Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <000000000000ac6a360592eb26c1@google.com> <d9a957b3-9f0a-20b5-588a-64ca4722d433@rasmusvillemoes.dk>
 <20190919211013.GN5340@magnolia>
In-Reply-To: <20190919211013.GN5340@magnolia>
From:   Andreas Gruenbacher <agruenba@redhat.com>
Date:   Mon, 14 Oct 2019 22:40:44 +0200
Message-ID: <CAHc6FU7drv7r7yu4BzTGKycnKi_wUDGsvND6XyhDLq7B=HCM8g@mail.gmail.com>
Subject: Re: INFO: task hung in pipe_write (2)
To:     "Darrick J. Wong" <darrick.wong@oracle.com>
Cc:     Rasmus Villemoes <linux@rasmusvillemoes.dk>,
        syzbot <syzbot+3c01db6025f26530cf8d@syzkaller.appspotmail.com>,
        linux-fsdevel <linux-fsdevel@vger.kernel.org>,
        LKML <linux-kernel@vger.kernel.org>,
        syzkaller-bugs@googlegroups.com,
        Alexander Viro <viro@zeniv.linux.org.uk>,
        Bob Peterson <rpeterso@redhat.com>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Hi Darrick,

On Thu, Sep 19, 2019 at 11:10 PM Darrick J. Wong
<darrick.wong@oracle.com> wrote:
> On Thu, Sep 19, 2019 at 10:55:44PM +0200, Rasmus Villemoes wrote:
> > On 19/09/2019 19.19, syzbot wrote:
> > > Hello,
> > >
> > > syzbot found the following crash on:
> > >
> > > HEAD commit:    288b9117 Add linux-next specific files for 20190918
> > > git tree:       linux-next
> > > console output: https://syzkaller.appspot.com/x/log.txt?x=17e86645600000
> > > kernel config:  https://syzkaller.appspot.com/x/.config?x=f6126e51304ef1c3
> > > dashboard link:
> > > https://syzkaller.appspot.com/bug?extid=3c01db6025f26530cf8d
> > > compiler:       gcc (GCC) 9.0.0 20181231 (experimental)
> > > syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=11855769600000
> > > C reproducer:   https://syzkaller.appspot.com/x/repro.c?x=143580a1600000
> > >
> > > The bug was bisected to:
> > >
> > > commit cfb864757d8690631aadf1c4b80022c18ae865b3
> > > Author: Darrick J. Wong <darrick.wong@oracle.com>
> > > Date:   Tue Sep 17 16:05:22 2019 +0000
> > >
> > >     splice: only read in as much information as there is pipe buffer space
> >
> > The middle hunk (the one before splice_pipe_to_pipe()) accesses
> > opipe->{buffers, nrbufs}, but opipe is not locked at that point. So
> > maybe we end up passing len==0, which seems (once there's room in opipe)
> > it would put a zero-length pipe_buffer in opipe - and that probably
> > violates an invariant somewhere.
> >
> > But does the splice_pipe_to_pipe() case even need that extra logic?
> > Doesn't it handle short writes correctly already?
>
> Yep.  I missed the part where splice_pipe_to_pipe is already perfectly
> capable of detecting insufficient space in opipe and kicking opipe's
> readers to clear out the buffer.  So that hunk isn't needed, and now I'm
> wondering how in the other clause we return 0 from wait_for_space yet
> still don't have buffer space...
>
> Oh well, back to the drawing board.  Good catch, though now it's become
> painfully clear that xfstests lacks rigorous testing of splice()...

have you had any luck figuring out how to fix this? We're still
suffering from the regression I've reported a while ago (*).

If not, I wonder if reverting commit 8f67b5adc030 would make sense for now.

* https://lore.kernel.org/linux-fsdevel/CAHpGcM+WQYFHOOC8SzKq+=DuHVZ4fw4RHLTMUDN-o6GX3YtGvQ@mail.gmail.com/T/#u

Thanks,
Andreas