Re: [PATCH] /dev/mem: Bail out upon SIGKILL when reading memory.

From: Dmitry Vyukov <dvyukov@google.com>
To: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>,
	Arnd Bergmann <arnd@arndb.de>,
	LKML <linux-kernel@vger.kernel.org>,
	syzbot <syzbot+8ab2d0f39fb79fe6ca40@syzkaller.appspotmail.com>,
	Eric Biggers <ebiggers@kernel.org>
Subject: Re: [PATCH] /dev/mem: Bail out upon SIGKILL when reading memory.
Date: Thu, 22 Aug 2019 10:11:20 -0700	[thread overview]
Message-ID: <CACT4Y+Z0yCAwie83Oqd7XBNgQjWtEkuEg5WJCd6rW-ZMWqosxg@mail.gmail.com> (raw)
In-Reply-To: <20190822164249.GA12551@kroah.com>

On Thu, Aug 22, 2019 at 9:42 AM Greg Kroah-Hartman
<gregkh@linuxfoundation.org> wrote:
>
> On Thu, Aug 22, 2019 at 11:00:59PM +0900, Tetsuo Handa wrote:
> > On 2019/08/22 22:35, Greg Kroah-Hartman wrote:
> > > On Thu, Aug 22, 2019 at 06:59:25PM +0900, Tetsuo Handa wrote:
> > >> Tetsuo Handa wrote:
> > >>> Greg Kroah-Hartman wrote:
> > >>>> Oh, nice!  This shouldn't break anything that is assuming that the read
> > >>>> will complete before a signal is delivered, right?
> > >>>>
> > >>>> I know userspace handling of "short" reads is almost always not there...
> > >>>
> > >>> Since this check will give up upon SIGKILL, userspace won't be able to see
> > >>> the return value from read(). Thus, returning 0 upon SIGKILL will be safe. ;-)
> > >>> Maybe we also want to add cond_resched()...
> > >>>
> > >>> By the way, do we want similar check on write_mem() side?
> > >>> If aborting "write to /dev/mem" upon SIGKILL (results in partial write) is
> > >>> unexpected, we might want to ignore SIGKILL for write_mem() case.
> > >>> But copying data from killed threads (especially when killed by OOM killer
> > >>> and userspace memory is reclaimed by OOM reaper before write_mem() returns)
> > >>> would be after all unexpected. Then, it might be preferable to check SIGKILL
> > >>> on write_mem() side...
> > >>>
> > >>
> > >> Ha, ha. syzbot reported the same problem using write_mem().
> > >> https://syzkaller.appspot.com/text?tag=CrashLog&x=1018055a600000
> > >> We want fatal_signal_pending() check on both sides.
> > >
> > > Ok, want to send a patch for that?
> >
> > Yes. But before sending a patch, I'm trying to dump values using debug printk().
> >
> > >
> > > And does anything use /dev/mem anymore?  I think X stopped using it a
> > > long time ago.
> > >
> > >> By the way, write_mem() worries me whether there is possibility of replacing
> > >> kernel code/data with user-defined memory data supplied from userspace.
> > >> If write_mem() were by chance replaced with code that does
> > >>
> > >>    while (1);
> > >>
> > >> we won't be able to return from write_mem() even if we added fatal_signal_pending() check.
> > >> Ditto for replacing local variables with unexpected values...
> > >
> > > I'm sorry, I don't really understand what you mean here, but I haven't
> > > had my morning coffee...  Any hints as to an example?
> >
> > Probably similar idea: "lockdown: Restrict /dev/{mem,kmem,port} when the kernel is locked down"
> >
> > https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/drivers/char/mem.c?h=next-20190822&id=9b9d8dda1ed72e9bd560ab0ca93d322a9440510e
> >
> > Then, syzbot might want to blacklist writing to /dev/mem .
>
> syzbot should probably blacklist that now, you can do a lot of bad
> things writing to that device node :(

Agree. It wasn't supposed to reach it, but it figured out how to mount
devfs and then open "./mem"  bypassing all checks. Fortunately there
is a config to disable /dev/mem, so we are going to turn it off.