2017-04-12 09:25, Stephen Hemminger:
> On Tue, 11 Apr 2017 17:44:07 +0200
> Gaetan Rivet <[email protected]> wrote:
>
> > I rebased the following work of Jan:
> >
> > This series is removing the PCI specific struct eth_driver from rte_ether.
> > The
> > PCI drivers are changed to use the newly introduced header-only helpers
> > instead. Although the virtual drivers did not make use of the ethdev's
> > driver
> > field they are converted to use the VDEV specific allocation helpers. The
> > motivation for this change is to properly embed a reference to the generic
> > rte_device in the ethdev.
> >
> > The series is based on:
> >
> > * http://dpdk.org/dev/patchwork/patch/20416/
> > * http://dpdk.org/dev/patchwork/patch/20417/
> > * The "Rework vdev probing to use rte_bus infrastructure" series
> > * http://dpdk.org/dev/patchwork/patch/21058/
> >
> > Changes in v2:
> > * Removed the patch "net/vmxnet3: use library function for DMA zone
> > reserve"
> > It is essentially duplicating the following commit:
> > 04df93d1edac ("net/vmxnet3: fix queue size changes")
> > * Fixed header includes for mlx4 and mlx5 PMD
> > * Followed the changes of the series for new PMDs (AVP, liquidio)
> > * Made KNI use the new API
> > * Made PCAP use the new API
> > * Fixed undefined behavior on vdev driver arg parsing
> > * Fixed virtual_pmd for unit test
>
> Looks good.
>
> Acked-by: Stephen Hemminger <[email protected]>
Applied, thanks
Now we need to remove
struct rte_pci_driver pci_drv;
from cryptodev and eventdev.
Sooner is better :)
Thanks
