Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20230808051110.3492693-1-yuanyaogoog@chromium.org>
 <CACGkMEt53ziY_bmgJHVdJ6pkppTyVqKX3=Czygv+yhJR8_KiFA@mail.gmail.com> <CAOJyEHaXqmHStJnHrT0H4QsTJBxjBxVe+33EuWm9H3wApPKtxQ@mail.gmail.com>
In-Reply-To: <CAOJyEHaXqmHStJnHrT0H4QsTJBxjBxVe+33EuWm9H3wApPKtxQ@mail.gmail.com>
From:   Jason Wang <jasowang@redhat.com>
Date:   Tue, 8 Aug 2023 14:05:34 +0800
Message-ID: <CACGkMEuiVnLFRRDHaTH3Jnpr+znU9L33gLyRynbBabVqumN2ug@mail.gmail.com>
Subject: Re: [PATCH] virtio_ring: fix avail_wrap_counter in virtqueue_add_packed
To:     Yuan Yao <yuanyaogoog@chromium.org>
Cc:     LKML <linux-kernel@vger.kernel.org>,
        Keiichi Watanabe <keiichiw@chromium.org>,
        Daniel Verkamp <dverkamp@chromium.org>,
        Takaya Saeki <takayas@chromium.org>,
        Junichi Uekawa <uekawa@chromium.org>,
        "David S. Miller" <davem@davemloft.net>,
        "Michael S. Tsirkin" <mst@redhat.com>,
        Tiwei Bie <tiwei.bie@intel.com>,
        Xuan Zhuo <xuanzhuo@linux.alibaba.com>,
        virtualization@lists.linux-foundation.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Tue, Aug 8, 2023 at 1:59=E2=80=AFPM Yuan Yao <yuanyaogoog@chromium.org> =
wrote:
>
> Since there is a check for 0-length chain in code;BUG_ON(total_sg =3D=3D =
0), we won=E2=80=99t get a 0-length chain in practice. So, I think use (i <=
=3D head) makes the commit as small as possible.

Ok, offered ack in other mail.

Thanks

>
>
> Best regards
>
>
>
> On Tue, Aug 8, 2023 at 2:43=E2=80=AFPM Jason Wang <jasowang@redhat.com> w=
rote:
>>
>> On Tue, Aug 8, 2023 at 1:11=E2=80=AFPM Yuan Yao <yuanyaogoog@chromium.or=
g> wrote:
>> >
>> > In current packed virtqueue implementation, the avail_wrap_counter won=
't
>> > flip, in the case when the driver supplies a descriptor chain with a
>> > length equals to the queue size; total_sg =3D=3D vq->packed.vring.num.
>> >
>> > Let=E2=80=99s assume the following situation:
>> > vq->packed.vring.num=3D4
>> > vq->packed.next_avail_idx: 1
>> > vq->packed.avail_wrap_counter: 0
>> >
>> > Then the driver adds a descriptor chain containing 4 descriptors.
>> >
>> > We expect the following result with avail_wrap_counter flipped:
>> > vq->packed.next_avail_idx: 1
>> > vq->packed.avail_wrap_counter: 1
>> >
>> > But, the current implementation gives the following result:
>> > vq->packed.next_avail_idx: 1
>> > vq->packed.avail_wrap_counter: 0
>> >
>> > To reproduce the bug, you can set a packed queue size as small as
>> > possible, so that the driver is more likely to provide a descriptor
>> > chain with a length equal to the packed queue size. For example, in
>> > qemu run following commands:
>> > sudo qemu-system-x86_64 \
>> > -enable-kvm \
>> > -nographic \
>> > -kernel "path/to/kernel_image" \
>> > -m 1G \
>> > -drive file=3D"path/to/rootfs",if=3Dnone,id=3Ddisk \
>> > -device virtio-blk,drive=3Ddisk \
>> > -drive file=3D"path/to/disk_image",if=3Dnone,id=3Drwdisk \
>> > -device virtio-blk,drive=3Drwdisk,packed=3Don,queue-size=3D4,\
>> > indirect_desc=3Doff \
>> > -append "console=3DttyS0 root=3D/dev/vda rw init=3D/bin/bash"
>> >
>> > Inside the VM, create a directory and mount the rwdisk device on it. T=
he
>> > rwdisk will hang and mount operation will not complete.
>> >
>> > This commit fixes the wrap counter error by flipping the
>> > packed.avail_wrap_counter, when start of descriptor chain equals to th=
e
>> > end of descriptor chain (head =3D=3D i).
>> >
>> > Fixes: 1ce9e6055fa0 ("virtio_ring: introduce packed ring support")
>> > Signed-off-by: Yuan Yao <yuanyaogoog@chromium.org>
>> > ---
>> >
>> >  drivers/virtio/virtio_ring.c | 2 +-
>> >  1 file changed, 1 insertion(+), 1 deletion(-)
>> >
>> > diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring=
.c
>> > index c5310eaf8b46..da1150d127c2 100644
>> > --- a/drivers/virtio/virtio_ring.c
>> > +++ b/drivers/virtio/virtio_ring.c
>> > @@ -1461,7 +1461,7 @@ static inline int virtqueue_add_packed(struct vi=
rtqueue *_vq,
>> >                 }
>> >         }
>> >
>> > -       if (i < head)
>> > +       if (i <=3D head)
>> >                 vq->packed.avail_wrap_counter ^=3D 1;
>>
>> Would it be better to move the flipping to the place where we flip
>> avail_used_flags?
>>
>>                         if ((unlikely(++i >=3D vq->packed.vring.num))) {
>>                                 i =3D 0;
>>                                 vq->packed.avail_used_flags ^=3D
>>                                         1 << VRING_PACKED_DESC_F_AVAIL |
>>                                         1 << VRING_PACKED_DESC_F_USED;
>>                         }
>>
>> Thanks
>>
>> >
>> >         /* We're using some buffers from the free list. */
>> > --
>> > 2.41.0.640.ga95def55d0-goog
>> >
>>