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: <0000000000001777f605fce42c5f@google.com> <20230530072310-mutt-send-email-mst@kernel.org>
 <CAGxU2F7O7ef3mdvNXtiC0VtWiS2DMnoiGwSR=Z6SWbzqcrBF-g@mail.gmail.com>
 <CAGxU2F7HK5KRggiY7xnKHeXFRXJmqcKbjf3JnXC3mbmn9xqRtw@mail.gmail.com>
 <e4589879-1139-22cc-854f-fed22cc18693@oracle.com> <6p7pi6mf3db3gp3xqarap4uzrgwlzqiz7wgg5kn2ep7hvrw5pg@wxowhbw4e7w7>
 <035e3423-c003-3de9-0805-2091b9efb45d@oracle.com>
In-Reply-To: <035e3423-c003-3de9-0805-2091b9efb45d@oracle.com>
From:   Stefano Garzarella <sgarzare@redhat.com>
Date:   Wed, 31 May 2023 09:27:54 +0200
Message-ID: <CAGxU2F5oTLY_weLixRKMQVqmjpDG_09yL6tS2rF8mwJ7K+xP0Q@mail.gmail.com>
Subject: Re: [syzbot] [kvm?] [net?] [virt?] general protection fault in vhost_work_queue
To:     michael.christie@oracle.com
Cc:     "Michael S. Tsirkin" <mst@redhat.com>,
        syzbot <syzbot+d0d442c22fa8db45ff0e@syzkaller.appspotmail.com>,
        jasowang@redhat.com, kvm@vger.kernel.org,
        linux-kernel@vger.kernel.org, netdev@vger.kernel.org,
        syzkaller-bugs@googlegroups.com,
        virtualization@lists.linux-foundation.org, stefanha@redhat.com
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Tue, May 30, 2023 at 6:30=E2=80=AFPM <michael.christie@oracle.com> wrote=
:
>
> On 5/30/23 11:17 AM, Stefano Garzarella wrote:
> > On Tue, May 30, 2023 at 11:09:09AM -0500, Mike Christie wrote:
> >> On 5/30/23 11:00 AM, Stefano Garzarella wrote:
> >>> I think it is partially related to commit 6e890c5d5021 ("vhost: use
> >>> vhost_tasks for worker threads") and commit 1a5f8090c6de ("vhost: mov=
e
> >>> worker thread fields to new struct"). Maybe that commits just
> >>> highlighted the issue and it was already existing.
> >>
> >> See my mail about the crash. Agree with your analysis about worker->vt=
sk
> >> not being set yet. It's a bug from my commit where I should have not s=
et
> >> it so early or I should be checking for
> >>
> >> if (dev->worker && worker->vtsk)
> >>
> >> instead of
> >>
> >> if (dev->worker)
> >
> > Yes, though, in my opinion the problem may persist depending on how the
> > instructions are reordered.
>
> Ah ok.
>
> >
> > Should we protect dev->worker() with an RCU to be safe?
>
> For those multiple worker patchsets Jason had asked me about supporting
> where we don't have a worker while we are swapping workers around. To do
> that I had added rcu around the dev->worker. I removed it in later patchs=
ets
> because I didn't think anyone would use it.
>
> rcu would work for your case and for what Jason had requested.

Yeah, so you already have some patches?

Do you want to send it to solve this problem?

Thanks,
Stefano