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: <000000000000708b1005f79acf5c@google.com> <CAGxU2F4ZiNEyrZzEJnYjYDz6CxniPGNW7AwyMLPLTxA2UbBWhA@mail.gmail.com>
 <CAGxU2F6m4KWXwOF8StjWbb=S6HRx=GhV_ONDcZxCZsDkvuaeUg@mail.gmail.com>
 <CAGxU2F7XjdKgdKwfZMT-sdJ+JK10p_2zNdaQeGBwm3jpEe1Xaw@mail.gmail.com> <46ba9b55-c6ff-925c-d51a-8da9d1abd2f2@sberdevices.ru>
In-Reply-To: <46ba9b55-c6ff-925c-d51a-8da9d1abd2f2@sberdevices.ru>
From:   Stefano Garzarella <sgarzare@redhat.com>
Date:   Fri, 24 Mar 2023 10:30:48 +0100
Message-ID: <CAGxU2F4KF=YGQkGJygurNyK=KU-pxPomAbr_v3eNvgMmwKQxmQ@mail.gmail.com>
Subject: Re: [syzbot] [kvm?] [net?] [virt?] general protection fault in virtio_transport_purge_skbs
To:     Arseniy Krasnov <avkrasnov@sberdevices.ru>
Cc:     syzbot <syzbot+befff0a9536049e7902e@syzkaller.appspotmail.com>,
        Bobby Eshleman <bobby.eshleman@bytedance.com>,
        Bobby Eshleman <bobby.eshleman@gmail.com>, davem@davemloft.net,
        edumazet@google.com, kuba@kernel.org, kvm@vger.kernel.org,
        linux-kernel@vger.kernel.org, netdev@vger.kernel.org,
        pabeni@redhat.com, stefanha@redhat.com,
        syzkaller-bugs@googlegroups.com,
        virtualization@lists.linux-foundation.org,
        Krasnov Arseniy <oxffffaa@gmail.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Fri, Mar 24, 2023 at 10:10=E2=80=AFAM Arseniy Krasnov
<avkrasnov@sberdevices.ru> wrote:
> On 24.03.2023 12:06, Stefano Garzarella wrote:
> > On Fri, Mar 24, 2023 at 9:55=E2=80=AFAM Stefano Garzarella <sgarzare@re=
dhat.com> wrote:
> >>
> >> On Fri, Mar 24, 2023 at 9:31=E2=80=AFAM Stefano Garzarella <sgarzare@r=
edhat.com> wrote:
> >>>
> >>> Hi Bobby,
> >>> can you take a look at this report?
> >>>
> >>> It seems related to the changes we made to support skbuff.
> >>
> >> Could it be a problem of concurrent access to pkt_queue ?
> >>
> >> IIUC we should hold pkt_queue.lock when we call skb_queue_splice_init(=
)
> >> and remove pkt_list_lock. (or hold pkt_list_lock when calling
> >> virtio_transport_purge_skbs, but pkt_list_lock seems useless now that
> >> we use skbuff)
> >>
> >
> > In the previous patch was missing a hunk, new one attached:
> >
> > #syz test https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linu=
x.git fff5a5e7f528
> >
> > --- a/net/vmw_vsock/vsock_loopback.c
> > +++ b/net/vmw_vsock/vsock_loopback.c
> > @@ -15,7 +15,6 @@
> >  struct vsock_loopback {
> >         struct workqueue_struct *workqueue;
> >
> > -       spinlock_t pkt_list_lock; /* protects pkt_list */
> >         struct sk_buff_head pkt_queue;
> >         struct work_struct pkt_work;
> >  };
> > @@ -32,9 +31,7 @@ static int vsock_loopback_send_pkt(struct sk_buff *sk=
b)
> >         struct vsock_loopback *vsock =3D &the_vsock_loopback;
> >         int len =3D skb->len;
> >
> > -       spin_lock_bh(&vsock->pkt_list_lock);
> >         skb_queue_tail(&vsock->pkt_queue, skb);
> Hello Stefano and Bobby,
>
> Small remark, may be here we can use virtio_vsock_skb_queue_tail() instea=
d of skb_queue_tail().
> skb_queue_tail() disables irqs during spinlock access, while  virtio_vsoc=
k_skb_queue_tail()
> uses spin_lock_bh(). vhost and virtio transports use virtio_vsock_skb_que=
ue_tail().
>

Yep, but this shouldn't be related.
I would make this change in a separate patch. ;-)

Thanks,
Stefano