +0000 (13:01 +0000)]
drivers: copy PCI device info to ethdev data

Use new function rte_eth_copy_pci_info.
Copy device info for the following pdevs:

bnx2x
cxgbe
e1000
enic
fm10k
i40e
ixgbe
mlx4
mlx5
virtio
vmxnet3

Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
9 years agoethdev: add device fields from PCI layer
Bernard Iremonger [Tue, 3 Nov 2015 13:01:55 +0000 (13:01 +0000)]
ethdev: add device fields from PCI layer

The driver fields have been added the rte_eth_dev_data so that it is no
longer necessary access this data through the pci_dev pointer.

The following fields have been added to rte_eth_dev_data:

dev_flags, and macros for dev_flags.
kdrv
numa_node
drv_name

Add function rte_eth_copy_pci_info

Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
[Thomas: remove useless flags]

9 years agoeal: set name to threads
Ravi Kerur [Wed, 29 Jul 2015 00:51:44 +0000 (17:51 -0700)]
eal: set name to threads

This patch adds support for pthread_setname_np on Linux and
pthread_set_name_np on FreeBSD.

Signed-off-by: Ravi Kerur <rkerur@gmail.com>
Acked-by: Stephen Hemminger <stephen@networkplumber.org>
[Thomas: add name in tep_termination example]

9 years agoethdev: clean port id retrieval when attaching
Ravi Kerur [Wed, 23 Sep 2015 21:16:17 +0000 (14:16 -0700)]
ethdev: clean port id retrieval when attaching

Removed following functions
   > rte_eth_dev_save and
   > rte_eth_dev_get_changed_port

Added 2 new functions
   > rte_eth_dev_get_port_by_name
   > rte_eth_dev_get_port_by_addr

Compiled on Linux for following targets
   > x86_64-native-linuxapp-gcc
   > x86_64-native-linuxapp-clang
   > x86_x32-native-linuxapp-gcc

Compiled on FreeBSD for following targets
   > x86_64-native-bsdapp-clang
   > x86_64-native-bsdapp-gcc

Tested on Linux/FreeBSD:
   > port attach eth_null
   > port start all
   > port stop all
   > port close all
   > port detach 0
   > port attach eth_null
   > port start all
   > port stop all
   > port close all
   > port detach 0

Successful run of checkpatch.pl on the diffs

Successful validate_abi on Linux for following targets

   > x86_64-native-linuxapp-gcc
   > x86_64-native-linuxapp-clang

Signed-off-by: Ravi Kerur <rkerur@gmail.com>
Acked-by: Tetsuya Mukawa <mukawa@igel.co.jp>
9 years agovhost: fix build with old kernels
Marcel Apfelbaum [Tue, 3 Nov 2015 11:28:41 +0000 (13:28 +0200)]
vhost: fix build with old kernels

Commit 15e9ee6982a4822ce395fd597dd500a61ceafa7c
uses the VIRTIO_F_VERSION_1 macro existing only in newer kernels.

Fixed it by manually defining it for older kernels.

Fixes: 15e9ee6982a4 ("vhost: enable virtio 1.0")

Reported-by: Qian Xu <qian.q.xu@intel.com>
Signed-off-by: Marcel Apfelbaum <marcel@redhat.com>
9 years agovirtio: fix size of MAC address array
Ivan Boule [Thu, 29 Oct 2015 08:46:15 +0000 (09:46 +0100)]
virtio: fix size of MAC address array

Make the virtio PMD allocate the array of unicast MAC addresses with
the maximum of entries (VIRTIO_MAX_MAC_ADDRS) that it exports.

Signed-off-by: Ivan Boule <ivan.boule@6wind.com>
Signed-off-by: David Marchand <david.marchand@6wind.com>
Reviewed-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
9 years agoixgbe: remove useless fields in checksum offload
Didier Pallard [Thu, 29 Oct 2015 08:47:53 +0000 (09:47 +0100)]
ixgbe: remove useless fields in checksum offload

According to Table 7-38: Valid Fields by Offload Option
of Intel ® 82599 10 GbE Controller Datasheet,
L4LEN field is not needed for L4 XSUM computation by the hardware.
So remove l4_len from tx_offload_mask in ixgbe_set_xmit_ctx
function used to build the context transmitted to the hardware.

Signed-off-by: Didier Pallard <didier.pallard@6wind.com>
Signed-off-by: David Marchand <david.marchand@6wind.com>
Acked-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
9 years agomlx5: support RETA query and update
Nelio Laranjeiro [Mon, 2 Nov 2015 18:11:57 +0000 (19:11 +0100)]
mlx5: support RETA query and update

ConnectX-4 adapters do not have a constant indirection table size, which is
set at runtime from the number of RX queues. The maximum size is retrieved
using a hardware query and is normally 512.

Since the current RETA API cannot handle a variable size, any query/update
command causes it to be silently updated to RSS_INDIRECTION_TABLE_SIZE
entries regardless of the original size.

Also due to the underlying type of the configuration structure, the maximum
size is limited to RSS_INDIRECTION_TABLE_SIZE (currently 128, at most 256
entries).

A port stop/start must be done to apply the new RETA configuration.

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
9 years agoi40evf: support AQ based RSS config
Helin Zhang [Tue, 13 Oct 2015 07:19:50 +0000 (15:19 +0800)]
i40evf: support AQ based RSS config

It supports both Admin queue based and directly writing registers
based RSS hash key and lookup table configuration, as X722 supports
AQ based configuration.

Signed-off-by: Helin Zhang <helin.zhang@intel.com>
Acked-by: Wenzhuo Lu <wenzhuo.lu@intel.com>