linux-kernel - Re: [PATCH] vfs: fsmount: add missing mntget()

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Thu, 13 Jun 2019 09:41:07 -0700
From:   Eric Biggers <ebiggers@...nel.org>
To:     Miklos Szeredi <miklos@...redi.hu>
Cc:     David Howells <dhowells@...hat.com>,
        Alexander Viro <viro@...iv.linux.org.uk>,
        linux-fsdevel@...r.kernel.org, Mark Rutland <mark.rutland@....com>,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] vfs: fsmount: add missing mntget()

On Thu, Jun 13, 2019 at 10:47:28AM +0200, Miklos Szeredi wrote:
> On Wed, Jun 12, 2019 at 11:43:13AM -0700, Eric Biggers wrote:
> > From: Eric Biggers <ebiggers@...gle.com>
> > 
> > sys_fsmount() needs to take a reference to the new mount when adding it
> > to the anonymous mount namespace.  Otherwise the filesystem can be
> > unmounted while it's still in use, as found by syzkaller.
> 
> So it needs one count for the file (which dentry_open() obtains) and one for the
> attachment into the anonymous namespace.  The latter one is dropped at cleanup
> time, so your patch appears to be correct at getting that ref.

Yes.

> 
> I wonder why such a blatant use-after-free was missed in normal testing.  RCU
> delayed freeing, I guess?

It's because mount freeing is delayed by task_work_add(), so normally the refcnt
would be dropped to -1 when the file is closed without problems.  The problems
only showed up if you took another reference, e.g. by fchdir().

> 
> How about this additional sanity checking patch?

Seems like a good idea.  Without my fix, the WARNING is triggered by the
following program (no fchdir() needed):

	int main(void)
	{
		int fs;

		fs = syscall(__NR_fsopen, "ramfs", 0);
		syscall(__NR_fsconfig, fs, 6, 0, 0, 0);
		syscall(__NR_fsmount, fs, 0, 0);
	}

Can you send it as a formal patch?

- Eric