Re: [PATCH] virtio_ring: add an error code check in virtqueue_resize

From: "Michael S. Tsirkin" <mst@redhat.com>
To: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Cc: Su Hui <suhui@nfschina.com>,
	kernel-janitors@vger.kernel.org, linux-kernel@vger.kernel.org,
	virtualization@lists.linux-foundation.org
Subject: Re: [PATCH] virtio_ring: add an error code check in virtqueue_resize
Date: Mon, 23 Oct 2023 07:17:30 -0400	[thread overview]
Message-ID: <20231023071548-mutt-send-email-mst@kernel.org> (raw)
In-Reply-To: <1698054722.2894735-1-xuanzhuo@linux.alibaba.com>

On Mon, Oct 23, 2023 at 05:52:02PM +0800, Xuan Zhuo wrote:
> On Mon, 23 Oct 2023 17:50:46 +0800, Su Hui <suhui@nfschina.com> wrote:
> > On 2023/10/23 13:46, Xuan Zhuo wrote:
> > >>>>>>>> Well, what are the cases where it can happen practically?
> > >>>>>>> Device error. Such as vp_active_vq()
> > >>>>>>>
> > >>>>>>> Thanks.
> > >>>>>> Hmm interesting. OK. But do callers know to recover?
> > >>>>> No.
> > >>>>>
> > >>>>> So I think WARN + broken is suitable.
> > >>>>>
> > >>>>> Thanks.
> > >>>> Sorry for the late, is the following code okay?
> > >>>>
> > >>>> @@ -2739,7 +2739,7 @@ int virtqueue_resize(struct virtqueue *_vq, u32 num,
> > >>>>                         void (*recycle)(struct virtqueue *vq, void *buf))
> > >>>>     {
> > >>>>            struct vring_virtqueue *vq = to_vvq(_vq);
> > >>>> -       int err;
> > >>>> +       int err, err_reset;
> > >>>>
> > >>>>            if (num > vq->vq.num_max)
> > >>>>                    return -E2BIG;
> > >>>> @@ -2759,7 +2759,15 @@ int virtqueue_resize(struct virtqueue *_vq, u32 num,
> > >>>>            else
> > >>>>                    err = virtqueue_resize_split(_vq, num);
> > >>>>
> > >>>> -       return virtqueue_enable_after_reset(_vq);
> > >>>> +       err_reset = virtqueue_enable_after_reset(_vq);
> > >>>> +
> > >>>> +       if (err) {
> > >>> No err.
> > >>>
> > >>> err is not important.
> > >>> You can remove that.
> > >> Emm, I'm a little confused that which code should I remove ?
> > >>
> > >>
> > >> like this:
> > >> 	if (vq->packed_ring)
> > >> 		virtqueue_resize_packed(_vq, num);
> > >> 	else
> > >> 		virtqueue_resize_split(_vq, num);
> > >>
> > >> And we should set broken and warn inside virtqueue_enable_after_reset()?
> >
> > In my opinion, we should return the error code of virtqueue_resize_packed() / virtqueue_resize_split().
> > But if this err is not important, this patch makes no sense.
> > Maybe I misunderstand somewhere...
> > If you think it's worth sending a patch, you can send it :).(I'm not familiar with this code).
> 
> OK.
> 
> Thanks.

I would first try to recover by re-enabling.
If that fails we can set broken.

> 
> >
> > Thanks,
> > Su Hui
> >