dpdk.git
6 years agonet/mlx4: mitigate Tx path memory barriers
Matan Azrad [Thu, 2 Nov 2017 16:42:51 +0000 (16:42 +0000)]
net/mlx4: mitigate Tx path memory barriers

Replace most of the memory barriers by IO memory barriers since they
are all targeted to the DRAM; This improves code efficiency for
systems which force store order between different addresses.

Only the doorbell register store should be protected by memory barrier
since it is targeted to the PCI memory domain.

Limit pre byte count store IO memory barrier for systems with cache
line size smaller than 64B (TXBB size).

This patch improves Tx performance by 0.2MPPS for one segment 64B
packets via 1 queue with 1 core test.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
6 years agonet/mlx4: fix HW memory optimizations careless
Matan Azrad [Thu, 2 Nov 2017 16:42:50 +0000 (16:42 +0000)]
net/mlx4: fix HW memory optimizations careless

Volatilize all Rx/Tx HW negotiation memories to be sure no compiler
optimization prevents either load or store commands.

Fixes: c3c977bbecbd ("net/mlx4: add Tx bypassing Verbs")
Fixes: 9f57340a8087 ("net/mlx4: restore Rx offloads")
Fixes: 6681b845034c ("net/mlx4: add Rx bypassing Verbs")
Fixes: 62e96ffb93ad ("net/mlx4: fix no Rx interrupts")

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
6 years agonet/mlx4: separate Tx segment cases
Matan Azrad [Thu, 2 Nov 2017 16:42:49 +0000 (16:42 +0000)]
net/mlx4: separate Tx segment cases

Optimize single segment case by processing it in different block which
prevents checks, calculations and barriers relevant only for multi
segment case.

Call a dedicated function for handling multi segments case.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Signed-off-by: Ophir Munk <ophirmu@mellanox.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
6 years agonet/mlx4: remove duplicate handling in Tx burst
Matan Azrad [Thu, 2 Nov 2017 16:42:48 +0000 (16:42 +0000)]
net/mlx4: remove duplicate handling in Tx burst

Remove usage of variable which count the packets for completion and
doesn't add more information than packets counter.

Remove no space in elements ring check which is already covered by
regular Tx flow.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
6 years agonet/mlx4: merge Tx path functions
Matan Azrad [Thu, 2 Nov 2017 16:42:47 +0000 (16:42 +0000)]
net/mlx4: merge Tx path functions

Merge tx_burst and mlx4_post_send functions to prevent
double asking about WQ remain space.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
6 years agonet/mlx4: fix ring wraparound compiler hint
Matan Azrad [Thu, 2 Nov 2017 16:42:46 +0000 (16:42 +0000)]
net/mlx4: fix ring wraparound compiler hint

Remove unlikely hint from WQ wraparound check because it is
expected case.

Fixes: c3c977bbecbd ("net/mlx4: add Tx bypassing Verbs")

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
6 years agonet/mlx4: associate MR to MP in a short function
Ophir Munk [Thu, 2 Nov 2017 16:42:45 +0000 (16:42 +0000)]
net/mlx4: associate MR to MP in a short function

Associate memory region to mempool (on data path) in a short function.
Handle the less common case of adding a new memory region to mempool
in a separate function.

Signed-off-by: Ophir Munk <ophirmu@mellanox.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
6 years agonet/mlx4: remove error flows from Tx fast path
Matan Azrad [Thu, 2 Nov 2017 16:42:44 +0000 (16:42 +0000)]
net/mlx4: remove error flows from Tx fast path

Move unnecessary error flows to DEBUG mode.

Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
6 years agonet/i40e: fix Rx queue interrupt mapping in VF
Wei Dai [Fri, 3 Nov 2017 08:47:30 +0000 (16:47 +0800)]
net/i40e: fix Rx queue interrupt mapping in VF

When a VF port is bound to VFIO-PCI, miscellaneous interrupt is
mapped to MSI-X vector 0 and Rx queues interrupt are mapped to
other vectors in vfio_enable_msix( ). To simplify implementation,
all VFIO-PCI bound i40e VF Rx queue interrupts can be mapped in
vector 1. And as current igb_uio only support only one vector,
i40e VF PMD should use vector 0 for igb_uio and vector 1 for
VFIO-PCI. Without this patch, VF Rx queue interrupt is mapped
to vector 0 in register settings and mapped to VFIO vector 1
in vfio_enable_msix( ), and then all Rx queue interrupts will
be missed.
Also remove 2 unused macro definitions.

Fixes: 4b90a3ff26c5 ("i40evf: support Rx interrupt")
Fixes: 975ffea6f671 ("net/i40e: remove DPDK PF version specific code")
Cc: stable@dpdk.org
Signed-off-by: Wei Dai <wei.dai@intel.com>
Tested-by: Lei Yao <lei.a.yao@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
6 years agonet/i40e: fix VFIO interrupt mapping in VF
Wei Dai [Fri, 3 Nov 2017 08:47:29 +0000 (16:47 +0800)]
net/i40e: fix VFIO interrupt mapping in VF

When a VF port is bound to VFIO-PIC, only miscellaneous interrupt
is mapped to VFIO vector 0 in i40evf_dev_init( ).
In i40evf_dev_interrupt_handle( ) and i40evf_dev_rx_queue_intr_enable( ),
if previous VFIO interrupt mapping set in i40evf_dev_init( ) is not
cleared, it will fail when PMD tries to map Rx queue interrupt to other
VFIO vectors by calling rte_intr_enable( ).

This patch clears the VFIO interrupt mappings before setting both
miscellaneous and Rx queue interrupt mappings again to avoid failure.
And remove the calling of rte_intr_enable( ) in
i40evf_dev_interrupt_handler( ) as there is no need to map VFIO interrupt
in this function repeatedly.

Fixes: 4b90a3ff26c5 ("i40evf: support Rx interrupt")
Cc: stable@dpdk.org
Signed-off-by: Wei Dai <wei.dai@intel.com>
Tested-by: Lei Yao <lei.a.yao@intel.com>
Acked-by: Jingjing Wu <jingjing.wu@intel.com>
6 years agonet/ixgbe: fix filter parser for L2 tunnel
Wei Zhao [Fri, 3 Nov 2017 07:38:38 +0000 (15:38 +0800)]
net/ixgbe: fix filter parser for L2 tunnel

The action for L2 tunnel should be VF or PF, not QUEUE.

Fixes: 99e7003831c ("net/ixgbe: parse L2 tunnel filter")
Cc: stable@dpdk.org
Signed-off-by: Wei Zhao <wei.zhao1@intel.com>
Tested-by: Yuan Peng <yuan.peng@intel.com>
Acked-by: Wei Dai <wei.dai@intel.com>
6 years agonet/vmxnet3: fix memory leak when releasing queues
Chas Williams [Wed, 1 Nov 2017 20:38:20 +0000 (16:38 -0400)]
net/vmxnet3: fix memory leak when releasing queues

At the end of the queue release, we can free the containers for the
queue objects.

Fixes: dfaff37fc46d ("vmxnet3: import new vmxnet3 poll mode driver implementation")
Cc: stable@dpdk.org
Signed-off-by: Chas Williams <chas3@att.com>
Reviewed-by: Luca Boccassi <bluca@debian.org>
6 years agonet/mlx4: fix missing include
Adrien Mazarguil [Thu, 2 Nov 2017 13:41:25 +0000 (14:41 +0100)]
net/mlx4: fix missing include

Fixes: 76df01ff622f ("net/mlx4: separate debugging macros")

Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
6 years agonet/mlx5: fix array overflow
Nélio Laranjeiro [Thu, 2 Nov 2017 13:30:03 +0000 (14:30 +0100)]
net/mlx5: fix array overflow

VLAN id is limited to MLX5_MAX_VLAN_IDS which is not verified by the code
before trying to add a new VLAN filter.

Fixes: 272733b5ebfd ("net/mlx5: use flow to enable unicast traffic")

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
6 years agonet/mrvl: reset errno before library call
Tomasz Duszynski [Thu, 2 Nov 2017 08:37:25 +0000 (09:37 +0100)]
net/mrvl: reset errno before library call

In case errno contains some non-zero value and call to strtoul() is
successful -2 error will be returned.

Setting errno to zero before calling strtoul() makes sure
errors will be handled correctly.

Fixes: 0ddc9b815b11 ("net/mrvl: add net PMD skeleton")

Signed-off-by: Tomasz Duszynski <tdu@semihalf.com>
6 years agonet/enic: fix TSO for packets greater than 9208 bytes
John Daley [Thu, 2 Nov 2017 05:47:10 +0000 (22:47 -0700)]
net/enic: fix TSO for packets greater than 9208 bytes

A check was previously added to drop Tx packets greater than what the Nic
is capable of sending since such packets can freeze the send queue. The
check did not account for TSO packets however, so TSO was limited to 9208
bytes.

Check packet length only for non-TSO packets. Also insure that TSO packet
segment size plus the headers do not exceed what the Nic is capable of
since this also can freeze the send queue.

Use the PKT_TX_TCP_SEG ol_flag instead of m->tso_segsz which is the
preferred way to check for TSO.

Fixes: ed6e564c214e ("net/enic: fix memory leak with oversized Tx packets")
Cc: stable@dpdk.org
Signed-off-by: John Daley <johndale@cisco.com>
6 years agodoc: add clarification for i40e 16 byte desc
Wenzhuo Lu [Wed, 1 Nov 2017 02:45:28 +0000 (10:45 +0800)]
doc: add clarification for i40e 16 byte desc

The description of i40e 16 byte descriptor is not clear
enough. Add more details.

Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
6 years agodoc: add how to set VF MAC address on ixgbe
Wenzhuo Lu [Tue, 24 Oct 2017 00:57:24 +0000 (08:57 +0800)]
doc: add how to set VF MAC address on ixgbe

Add the description of how to add a MAC address for a VF
from PF on ixgbe.

Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
6 years agonet/mlx5: fix flow director TCPv6 filter
Nélio Laranjeiro [Tue, 31 Oct 2017 15:51:14 +0000 (16:51 +0100)]
net/mlx5: fix flow director TCPv6 filter

Flow director to generic flow conversion wrongly handles TCPv6 layer.

Fixes: 4c3e9bcdd52e ("net/mlx5: support flow director")

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Acked-by: Yongseok Koh <yskoh@mellanox.com>
6 years agonet/mlx5: fix flow director filter removal
Nélio Laranjeiro [Tue, 31 Oct 2017 15:51:13 +0000 (16:51 +0100)]
net/mlx5: fix flow director filter removal

Initialise struct mlx5_fdir to avoid passing non initialised
items/actions arrays elements to the generic flow interpreter.

Fixes: 4c3e9bcdd52e ("net/mlx5: support flow director")

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Acked-by: Yongseok Koh <yskoh@mellanox.com>
6 years agonet/mlx5: fix flows when VXLAN tunnel is 0
Nélio Laranjeiro [Tue, 31 Oct 2017 15:51:11 +0000 (16:51 +0100)]
net/mlx5: fix flows when VXLAN tunnel is 0

Fix a strange behavior from the NIC, when the flow starts with a VXLAN
layer with a VNI equals to zero all the traffic will match within this
rule.

Fixes: 2e709b6aa0f5 ("net/mlx5: support VXLAN flow item")
Cc: stable@dpdk.org
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Acked-by: Yongseok Koh <yskoh@mellanox.com>
6 years agonet/mlx5: fix parser inner value
Nélio Laranjeiro [Tue, 31 Oct 2017 15:51:10 +0000 (16:51 +0100)]
net/mlx5: fix parser inner value

In validate() it sets the value to 1 whereas the expected one is
IBV_FLOW_SPEC_INNER.

Fixes: c2c6ddd51da4 ("net/mlx5: limit flow API rules to one tunnel")
Cc: stable@dpdk.org
Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Acked-by: Yongseok Koh <yskoh@mellanox.com>
6 years agonet/dpaa2: set queues after reconfiguration
Akhil Goyal [Wed, 1 Nov 2017 08:16:41 +0000 (13:46 +0530)]
net/dpaa2: set queues after reconfiguration

if dpaa2_dev_tx_queue_setup is called multiple times, the
assignment of device->data->tx_queues is not done, as a result
tx_queues remain NULL after reconfiguration.

This patch sets the tx_queues from the device private data to the
usable device tx queues.

Fixes: 7ae777d064e8 ("net/dpaa2: add support for congestion notification")
Cc: stable@dpdk.org
Signed-off-by: Akhil Goyal <akhil.goyal@nxp.com>
Acked-by: Hemant Agrawal <hemant.agrawal@nxp.com>
6 years agonet/mlx5: fix counter set destroy order
Ori Kam [Mon, 30 Oct 2017 12:46:54 +0000 (14:46 +0200)]
net/mlx5: fix counter set destroy order

The counter set should be destroyed only after the flow was destroyed.

Fixes: 9a761de8ea14 ("net/mlx5: flow counter support")

Signed-off-by: Ori Kam <orika@mellanox.com>
Acked-by: Yongseok Koh <yskoh@mellanox.com>
6 years agonet/bonding: fix default aggregator mode to stable
Daniel Mrzyglod [Fri, 20 Oct 2017 14:32:54 +0000 (16:32 +0200)]
net/bonding: fix default aggregator mode to stable

This patch change default aggregator mode to stable
when using rte_eth_bond_create()

Fixes: 6d72657ce379 ("net/bonding: add other aggregator modes")
Cc: stable@dpdk.org
Signed-off-by: Daniel Mrzyglod <danielx.t.mrzyglod@intel.com>
Acked-by: Michal Jastrzebski <michalx.k.jastrzebski@intel.com>
6 years agonet/ixgbe: fix IPsec Rx with SSE
Radu Nicolau [Wed, 1 Nov 2017 11:28:32 +0000 (11:28 +0000)]
net/ixgbe: fix IPsec Rx with SSE

And small performance improvement

Fixes: 9a0752f498d2 ("net/ixgbe: enable inline IPsec")

Signed-off-by: Radu Nicolau <radu.nicolau@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
6 years agonet/i40e: fix Rx packets number for NEON
Jianbo Liu [Tue, 31 Oct 2017 05:52:44 +0000 (13:52 +0800)]
net/i40e: fix Rx packets number for NEON

Fix i40e stop receiving on ARM, as the statuses of RX descriptors
are not consistent, which is caused by cacheable hugepages.

Fixes: ae0eb310f253 ("net/i40e: implement vector PMD for ARM")
Cc: stable@dpdk.org
Signed-off-by: Jianbo Liu <jianbo.liu@arm.com>
6 years agonet/bnxt: fix VLAN spoof configuration
Ajit Khaparde [Tue, 31 Oct 2017 14:45:57 +0000 (09:45 -0500)]
net/bnxt: fix VLAN spoof configuration

We are not programming the vlan_spoof setting currently owing to
an unnecessary check. This patch fixes that.

Fixes: 36735a932ca7 ("net/bnxt: support set VF QOS and MAC anti spoof")
Cc: stable@dpdk.org
Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
6 years agonet/mlx4: fix queue index check on flow rules
Adrien Mazarguil [Tue, 31 Oct 2017 10:31:07 +0000 (11:31 +0100)]
net/mlx4: fix queue index check on flow rules

Users are not prevented from creating flow rules targeting nonexistent
queues, which silently makes such rules drop-like.

While it can be thought as a feature, reporting an error instead is
actually far more useful in order to catch common mistakes.

Fixes: 078b8b452e6b ("net/mlx4: add RSS flow rule action support")

Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
6 years agonet/mlx4: fix Rx after updating number of queues
Adrien Mazarguil [Tue, 31 Oct 2017 10:31:04 +0000 (11:31 +0100)]
net/mlx4: fix Rx after updating number of queues

When not in isolated mode, internal flow rules are automatically
maintained by the PMD to receive traffic according to global device
settings (MAC, VLAN, promiscuous mode and so on).

Since RSS support was added to the mix, it must also check whether Rx
queue configuration has changed when refreshing flow rules to prevent
the following from happening:

- With a smaller number of Rx queues, traffic is implicitly dropped
  since the existing RSS context cannot be re-applied.
- With a larger number of Rx queues, traffic remains balanced within the
  original (smaller) set of queues.

One workaround before this commit was to temporarily enter/leave
isolated mode to make it regenerate internal flow rules.

Fixes: 7d8675956f57 ("net/mlx4: add RSS support outside flow API")

Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
6 years agonet/i40e: fix mirror with firmware 6.0
Qi Zhang [Tue, 31 Oct 2017 14:26:42 +0000 (10:26 -0400)]
net/i40e: fix mirror with firmware 6.0

Value of I40E_GL_SWR_PRI_JOIN_MAP_0 need to be corrected or mirror
does not work on latest firmware 6.0.

Fixes: 77370db964f7 ("net/i40e: fix ethertype filter for new FW")
Cc: stable@dpdk.org
Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
Acked-by: Beilei Xing <beilei.xing@intel.com>
6 years agonet/nfp: avoid unplug if multiport
Alejandro Lucero [Tue, 31 Oct 2017 11:04:52 +0000 (11:04 +0000)]
net/nfp: avoid unplug if multiport

PF multiport support has some limitations like not allowing plugging or
unplugging of single ports as they all belong to same PCI device.

A previous patch for removing detachable flag introduced a wrong check.

Fixes: 00a3d8104ac5 ("ethdev: remove detachable device flag")

Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
6 years agonet/mlx5: fix adding counter when dev not started
Ori Kam [Mon, 30 Oct 2017 07:32:24 +0000 (09:32 +0200)]
net/mlx5: fix adding counter when dev not started

When adding count action to a drop flow rule while the device was not
started, the counter was not saved in the flow.

Fixes: 9a761de8ea14 ("net/mlx5: flow counter support")

Signed-off-by: Ori Kam <orika@mellanox.com>
Acked-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
6 years agodoc: add octeontx to release notes
Jerin Jacob [Sun, 29 Oct 2017 13:28:50 +0000 (18:58 +0530)]
doc: add octeontx to release notes

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
6 years agonet/mlx5: fix flow director matching rules
Nélio Laranjeiro [Fri, 27 Oct 2017 06:50:00 +0000 (08:50 +0200)]
net/mlx5: fix flow director matching rules

Flow director API does not provide a layer 2 configuration when the
filter is for layer 3 and 4 causing the translation to generic flow API
to be wrong, as not providing a mask for layer ends by using the
default one.
In this case, the Ethernet mask layer is full whereas it must be empty.

Fixes: 4c3e9bcdd52e ("net/mlx5: support flow director")

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Acked-by: Yongseok Koh <yskoh@mellanox.com>
6 years agonet/bnxt: fix HWRM command failures during VF unload
Ajit Khaparde [Mon, 30 Oct 2017 16:08:08 +0000 (11:08 -0500)]
net/bnxt: fix HWRM command failures during VF unload

In some cases when a VF driver is unloaded after the PF driver,
certain HWRM commands are returned with an error.
Instead the PF can tell the FW to permit these commands in order
to allow a clean unload.

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
6 years agonet/bnxt: fix flow director filter
Ajit Khaparde [Fri, 27 Oct 2017 15:53:34 +0000 (10:53 -0500)]
net/bnxt: fix flow director filter

Set the filter_type before we match a new filter against existing
filters. Otherwise we are missing the existing filters.

Fixes: 2d64da097aa0 ("net/bnxt: support FDIR")

Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
6 years agonet/bnxt: do not set hash type unnecessarily
Ajit Khaparde [Fri, 27 Oct 2017 02:30:24 +0000 (21:30 -0500)]
net/bnxt: do not set hash type unnecessarily

We are wrongly "OR-ing" the vnic->hash_type instead of assigning
the new hash type thereby wrongly configuring hash settings.
Fixing it.

Fixes: fcc0aa1edc10 ("net/bnxt: add RSS hash configuration")
Cc: stable@dpdk.org
Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
6 years agonet/mlx5: fix RSS action for tunneled packets
Shahaf Shuler [Thu, 26 Oct 2017 17:41:57 +0000 (20:41 +0300)]
net/mlx5: fix RSS action for tunneled packets

The flow engine in mlx5 searches for the most specific layer in the
pattern in order to set the flow rule priority properly.

Since the RSS can be currently performed only for the outer headers, avoid
updating the layer for the inner headers.

Fixes: 8086cf08b2f0 ("net/mlx5: handle RSS hash configuration in RSS flow")

Signed-off-by: Shahaf Shuler <shahafs@mellanox.com>
Acked-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
6 years agonet/ixgbe: fix build with gcc 4.5.1
Radu Nicolau [Mon, 6 Nov 2017 11:36:47 +0000 (11:36 +0000)]
net/ixgbe: fix build with gcc 4.5.1

On SUSE11 with gcc 4.5.1 the following build error occurred,
most likely because of improper handling of annonymous unions:

drivers/net/ixgbe/ixgbe_ipsec.c:579:4: error:
unknown field ‘ipsec’ specified in initializer

Fixes: 9a0752f498d2 ("net/ixgbe: enable inline IPsec")

Signed-off-by: Radu Nicolau <radu.nicolau@intel.com>
6 years agovhost: postpone ring address translations at kick time only
Maxime Coquelin [Fri, 3 Nov 2017 15:52:35 +0000 (16:52 +0100)]
vhost: postpone ring address translations at kick time only

If multiple queue pairs are created but all are not used, the
device is never started, as unused queues aren't enabled and
their ring addresses aren't translated. The device is changed
to running state when all rings addresses are translated.

This patch fixes this by postponning rings addresses translation
at kick time unconditionnaly, VHOST_USER_F_PROTOCOL_FEATURES
being negotiated or not.

Reported-by: Lei Yao <lei.a.yao@intel.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Tested-by: Lei Yao <lei.a.yao@intel.com>
Acked-by: Yuanhan Liu <yliu@fridaylinux.org>
6 years agocfgfile: fix leak on creation error
Jacek Piasecki [Thu, 26 Oct 2017 06:21:09 +0000 (08:21 +0200)]
cfgfile: fix leak on creation error

Unsuccesfull memory allocation for elements inside cfgfile
structure could result in resource leak.
Fixed by pointer verification after each malloc,
if malloc fail - error branch is proceeded with freeing memory.

Coverity issue: 195032
Fixes: d4cb8197589d ("cfgfile: support runtime modification")

Signed-off-by: Jacek Piasecki <jacekx.piasecki@intel.com>
Acked-by: Michal Jastrzebski <michalx.k.jastrzebski@intel.com>
6 years agocfgfile: fix null pointer dereference in parsing
Jacek Piasecki [Thu, 26 Oct 2017 06:24:06 +0000 (08:24 +0200)]
cfgfile: fix null pointer dereference in parsing

Function memchr() could return NULL and assign it to split[1] pointer.
Additional check and error handing is made after memchr() call.

Coverity issue: 195004
Fixes: a6a47ac9c2c9 ("cfgfile: rework load function")

Signed-off-by: Jacek Piasecki <jacekx.piasecki@intel.com>
Acked-by: Michal Jastrzebski <michalx.k.jastrzebski@intel.com>
6 years agotable: fix LRU hash table parameters check
Jasvinder Singh [Fri, 27 Oct 2017 10:47:30 +0000 (11:47 +0100)]
table: fix LRU hash table parameters check

Fixes the copy paste error in lru hash table parameters check.

Coverity issue: 198433
Fixes: b5cde2cb8c81 ("table: rework variable size key lru hash table")

Signed-off-by: Jasvinder Singh <jasvinder.singh@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
6 years agodistributor: fix some typos
Pavel Shirshov [Mon, 30 Oct 2017 07:03:01 +0000 (00:03 -0700)]
distributor: fix some typos

Signed-off-by: Pavel Shirshov <pavel.shirshov@gmail.com>
6 years agolib: fix some typos
Thomas Monjalon [Thu, 2 Nov 2017 19:30:17 +0000 (20:30 +0100)]
lib: fix some typos

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Signed-off-by: Rami Rosen <rami.rosen@intel.com>
6 years agoservice: fix race in service on app lcore function
Harry van Haaren [Wed, 1 Nov 2017 18:48:01 +0000 (18:48 +0000)]
service: fix race in service on app lcore function

This commit fixes a possible race condition if an application
uses the service-cores infrastructure and the function to run
a service on an application lcore at the same time.

The fix is to change the num_mapped_cores variable to be an
atomic variable. This causes concurrent accesses by multiple
threads to a service using rte_service_run_iter_on_app_lcore()
to detect if another core is currently mapped to the service,
and refuses to run if it is not multi-thread safe.

The run iteration on app lcore function has two arguments, the
service id to run, and if atomics should be used to serialize access
to multi-thread unsafe services. This allows applications to choose
if they wish to use use the service-cores feature, or if they
take responsibility themselves for serializing invoking a service.
See doxygen documentation for more details.

Two unit tests were added to verify the behaviour of the
function to run a service on an application core, testing both
a multi-thread safe service, and a multi-thread unsafe service.

The doxygen API documentation for the function has been updated
to reflect the current and correct behaviour.

Fixes: e9139a32f6e8 ("service: add function to run on app lcore")

Signed-off-by: Harry van Haaren <harry.van.haaren@intel.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
6 years agoeal: fix check for default plugin directory
Bruce Richardson [Mon, 6 Nov 2017 13:58:00 +0000 (13:58 +0000)]
eal: fix check for default plugin directory

The check for the existence of the default plugin directory calls stat
using an incorrect variable, which will cause a NULL pointer dereference
error.

Coverity issue: 198440
Fixes: d6a4399cdfc9 ("eal: avoid error for non-existent default PMD path")

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Acked-by: Aaron Conole <aconole@redhat.com>
6 years agoeal/linux: add interrupt counter size for vdev
Jingjing Wu [Thu, 24 Aug 2017 02:10:56 +0000 (10:10 +0800)]
eal/linux: add interrupt counter size for vdev

For virtual device, the rte_intr_handle struct is
initialized by the virtual device driver, including
the event fd assignment. If the event fd need to be
read for clean, an argument is required for the proper
event fd read.

This patch adds efd_counter_size in rte_intr_handle
struct to tell the rx interrupt process the read size.

Signed-off-by: Jingjing Wu <jingjing.wu@intel.com>
Reviewed-by: Jianfeng Tan <jianfeng.tan@intel.com>
6 years agoeal/x86: revert select optimized memcpy at run-time
Xiaoyun Li [Fri, 3 Nov 2017 12:47:23 +0000 (20:47 +0800)]
eal/x86: revert select optimized memcpy at run-time

Revert the patchset run-time Linking support including the following
3 commits:

Fixes: 84cc318424d4 ("eal/x86: select optimized memcpy at run-time")
Fixes: c7fbc80fe60f ("test: select memcpy alignment unit at run-time")
Fixes: 5f180ae32962 ("efd: move AVX2 lookup in its own compilation unit")

The patchset would cause perf drop in vhost/virtio loopback performance
test. Because the run-time dispatch must cost at least a function call
comparing to the compile-time dispatch. And the reference cpu cycles value
is small. And in the test, when using 128-256 bytes packet, it would cause
16%-20% perf drop with mergeble path. When using 256 bytes packet, it would
cause 13% perf drop with vector path.

Signed-off-by: Xiaoyun Li <xiaoyun.li@intel.com>
6 years agoigb_uio: prevent reset for bnx2x devices
Ferruh Yigit [Mon, 6 Nov 2017 18:48:15 +0000 (18:48 +0000)]
igb_uio: prevent reset for bnx2x devices

Some devices are having problem on device reset that happens during DPDK
application exit [1].

Create a static list of devices and exclude them from device reset.

[1]
http://dpdk.org/ml/archives/dev/2017-November/080927.html

Fixes: b58eedfc7dd5 ("igb_uio: issue FLR during open and release of device file")
Cc: stable@dpdk.org
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
6 years agobus/pci: fix namespace of sysfs path function
Thomas Monjalon [Mon, 6 Nov 2017 01:56:09 +0000 (02:56 +0100)]
bus/pci: fix namespace of sysfs path function

The function pci_get_sysfs_path was moved from EAL to the PCI driver.

The namespace is now fixed by adding "rte_" prefix.
The map files are fixed by removing the symbol from EAL and adding
it to the PCI driver.

It is an API break but it is probably not used by applications.
Anyway this API is already broken by the move in a new header file.

Fixes: c752998b5e2e ("pci: introduce library and driver")

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Gaetan Rivet <gaetan.rivet@6wind.com>
6 years agobus/pci: fix IOMMU class for sPAPR
Jonas Pfefferle [Fri, 3 Nov 2017 12:05:19 +0000 (13:05 +0100)]
bus/pci: fix IOMMU class for sPAPR

PPC64 sPAPR iommu does not support iova as va.
Use pa mode instead.

Fixes: 815c7deaed2d ("pci: get IOMMU class on Linux")

Signed-off-by: Jonas Pfefferle <jpf@zurich.ibm.com>
6 years agoeal/linux: force IOVA as PA mode if KNI module inserted
Ferruh Yigit [Thu, 2 Nov 2017 00:06:00 +0000 (00:06 +0000)]
eal/linux: force IOVA as PA mode if KNI module inserted

Fix kernel crash with KNI because KNI requires physical addresses.

When IOVA VA mode used, memzones and mbufs physical address fields
contain virtual addresses. But KNI relies on these fields to enable
kernel access for buffers. Those fields having virtual address cause
crash in kernel.

This is a workaround until KNI fixed properly to work with virtual
addresses.

Fixes: 72d013644bd6 ("mem: honor IOVA mode in malloc virt2phy")

Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
6 years agoeal: fix version map experimental section
Harry van Haaren [Wed, 25 Oct 2017 12:29:49 +0000 (13:29 +0100)]
eal: fix version map experimental section

Before this commit, the EXPERIMENTAL version of ABI
derived from the DPDK_17.08 tag. In parallel there
was a DPDK_17.11 tag.

Experimental map should always derive from the latest ABI,
so this patch moves the 17.11 section above EXPERIMENTAL,
and updates EXPERIMENTAL to derive from the 17.11 map.

Fixes: aadc3eb002d3 ("pci: export match function")

Signed-off-by: Harry van Haaren <harry.van.haaren@intel.com>
6 years agomempool: increase ABI version
Thomas Monjalon [Mon, 6 Nov 2017 10:54:26 +0000 (11:54 +0100)]
mempool: increase ABI version

API and ABI of mempool library has been changed in 17.11.

Fixes: 02604520b2f2 ("mempool: remove unused flags argument")
Fixes: 0cc0f8aaa35d ("mempool: change flags from int to unsigned int")
Fixes: 6eac187bff30 ("mempool: add flags arg in xmem size and usage")

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Santosh Shukla <santosh.shukla@caviumnetworks.com>
6 years agodoc: add IOVA aware API changes in release notes
Thomas Monjalon [Fri, 20 Oct 2017 12:31:35 +0000 (18:01 +0530)]
doc: add IOVA aware API changes in release notes

The wording changes have been done in the API without breaking
the ABI. The deprecated fields and symbols can be removed later
when an another ABI change will be required.
The deprecation notice can be removed.

The release notes describe the new available API with IOVA wording.

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Santosh Shukla <santosh.shukla@caviumnetworks.com>
Acked-by: John McNamara <john.mcnamara@intel.com>
6 years agodrivers/net: rename physical address type to IOVA
Santosh Shukla [Fri, 20 Oct 2017 12:31:31 +0000 (18:01 +0530)]
drivers/net: rename physical address type to IOVA

Renamed data type from phys_addr_t to rte_iova_t.

Signed-off-by: Santosh Shukla <santosh.shukla@caviumnetworks.com>
Reviewed-by: Anatoly Burakov <anatoly.burakov@intel.com>
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
6 years agocryptodev: rename physical address type to IOVA
Santosh Shukla [Fri, 20 Oct 2017 12:31:31 +0000 (18:01 +0530)]
cryptodev: rename physical address type to IOVA

Renamed data type from phys_addr_t to rte_iova_t.

Signed-off-by: Santosh Shukla <santosh.shukla@caviumnetworks.com>
Reviewed-by: Anatoly Burakov <anatoly.burakov@intel.com>
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
6 years agombuf: rename data address helpers to IOVA
Thomas Monjalon [Sun, 5 Nov 2017 23:22:55 +0000 (00:22 +0100)]
mbuf: rename data address helpers to IOVA

The following inline functions and macros have been renamed to be
consistent with the IOVA wording:

rte_mbuf_data_dma_addr         -> rte_mbuf_data_iova
rte_mbuf_data_dma_addr_default -> rte_mbuf_data_iova_default
rte_pktmbuf_mtophys            -> rte_pktmbuf_iova
rte_pktmbuf_mtophys_offset     -> rte_pktmbuf_iova_offset

The deprecated functions and macros are kept to avoid breaking the API.

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
6 years agombuf: rename physical address to IOVA
Santosh Shukla [Fri, 20 Oct 2017 12:31:32 +0000 (18:01 +0530)]
mbuf: rename physical address to IOVA

Rename buf_physaddr to buf_iova.
Keep the deprecated name in an anonymous union to avoid breaking
the API.

Signed-off-by: Santosh Shukla <santosh.shukla@caviumnetworks.com>
Reviewed-by: Anatoly Burakov <anatoly.burakov@intel.com>
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
6 years agomempool: rename populate functions to IOVA
Thomas Monjalon [Sun, 5 Nov 2017 22:26:24 +0000 (23:26 +0100)]
mempool: rename populate functions to IOVA

The functions rte_mempool_populate_phys() and
rte_mempool_populate_phys_tab() are renamed to
rte_mempool_populate_iova() and rte_mempool_populate_iova_tab().
The deprecated functions are kept as aliases to avoid breaking the API.

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
6 years agomempool: rename address mapping function to IOVA
Thomas Monjalon [Sun, 5 Nov 2017 18:02:29 +0000 (19:02 +0100)]
mempool: rename address mapping function to IOVA

The function rte_mempool_virt2phy() is renamed to rte_mempool_virt2iova().
The new function has one less parameter because it is unused.
The deprecated function is kept as an alias to avoid breaking the API.

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Santosh Shukla <santosh.shukla@caviumnetworks.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
6 years agomempool: rename addresses from physical to IOVA
Thomas Monjalon [Fri, 20 Oct 2017 12:31:31 +0000 (18:01 +0530)]
mempool: rename addresses from physical to IOVA

The struct fields phys_addr_t rte_mempool_objhdr.physaddr and
rte_mempool_memhdr.phys_addr are renamed to rte_iova_t iova.
The deprecated names are kept in an anonymous union to avoid breaking
the API.

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Santosh Shukla <santosh.shukla@caviumnetworks.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
6 years agomemzone: rename address from physical to IOVA
Thomas Monjalon [Sat, 4 Nov 2017 01:22:28 +0000 (02:22 +0100)]
memzone: rename address from physical to IOVA

The struct rte_memzone field .phys_addr is renamed to .iova.
The deprecated name is kept in an anonymous union to avoid breaking
the API.

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Santosh Shukla <santosh.shukla@caviumnetworks.com>
6 years agomalloc: use pointer diff macro in IOVA mapping
Santosh Shukla [Fri, 20 Oct 2017 12:31:36 +0000 (18:01 +0530)]
malloc: use pointer diff macro in IOVA mapping

Use RTE_PTR_DIFF macro in rte_mem_virt2iova api.

Suggested-by: Anatoly Burakov <anatoly.burakov@intel.com>
Signed-off-by: Santosh Shukla <santosh.shukla@caviumnetworks.com>
6 years agomalloc: rename address mapping function to IOVA
Thomas Monjalon [Sat, 4 Nov 2017 01:22:21 +0000 (02:22 +0100)]
malloc: rename address mapping function to IOVA

The function rte_malloc_virt2phy() is renamed to rte_malloc_virt2iova().
The deprecated name is kept as an alias to avoid breaking the API.

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Santosh Shukla <santosh.shukla@caviumnetworks.com>
6 years agomem: rename address mapping function to IOVA
Thomas Monjalon [Sat, 4 Nov 2017 16:15:04 +0000 (17:15 +0100)]
mem: rename address mapping function to IOVA

The function rte_mem_virt2phy() is kept and used in functions which
works only with physical addresses.
For all other calls this function is replaced by rte_mem_virt2iova()
which does a direct mapping (no conversion) in the VA case.

Note: the new function rte_mem_virt2iova() function matches the
behaviour implemented in rte_mem_virt2phy() by the commit
680f6c12600f ("mem: honor IOVA mode in virt2phy")

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Santosh Shukla <santosh.shukla@caviumnetworks.com>
6 years agomem: rename segment address from physical to IOVA
Santosh Shukla [Fri, 20 Oct 2017 12:31:33 +0000 (18:01 +0530)]
mem: rename segment address from physical to IOVA

Renaming rte_memseg {.phys_addr} to {.iova}
Keep the deprecated name in an anonymous union to avoid breaking
the API.

Use rte_iova_t and RTE_BAD_IOVA where appropriate in
memory segment handling.

Signed-off-by: Santosh Shukla <santosh.shukla@caviumnetworks.com>
Reviewed-by: Anatoly Burakov <anatoly.burakov@intel.com>
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
6 years agomem: introduce IOVA type
Thomas Monjalon [Fri, 3 Nov 2017 23:36:47 +0000 (00:36 +0100)]
mem: introduce IOVA type

The IO virtual addresses may be used instead of physical addresses.
As IOVA is more generic, it should be used in most places instead
of physical address wording.

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Santosh Shukla <santosh.shukla@caviumnetworks.com>
6 years agomem: hide physical address error in VA mode
Thomas Monjalon [Sat, 4 Nov 2017 22:09:03 +0000 (23:09 +0100)]
mem: hide physical address error in VA mode

If the IOVA mode is not using physical addresses,
no need to log an error about physical address issue.

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Santosh Shukla <santosh.shukla@caviumnetworks.com>
6 years agomem: remove old function from symbol list
Thomas Monjalon [Fri, 3 Nov 2017 11:21:04 +0000 (12:21 +0100)]
mem: remove old function from symbol list

The function rte_mem_phy2mch() was removed with the support
of Xen dom0.

Fixes: a7cb2e20d23c ("mem: remove API to get physical address in dom0")

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
6 years agoremove useless memzone includes
Thomas Monjalon [Sat, 4 Nov 2017 00:43:57 +0000 (01:43 +0100)]
remove useless memzone includes

The memzone header is often included without good reason.

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
6 years agoremove include of generated config header
Thomas Monjalon [Thu, 2 Nov 2017 21:35:02 +0000 (22:35 +0100)]
remove include of generated config header

The file rte_config.h is generated and automatically included
with -include option.
The explicit includes in drivers and libraries are useless.

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
6 years agouse macro to declare constructor functions
Thomas Monjalon [Thu, 2 Nov 2017 22:06:38 +0000 (23:06 +0100)]
use macro to declare constructor functions

It is easier to find all constructor functions when they use
the same macros RTE_INIT or RTE_INIT_PRIO.

The macro definitions are moved from rte_eal.h to rte_common.h.

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
6 years agomk: remove useless list of core libraries
Thomas Monjalon [Mon, 6 Nov 2017 11:21:20 +0000 (12:21 +0100)]
mk: remove useless list of core libraries

When moving the library dependencies handling from top Makefiles
to library Makefiles, the list core-libs became useless.

Fixes: cbc12b0a96f5 ("mk: do not generate LDLIBS from directory dependencies")

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
6 years agoconfig: fix bnx2x option for armv7a
Ferruh Yigit [Tue, 31 Oct 2017 17:39:24 +0000 (17:39 +0000)]
config: fix bnx2x option for armv7a

Fixes: 02a8686263de ("mk: introduce ARMv7 architecture")
Cc: stable@dpdk.org
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
6 years agobuildtools: fix icc build
Ferruh Yigit [Thu, 2 Nov 2017 00:25:10 +0000 (00:25 +0000)]
buildtools: fix icc build

There are random build errors in test reports [1]. Build error
is not directly related to DPDK but observed during DPDK build.

When I get similar unexpected build errors in my system, found
out that /dev/null is invalid.

It seems ICC overwrites the /dev/null with "icc -o /dev/null" instead
of sending output to /dev/null. This is not always reproducible, so
hard to say what exactly is triggering the error.

I suspect test-report build errors can be because of the same reason,
and it is good to add a protection for this case.

Instead of sending output to /dev/null save it to the tmp folder and
remove it back when done.

[1]
http://dpdk.org/ml/archives/test-report/2017-November/034053.html
Failure #3

/usr/src/linux-headers-4.4.0-97-generic/include/linux/sysfs.h:517:37:
error: pointer targets in passing argument 2 of ‘kernfs_find_and_get’
differ in signedness [-Werror=pointer-sign]
  return kernfs_find_and_get(parent, name);

Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
6 years agopci: fix namespace prefix of new functions
Gaetan Rivet [Mon, 6 Nov 2017 16:08:59 +0000 (17:08 +0100)]
pci: fix namespace prefix of new functions

Some symbols were introduced with the wrong prefix.
Add the usual "rte_" prefix when needed.

Fixes: c752998b5e2e ("pci: introduce library and driver")

Signed-off-by: Gaetan Rivet <gaetan.rivet@6wind.com>
6 years agovfio: fix namespace prefix of newly exposed functions
Gaetan Rivet [Mon, 6 Nov 2017 16:08:58 +0000 (17:08 +0100)]
vfio: fix namespace prefix of newly exposed functions

Exposed VFIO functions simply uses a "vfio" prefix.
Use the proper "rte_vfio" prefix for those symbols.

Fixes: 279b581c897d ("vfio: expose functions")

Signed-off-by: Gaetan Rivet <gaetan.rivet@6wind.com>
6 years agobus/pci: fix VFIO device reset
Jerin Jacob [Sat, 28 Oct 2017 06:22:55 +0000 (11:52 +0530)]
bus/pci: fix VFIO device reset

If the device is not capable of resetting, then Linux kernel updates
the errno as EINVAL.
http://elixir.free-electrons.com/linux/v4.9/source/drivers/vfio/pci/vfio_pci.c#L887

Honor the EINVAL errno value to avoid pci vfio setup failure.

Fixes: f25f8f367644 ("bus/pci: check VFIO reset ioctl error")

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Reviewed-by: Jonas Pfefferle <jpf@zurich.ibm.com>
6 years agobus/pci: fix VFIO mode
Ferruh Yigit [Mon, 30 Oct 2017 22:32:27 +0000 (22:32 +0000)]
bus/pci: fix VFIO mode

Revert back to using VFIO_PRESENT as a marker to enable compilation
of VFIO-related segments.

VFIO_PRESENT is the combination of user configuration RTE_EAL_VFIO and
kernel version support check.

eal_vfio.h VFIO_PRESENT related check ordered to be compatible with
rte_vfio.h one, no functional modification.

Fixes: 279b581c897d ("vfio: expose functions")

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Signed-off-by: Gaetan Rivet <gaetan.rivet@6wind.com>
Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Tested-by: Bruce Richardson <bruce.richardson@intel.com>
6 years agoversion: 17.11-rc2
Thomas Monjalon [Thu, 26 Oct 2017 23:39:36 +0000 (01:39 +0200)]
version: 17.11-rc2

Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
6 years agonet/mlx5: fix TSO inline size calculation
Yongseok Koh [Wed, 25 Oct 2017 23:30:40 +0000 (16:30 -0700)]
net/mlx5: fix TSO inline size calculation

When "net/mlx5: replace network to host macros" is rebased for v4,
changes of "net/mlx5: fix calculating TSO inline size" have been
reverted.

Fixes: 6b30a6a8552a ("net/mlx5: replace network to host macros")
Cc: stable@dpdk.org
Signed-off-by: Yongseok Koh <yskoh@mellanox.com>
Acked-by: Shahaf Shuler <shahafs@mellanox.com>
Acked-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
6 years agonet/mlx4: fix no Rx interrupts
Moti Haimovsky [Wed, 25 Oct 2017 15:37:27 +0000 (18:37 +0300)]
net/mlx4: fix no Rx interrupts

This commit addresses the issue of Rx interrupts support with
the new Rx datapath introduced in DPDK version 17.11.
In order to generate an Rx interrupt an event queue is armed with the
consumer index of the Rx completion queue. Since version 17.11 this
index is handled by the PMD so it is now the responsibility of the
PMD to write this value when enabling Rx interrupts.

Fixes: 6681b845034c ("net/mlx4: add Rx bypassing Verbs")

Signed-off-by: Moti Haimovsky <motih@mellanox.com>
Acked-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
6 years agonet/mlx4: introducing consumer index mask
Moti Haimovsky [Wed, 25 Oct 2017 15:37:26 +0000 (18:37 +0300)]
net/mlx4: introducing consumer index mask

This commit defines MLX4_CQ_DB_CI_MASK which is used when updating
the consumer index of the completion queue instead of the hardcoded
0xffffff used until now.

Signed-off-by: Moti Haimovsky <motih@mellanox.com>
Acked-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
6 years agonet/liquidio: add support for 10GBase-T cards
Srisivasubramanian S [Tue, 24 Oct 2017 04:42:50 +0000 (10:12 +0530)]
net/liquidio: add support for 10GBase-T cards

Signed-off-by: Srisivasubramanian S <ssrinivasan@caviumnetworks.com>
Signed-off-by: Shijith Thotton <shijith.thotton@caviumnetworks.com>
6 years agonet/failsafe: fix Rx clean race
Matan Azrad [Sun, 22 Oct 2017 05:51:08 +0000 (05:51 +0000)]
net/failsafe: fix Rx clean race

When removing a device, the fail-safe checks that it is not within its
datapath before cleaning it.

When checking whether an Rx burst should be performed on a device, the
remove flag is not checked. Thus the port could still enter its datapath
and miss a removal round. Furthermore, there is a race between the
thread removing the device and the polling thread.

Check the remove flag before entering a sub-device Rx burst when in safe
mode. This check mitigates the aforementioned race condition.

Fixes: 72a57bfd9a0e ("net/failsafe: add fast burst functions")
Cc: stable@dpdk.org
Signed-off-by: Matan Azrad <matan@mellanox.com>
Acked-by: Gaetan Rivet <gaetan.rivet@6wind.com>
6 years agolpm6: fix compilation with -Og
Olivier Matz [Mon, 11 Sep 2017 15:13:31 +0000 (17:13 +0200)]
lpm6: fix compilation with -Og

The compilation with gcc-6.3.0 and EXTRA_CFLAGS=-Og gives the following
error:

  CC rte_lpm6.o
  rte_lpm6.c: In function ‘rte_lpm6_add_v1705’:
  rte_lpm6.c:442:11: error: ‘tbl_next’ may be used uninitialized in
                             this function [-Werror=maybe-uninitialized]
     if (!tbl[tbl_index].valid) {
             ^
  rte_lpm6.c:521:29: note: ‘tbl_next’ was declared here
    struct rte_lpm6_tbl_entry *tbl_next;
                               ^~~~~~~~

This is a false positive from gcc. Fix it by initializing tbl_next
to NULL.

Fixes: 5c510e13a9cb ("lpm: add IPv6 support")
Cc: stable@dpdk.org
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
6 years agodoc: add new introduction to sample app guides
Marko Kovacevic [Wed, 25 Oct 2017 15:51:00 +0000 (16:51 +0100)]
doc: add new introduction to sample app guides

Add new Introduction Section into the sample app guides.

Signed-off-by: Marko Kovacevic <marko.kovacevic@intel.com>
Acked-by: John McNamara <john.mcnamara@intel.com>
6 years agodoc: add generic build instructions for sample apps
Herakliusz Lipiec [Wed, 25 Oct 2017 15:50:59 +0000 (16:50 +0100)]
doc: add generic build instructions for sample apps

Moved duplicated, and occasionally outdated, doc sections from each
of the sample app guides chapters to a common chapter at the start.

This reduces the duplication in the docs and provides a single
point of reference for compiling the sample apps.

Signed-off-by: Herakliusz Lipiec <herakliusz.lipiec@intel.com>
Signed-off-by: Marko Kovacevic <marko.kovacevic@intel.com>
6 years agoexamples/multi_process: fix received message length
Xueming Li [Thu, 26 Oct 2017 08:29:23 +0000 (16:29 +0800)]
examples/multi_process: fix received message length

Simple_mp example receives message size less than 64 chars while send
side accepts chars less than 128, this leads to different result when
sending text length larger than 64.
This patch uses same buffer length on both message pool and command
line.

Fixes: af75078fece3 ("first public release")
Cc: stable@dpdk.org
Signed-off-by: Xueming Li <xuemingl@mellanox.com>
6 years agodoc: add event eth Rx adapter to release notes
Nikhil Rao [Tue, 24 Oct 2017 10:32:19 +0000 (16:02 +0530)]
doc: add event eth Rx adapter to release notes

Signed-off-by: Nikhil Rao <nikhil.rao@intel.com>
Acked-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
6 years agodoc: update software event device
Pavan Nikhilesh [Wed, 25 Oct 2017 14:50:33 +0000 (20:20 +0530)]
doc: update software event device

Update software event device documentation to include use of service
cores for event distribution.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
Acked-by: John McNamara <john.mcnamara@intel.com>
6 years agoeventdev: remove schedule API
Pavan Nikhilesh [Wed, 25 Oct 2017 14:50:32 +0000 (20:20 +0530)]
eventdev: remove schedule API

remove eventdev schedule api and enforce sw driver to use service core
feature for event scheduling.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
Acked-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
6 years agoexamples/eventdev: use service library
Pavan Nikhilesh [Wed, 25 Oct 2017 14:50:31 +0000 (20:20 +0530)]
examples/eventdev: use service library

Update the sample app eventdev_pipeline_sw_pmd to use service run iter for
event scheduling in case of sw eventdev.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
Acked-by: Harry van Haaren <harry.van.haaren@intel.com>
6 years agotest/eventdev: use service iter
Pavan Nikhilesh [Wed, 25 Oct 2017 14:50:30 +0000 (20:20 +0530)]
test/eventdev: use service iter

Use service run iter for event scheduling instead of calling the event
schedule api directly.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
Acked-by: Harry van Haaren <harry.van.haaren@intel.com>
6 years agoapp/testeventdev: use service cores
Pavan Nikhilesh [Wed, 25 Oct 2017 14:50:29 +0000 (20:20 +0530)]
app/testeventdev: use service cores

Use service cores for offloading event scheduling in case of
centralized scheduling instead of calling the schedule api directly.
This removes the dependency on dedicated scheduler core specified by
giving command line option --slcore.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
Acked-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
6 years agoevent/sw: extend service capability
Pavan Nikhilesh [Wed, 25 Oct 2017 14:50:28 +0000 (20:20 +0530)]
event/sw: extend service capability

Extend the service capability of the sw event device by exposing service id
to the application.
The application can use service id to configure service cores to run event
scheduling.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
Acked-by: Harry van Haaren <harry.van.haaren@intel.com>
6 years agoeventdev: add API to get service id
Pavan Nikhilesh [Wed, 25 Oct 2017 14:50:27 +0000 (20:20 +0530)]
eventdev: add API to get service id

In case of sw event device the scheduling can be done on a service core
using the service registered at the time of probe.
This patch adds a helper function to get the service id that can be used
by the application to assign a lcore for the service to run on.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
Acked-by: Harry van Haaren <harry.van.haaren@intel.com>