[dpdk-dev,8/9] net/virtio: keep Rx handler whatever the Tx queue config

  Split use_simple_rxtx into use_simple_rx and use_simple_tx,
and ensure that only use_simple_tx is updated when txq flags
forces to use the standard Tx handler.

This change is also useful for next commit (disable simple Rx
path when Rx checksum is requested).

Cc: stable@dpdk.org

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
---
 drivers/net/virtio/virtio_ethdev.c      | 25 ++++++++++++++++---------
 drivers/net/virtio/virtio_pci.h         |  3 ++-
 drivers/net/virtio/virtio_rxtx.c        | 18 +++++++++---------
 drivers/net/virtio/virtio_user_ethdev.c |  3 ++-
 4 files changed, 29 insertions(+), 20 deletions(-)
  

Platform description
--------------------

guest (dpdk)
+----------------+
|                |
|                |
|    port0       |
+----------------+
       |
       | virtio
       |
+----------------+
|     tap0       |
|                |
|                |
+----------------+
host (linux, vhost-net)

Host configuration
------------------

Start qemu with:

- a ne2k management interface to avoi any conflict with dpdk
- a virtio net device, connected to a tap interface through vhost-net
- mergeable buffers disabled

  /usr/bin/qemu-system-x86_64 -k fr -daemonize --enable-kvm -m 2G -cpu host \
    -smp 3 -serial telnet::40564,server,nowait -serial null \
    -qmp tcp::44340,server,nowait -monitor telnet::49229,server,nowait \
    -device ne2k_pci,mac=de:ad:de:01:02:03,netdev=user.0,addr=03 \
    -netdev user,id=user.0,hostfwd=tcp::34965-:22 \
    -netdev type=tap,id=vhostnet0,script=no,vhost=on,queues=8 \
    -device virtio-net-pci,mrg_rxbuf=off,netdev=vhostnet0,mq=on,vectors=17 \
    -hda "${VM_PATH}/ubuntu-16.04-template.qcow2" \
    -snapshot -vga none -display none

Guest configuration
-------------------

Compile dpdk:

  cd dpdk.org
  make config T=x86_64-native-linuxapp-gcc
  sed -i 's,CONFIG_RTE_LIBRTE_VIRTIO_DEBUG_INIT=n,CONFIG_RTE_LIBRTE_VIRTIO_DEBUG_INIT=y,' build/.config
  sed -i 's,CONFIG_RTE_LIBRTE_VIRTIO_DEBUG_DRIVER=n,CONFIG_RTE_LIBRTE_VIRTIO_DEBUG_DRIVER=y,' build/.config
  make -j4

Prepare environment:

  mkdir -p /mnt/huge
  mount -t hugetlbfs nodev /mnt/huge
  echo 256 > /sys/devices/system/node/node0/hugepages/hugepages-2048kB/nr_hugepages
  modprobe uio_pci_generic
  python usertools/dpdk-devbind.py -b uio_pci_generic 0000:00:02.0

  ./build/app/testpmd -l 0,1 --log-level 7 -- --total-num-mbufs=16384 \
    -i --port-topology=chained --disable-hw-vlan-filter \
    --disable-hw-vlan-strip --txqflags=0

Without the fix, standard path is used in rx and tx, but simple rx could
be used:

  ...
  PMD: set_rxtx_funcs(): virtio: using standard Rx path on port 0
  PMD: set_rxtx_funcs(): virtio: using standard Tx path on port 0
  ...

With the fix:

  ...
  PMD: set_rxtx_funcs(): virtio: using simple Rx path on port 0
  PMD: set_rxtx_funcs(): virtio: using standard Tx path on port 0
  ...
  

On Thu, Aug 31, 2017 at 03:40:14PM +0200, Olivier Matz wrote:
> Split use_simple_rxtx into use_simple_rx and use_simple_tx,
> and ensure that only use_simple_tx is updated when txq flags
> forces to use the standard Tx handler.

I think it's a good idea to split it.

> This change is also useful for next commit (disable simple Rx
> path when Rx checksum is requested).
> 
> Cc: stable@dpdk.org

But again, I don't think it's a good idea to put them (including the
next patch) to stable releases.

	--yliu
  

On Fri, Sep 01, 2017 at 05:25:38PM +0800, Yuanhan Liu wrote:
> On Thu, Aug 31, 2017 at 03:40:14PM +0200, Olivier Matz wrote:
> > Split use_simple_rxtx into use_simple_rx and use_simple_tx,
> > and ensure that only use_simple_tx is updated when txq flags
> > forces to use the standard Tx handler.
> 
> I think it's a good idea to split it.
> 
> > This change is also useful for next commit (disable simple Rx
> > path when Rx checksum is requested).
> > 
> > Cc: stable@dpdk.org
> 
> But again, I don't think it's a good idea to put them (including the
> next patch) to stable releases.

Yes, you're right this one is more an enhancement than a fix: it
just selects a faster handler if it's possible.

But next one is a fix: it must not select the simple handler if
offload is requested.

If these commits are not pushed in stable, it's not a problem for
me, so I let you decide. In that case, I will remove the Cc: stable
from the last 3 commits. Do you confirm?

Thanks for the review.
Olivier
  

On Fri, Sep 01, 2017 at 11:58:07AM +0200, Olivier MATZ wrote:
> On Fri, Sep 01, 2017 at 05:25:38PM +0800, Yuanhan Liu wrote:
> > On Thu, Aug 31, 2017 at 03:40:14PM +0200, Olivier Matz wrote:
> > > Split use_simple_rxtx into use_simple_rx and use_simple_tx,
> > > and ensure that only use_simple_tx is updated when txq flags
> > > forces to use the standard Tx handler.
> > 
> > I think it's a good idea to split it.
> > 
> > > This change is also useful for next commit (disable simple Rx
> > > path when Rx checksum is requested).
> > > 
> > > Cc: stable@dpdk.org
> > 
> > But again, I don't think it's a good idea to put them (including the
> > next patch) to stable releases.
> 
> Yes, you're right this one is more an enhancement than a fix: it
> just selects a faster handler if it's possible.

Exactly.

> But next one is a fix: it must not select the simple handler if
> offload is requested.
> 
> If these commits are not pushed in stable, it's not a problem for
> me, so I let you decide. In that case, I will remove the Cc: stable
> from the last 3 commits. Do you confirm?

Yes, I think it's better to drop the cc stable tag.

> Thanks for the review.

Thanks for the work!

	--yliu