[dpdk-dev] [PATCH] net/virtio: fix secondary process crash with PCI devices

David Marchand david.marchand at redhat.com
Wed Feb 3 18:03:48 CET 2021


On Wed, Feb 3, 2021 at 4:58 PM Maxime Coquelin
<maxime.coquelin at redhat.com> wrote:
>
> The Virtio rework series mistakenly moved the rte_pci_device
> pointer to struct virtio_hw, which is shared between the two
> processes. But this structure is per-process, so this change
> made secondary process to try accessing primary process-only
> memory, leading to a crash

crash.*

Ah the joy of multiprocess support :-).

>
> This patch reverts to proper behavior, by storing the
> rte_pci_device pointer into the pre-process

per*

> virtio_pci_internal struct. It provides also helper to get

also provides*

> the pointer from the virtio_hw struct pointer.
>
> Bugzilla ID: 633
> Fixes: c8d4b02f72ae ("net/virtio: move legacy IO to virtio PCI")
>
> Reported-by: Anatoly Burakov <anatoly.burakov at intel.com>
> Signed-off-by: Maxime Coquelin <maxime.coquelin at redhat.com>

The patch lgtm.
Reviewed-by: David Marchand <david.marchand at redhat.com>

Thanks.

-- 
David Marchand



More information about the dev mailing list