Re: [PATCH 7/8] VirtIOBlock: protect rq with its own lock

[Stefan Hajnoczi]

On Thu, Jun 09, 2022 at 10:37:26AM -0400, Emanuele Giuseppe Esposito wrote:
> @@ -946,17 +955,20 @@ static void virtio_blk_reset(VirtIODevice *vdev)
>       * stops all Iothreads.
>       */
>      blk_drain(s->blk);
> +    aio_context_release(ctx);
>  
>      /* We drop queued requests after blk_drain() because blk_drain() itself 
> can
>       * produce them. */
> +    qemu_mutex_lock(&s->req_mutex);
>      while (s->rq) {
>          req = s->rq;
>          s->rq = req->next;
> +        qemu_mutex_unlock(&s->req_mutex);
>          virtqueue_detach_element(req->vq, &req->elem, 0);
>          virtio_blk_free_request(req);
> +        qemu_mutex_lock(&s->req_mutex);

Why is req_mutex dropped temporarily? At this point we don't really need
the req_mutex (all I/O should be stopped and drained), but maybe we
should do:

  WITH_QEMU_MUTEX(&s->req_mutex) {
      req = s->rq;
      s->rq = NULL;
  }

  ...process req list...

Otherwise:
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>

signature.asc
Description: PGP signature