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: <000000000000701c3305818e4814@google.com> <CAOQ4uxhhRs6kBH2rgbPyjd=tXoxxuJG76Y2vpCYmD0c0HzLwhA@mail.gmail.com>
 <CAJfpegvjLgXa1HOCZkz6RKup7UdioMGyY7W4miyqcRtn9PL4QA@mail.gmail.com>
 <CAJfpegsJo8X+efgWSa4QwSGpAGADpoossZZg=m7x5ApPXhxCoA@mail.gmail.com>
 <CAOQ4uxj0nKK1csWGxxLT-EvC3hv8Zz5-HpvMF_a2r2iVcLXt1A@mail.gmail.com>
 <CAJfpegsadeK=3EON9HgZ8rLCDWCJXR9QrXxA3ZZjdjzq6FDRjw@mail.gmail.com> <CAOQ4uxhnK2-1jsNQPkfzF84-ZcLgroakSd4JDR62mM-h1Z6r4Q@mail.gmail.com>
In-Reply-To: <CAOQ4uxhnK2-1jsNQPkfzF84-ZcLgroakSd4JDR62mM-h1Z6r4Q@mail.gmail.com>
From:   Miklos Szeredi <miklos@szeredi.hu>
Date:   Mon, 11 Feb 2019 16:40:37 +0100
Message-ID: <CAJfpegstnt2tyhZuQ+DnR3rfn7B6xo9JhcBnb=GJV11i8013gA@mail.gmail.com>
Subject: Re: possible deadlock in pipe_lock (2)
To:     Amir Goldstein <amir73il@gmail.com>
Cc:     Jan Kara <jack@suse.cz>,
        linux-fsdevel <linux-fsdevel@vger.kernel.org>,
        linux-kernel <linux-kernel@vger.kernel.org>,
        syzkaller-bugs@googlegroups.com, Al Viro <viro@zeniv.linux.org.uk>,
        syzbot <syzbot+31d8b84465a7cbfd8515@syzkaller.appspotmail.com>,
        overlayfs <linux-unionfs@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Mon, Feb 11, 2019 at 4:07 PM Amir Goldstein <amir73il@gmail.com> wrote:
>
> On Mon, Feb 11, 2019 at 4:33 PM Miklos Szeredi <miklos@szeredi.hu> wrote:
> >
> > On Mon, Feb 11, 2019 at 2:08 PM Amir Goldstein <amir73il@gmail.com> wrote:
> > >
> > > On Mon, Feb 11, 2019 at 2:37 PM Miklos Szeredi <miklos@szeredi.hu> wrote:
> > > >
> > > > On Mon, Feb 11, 2019 at 1:06 PM Miklos Szeredi <miklos@szeredi.hu> wrote:
> > > > >
> > > > > On Mon, Feb 11, 2019 at 8:38 AM Amir Goldstein <amir73il@gmail.com> wrote:
> > > > > >
> > > > > > On Sun, Feb 10, 2019 at 8:23 PM syzbot
> > > > > > <syzbot+31d8b84465a7cbfd8515@syzkaller.appspotmail.com> wrote:
> > > > >
> > > > > > > -> #1 (&ovl_i_mutex_key[depth]){+.+.}:
> > > > > > >         down_write+0x38/0x90 kernel/locking/rwsem.c:70
> > > > > > >         inode_lock include/linux/fs.h:757 [inline]
> > > > > > >         ovl_write_iter+0x148/0xc20 fs/overlayfs/file.c:231
> > > > > > >         call_write_iter include/linux/fs.h:1863 [inline]
> > > > > > >         new_sync_write fs/read_write.c:474 [inline]
> > > > > > >         __vfs_write+0x613/0x8e0 fs/read_write.c:487
> > > > > > > kobject: 'loop4' (000000009e2b886d): kobject_uevent_env
> > > > > > >         __kernel_write+0x110/0x3b0 fs/read_write.c:506
> > > > > > >         write_pipe_buf+0x15d/0x1f0 fs/splice.c:797
> > > > > > >         splice_from_pipe_feed fs/splice.c:503 [inline]
> > > > > > >         __splice_from_pipe+0x39a/0x7e0 fs/splice.c:627
> > > > > > >         splice_from_pipe+0x108/0x170 fs/splice.c:662
> > > > > > >         default_file_splice_write+0x3c/0x90 fs/splice.c:809
> > > > >
> > > > > Irrelevant to the lockdep splat, but why isn't there an
> > > > > ovl_splice_write() that just recurses into realfile->splice_write()?
> > > > > Sounds like a much more efficient way to handle splice read and
> > > > > write...
> > > > >
> > > > > [...]
> > > > >
> > > > > > Miklos,
> > > > > >
> > > > > > Its good that this report popped up again, because I went to
> > > > > > look back at my notes from previous report [1].
> > > > > > If I was right in my previous analysis then we must have a real
> > > > > > deadlock in current "lazy copy up" WIP patches. Right?
> > > > >
> > > > > Hmm, AFAICS this circular dependency translated into layman's terms:
> > > > >
> > > > > pipe lock -> ovl inode lock  (splice to ovl file)
> > > > >
> > > > > ovl inode lock -> upper freeze lock (truncate of ovl file)
> > > > >
> > > > > upper freeze lock -> pipe lock (splice to upper file)
> > > >
> > > > So what can we do with this?
> > > >
> > > > The "freeze lock -> inode lock" dependency is fixed.   This is
> > > > reversed in overlay to "ovl inode lock -> upper freeze lock", which is
> > > > okay, because this is a nesting that cannot be reversed.   But in
> > > > splice the pipe locks comes in between: "freeze lock -> pipe lock ->
> > > > inode lock" which breaks this nesting direction and creates a true
> > > > reverse dependency between ovl inode lock and upper freeze lock.
> > > >
> > > > The only way I see this could be fixed is to move the freeze lock
> > > > inside the pipe lock.  But that would mean splice/sendfile/etc could
> > > > be frozen with the pipe lock held.  It doesn't look nice.
> > > >
> > > > Any other ideas?
> > > >
> > >
> > > [CC Jan]
> > >
> > > I think we are allowed to file_start_write_trylock(upper)
> > > before ovl_inode_lock(). This in ONLY needed to cover the corner
> > > case of upper being frozen in between "upper freeze lock -> pipe lock"
> > > and thread B being in between "ovl inode lock -> upper freeze lock".
> > > Is it OK to return a transient error in this corner copy up case?
> >
> > This case shouldn't happen assuming adherence to the "upper shall not
> > be modified while part of the overlay" rule.
> >
>
> Right. And unfreezing upper should untangle this deadlock,
> because both thread A and B are taking a shared sb_writers lock.

I don't think that'll work.   The deadlock would involve freezing for
sure, otherwise sb_start_write() won't block.  But there's no way to
cancel sb_wait_write() once it's called, so the deadlock is permanent.

> > Side note: I don't see that it has anything to do with copy-up, but I
> > may be missing something.
> >
>
> You are right. I was confusing your "ovl inode lock" with ovl_inode_lock(),
> but the latter is taken after upper freeze lock, so irrelevant.
>
> > My other thought is that perhaps sb_start_write() should invoke
> > s_ops->start_write() so that overlay can do the freeze protection on
> > the upper early.
> >
>
> Sorry, I don't see how that solves anything expect for the lockdep
> warning. In both cases threads A and B would block until upper
> in unfrozen, only without a lockdep warning.
> Also, I am quite sure that taking upper freeze lock early will generate
> many new lockdep warnings.

My thinking was to make the lock order:

  ovl freeze lock -> upper freeze lock -> ovl inode lock -> upper inode lock


> Anyway, what about the recursive do_splice_direct() issue
> with lazy copy up, do you have an easy solution for that?

Not sure.  I think splice_direct_to_actor() should be able to deal
with it.  E.g.

    pipe = current->splice_pipe;
    if (unlikely(!pipe)) {
        pipe = alloc_pipe_info();
        ...
    } else {
        current->splice_pipe = NULL;
    }

    ... do the actual splicing ...

    if (!current->splice_pipe) {
        current->splice_pipe = pipe
    } else {
        free_pipe_info(pipe);
    }

Thanks,
Miklos