On 24-03-21, 08:53, Jie Deng wrote:
>
> On 2021/3/23 17:38, Viresh Kumar wrote:
> > On 23-03-21, 14:31, Viresh Kumar wrote:
> > > On 23-03-21, 22:19, Jie Deng wrote:
> > > > +static int virtio_i2c_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs, int num)
> > > > +{
> > > > + struct virtio_i2c *vi = i2c_get_adapdata(adap);
> > > > + struct virtqueue *vq = vi->vq;
> > > > + struct virtio_i2c_req *reqs;
> > > > + unsigned long time_left;
> > > > + int ret, nr;
> > > > +
> > > > + reqs = kcalloc(num, sizeof(*reqs), GFP_KERNEL);
> > > > + if (!reqs)
> > > > + return -ENOMEM;
> > > > +
> > > > + mutex_lock(&vi->lock);
> > > > +
> > > > + ret = virtio_i2c_send_reqs(vq, reqs, msgs, num);
> > > > + if (ret == 0)
> > > > + goto err_unlock_free;
> > > > +
> > > > + nr = ret;
> > > > + reinit_completion(&vi->completion);
> > > I think I may have found a possible bug here. This reinit_completion() must
> > > happen before we call virtio_i2c_send_reqs(). It is certainly possible (surely
> > > in corner cases) that virtio_i2c_msg_done() may get called right after
> > > virtio_i2c_send_reqs() and before we were able to call reinit_completion(). And
> > > in that case we will never see the completion happen at all.
> > >
> > > > + virtqueue_kick(vq);
> > I may have misread this. Can the actually start before virtqueue_kick() is
> > called ?
I didn't write it properly here. I wanted to say,
"Can the _transfer_ actually start before virtqueue_kick() is called ?"
> No. It starts when wait_for_completion_timeout is called.
No, the transfer doesn't have anything to do with wait_for_completion_timeout().
And if complete() gets called before wait_for_completion_timeout() is called,
then wait_for_completion_timeout() will simply return back.
> So it should be fine here.
>
>
> > If not, then completion may be fine where it is.
> >
--
viresh
