Re: [PATCH] vdpa: fix emulated guest announce feature status handling

[Gautam Dawar <gdawar@amd.com>]

On 3/3/23 20:28, Eugenio Perez Martin wrote:
> Caution: This message originated from an External Source. Use proper caution when opening attachments, clicking links, or responding.
>
>
> On Fri, Mar 3, 2023 at 12:58 PM Gautam Dawar <gautam.dawar@amd.com> wrote:
>> Guest announce capability is emulated by qemu in the .avail_handler
>> shadow virtqueue operation. It updates the status to success in
>> `*s->status` but incorrectly fetches the command execution
>> status from local variable `status` later in call to iov_from_buf().
>> As `status` is initialized to VIRTIO_NET_ERR, it results in a
>> warning "Failed to ack link announce" in virtio_net driver's
>> virtnet_ack_link_announce() function after VM Live Migration.
>> Also, I noticed an invalid check in vhost_vdpa_net_handle_ctrl_avail()
>> that reports an error because status is not updated in call to
>> virtio_net_handle_ctrl_iov():
>>
> status should be updated through &in. It is declared as:
> const struct iovec in = {
>          .iov_base = &status,
>          .iov_len = sizeof(status),
>      };
>
> And it should be filled at the end of virtio_net_handle_ctrl_iov with a call to:
>      s = iov_from_buf(in_sg, in_num, 0, &status, sizeof(status));

Apologies for a delayed response. This totally makes sense and I've not 
been able to reproduce this issue.


> How do you obtain different values? Maybe const optimizations is
> invalid and the compiler thinks virtio_net_handle_ctrl_iov will never
> change status?
>
> Thanks!

I think the issue might have been a result of incorrectly returning 
VIRTIO_NET_F_GUEST_ANNOUNCE in the device features without handling of 
VIRTIO_NET_CTRL_ANNOUNCE_ACK in the parent vdpa driver.

We can drop this patch.

Gautam

>>      virtio_net_handle_ctrl_iov(svq->vdev, &in, 1, &out, 1);
>>      if (status != VIRTIO_NET_OK) {
>>          error_report("Bad CVQ processing in model");
>>      }
>> Adding an optional OUT status parameter to virtio_net_handle_ctrl_iov()
>> would help resolving this issue and also send the correct status
>> value to the virtio-net driver.
>>
>> Signed-off-by: Gautam Dawar <gautam.dawar@amd.com>
>> ---
>>   hw/net/virtio-net.c            | 9 +++++++--
>>   include/hw/virtio/virtio-net.h | 3 ++-
>>   net/vhost-vdpa.c               | 3 +--
>>   3 files changed, 10 insertions(+), 5 deletions(-)
>>
>> diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c
>> index 3ae909041a..36a75592da 100644
>> --- a/hw/net/virtio-net.c
>> +++ b/hw/net/virtio-net.c
>> @@ -1503,7 +1503,8 @@ static int virtio_net_handle_mq(VirtIONet *n, uint8_t cmd,
>>   size_t virtio_net_handle_ctrl_iov(VirtIODevice *vdev,
>>                                     const struct iovec *in_sg, unsigned in_num,
>>                                     const struct iovec *out_sg,
>> -                                  unsigned out_num)
>> +                                  unsigned out_num,
>> +                                 virtio_net_ctrl_ack *status_out)
>>   {
>>       VirtIONet *n = VIRTIO_NET(vdev);
>>       struct virtio_net_ctrl_hdr ctrl;
>> @@ -1514,6 +1515,8 @@ size_t virtio_net_handle_ctrl_iov(VirtIODevice *vdev,
>>       if (iov_size(in_sg, in_num) < sizeof(status) ||
>>           iov_size(out_sg, out_num) < sizeof(ctrl)) {
>>           virtio_error(vdev, "virtio-net ctrl missing headers");
>> +       if (status_out)
>> +               *status_out = status;
>>           return 0;
>>       }
>>
>> @@ -1540,6 +1543,8 @@ size_t virtio_net_handle_ctrl_iov(VirtIODevice *vdev,
>>       assert(s == sizeof(status));
>>
>>       g_free(iov2);
>> +    if (status_out)
>> +           *status_out = status;
>>       return sizeof(status);
>>   }
>>
>> @@ -1555,7 +1560,7 @@ static void virtio_net_handle_ctrl(VirtIODevice *vdev, VirtQueue *vq)
>>           }
>>
>>           written = virtio_net_handle_ctrl_iov(vdev, elem->in_sg, elem->in_num,
>> -                                             elem->out_sg, elem->out_num);
>> +                                             elem->out_sg, elem->out_num, NULL);
>>           if (written > 0) {
>>               virtqueue_push(vq, elem, written);
>>               virtio_notify(vdev, vq);
>> diff --git a/include/hw/virtio/virtio-net.h b/include/hw/virtio/virtio-net.h
>> index ef234ffe7e..da76cc414d 100644
>> --- a/include/hw/virtio/virtio-net.h
>> +++ b/include/hw/virtio/virtio-net.h
>> @@ -224,7 +224,8 @@ struct VirtIONet {
>>   size_t virtio_net_handle_ctrl_iov(VirtIODevice *vdev,
>>                                     const struct iovec *in_sg, unsigned in_num,
>>                                     const struct iovec *out_sg,
>> -                                  unsigned out_num);
>> +                                  unsigned out_num,
>> +                                 virtio_net_ctrl_ack *status);
>>   void virtio_net_set_netclient_name(VirtIONet *n, const char *name,
>>                                      const char *type);
>>
>> diff --git a/net/vhost-vdpa.c b/net/vhost-vdpa.c
>> index de5ed8ff22..c72b338633 100644
>> --- a/net/vhost-vdpa.c
>> +++ b/net/vhost-vdpa.c
>> @@ -638,8 +638,7 @@ static int vhost_vdpa_net_handle_ctrl_avail(VhostShadowVirtqueue *svq,
>>           return VIRTIO_NET_ERR;
>>       }
>>
>> -    status = VIRTIO_NET_ERR;
>> -    virtio_net_handle_ctrl_iov(svq->vdev, &in, 1, &out, 1);
>> +    virtio_net_handle_ctrl_iov(svq->vdev, &in, 1, &out, 1, &status);
>>       if (status != VIRTIO_NET_OK) {
>>           error_report("Bad CVQ processing in model");
>>       }
>> --
>> 2.30.1
>>