linux-kernel - Re: [PATCH] procfs: Fix refcnt leak on proc_self_follow

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]

Message-ID: <20100114015557.GF19799@ZenIV.linux.org.uk>
Date:	Thu, 14 Jan 2010 01:55:57 +0000
From:	Al Viro <viro@...IV.linux.org.uk>
To:	Andrew Morton <akpm@...ux-foundation.org>
Cc:	OGAWA Hirofumi <hirofumi@...l.parknet.co.jp>,
	linux-kernel@...r.kernel.org, linux-fsdevel@...r.kernel.org
Subject: Re: [PATCH] procfs: Fix refcnt leak on proc_self_follow_link()
	error path

On Wed, Jan 13, 2010 at 04:25:04PM -0800, Andrew Morton wrote:
> On Wed, 13 Jan 2010 19:43:01 +0000
> Al Viro <viro@...IV.linux.org.uk> wrote:
> 
> > On Tue, Jan 12, 2010 at 03:38:36AM +0900, OGAWA Hirofumi wrote:
> > > 
> > > If ->follow_link handler return the error, it should decrement
> > > nd->path refcnt.
> > > 
> > > This patch fix it.
> > 
> > It's OK for -stable, but for the next tree... not really.  I'd rather
> > kill vfs_follow_link() uses here and in gfs2; see #untested in vfs-2.6.git
> > for details.
> 
> Confused.  Is #untested planned for 2.6.33?  If not, how do we fix this
> bug in .33?

My preference would be a backport of corresponding bits - they _are_ safe.
I can live with procfs/gfs2 stuff getting into the tree as-is to be
converted later (and note that at least procfs one is fairly old - it goes
back to commit 488e5bc4560d0b510c1ddc451c51a6cc14e3a930
Author: Eric W. Biederman <ebiederm@...ssion.com>
Date:   Fri Feb 8 04:18:34 2008 -0800
    proc: proper pidns handling for /proc/self
so it's prime -stable fodder, whatever form of fix we prefer for that).

For post-2.6.33 we definitely have good reasons to fix that stuff by
providing ->put_link() and switching to nd_set_link() for those.  It
allows to kill fsckloads of symlink body copying when doing open() on
pathname that resolves to a symlink, which is worth a lot.

I'm still not 100% sure which way to go for -stable (and .33 - these are
equivalent in that respect).  Posted patches touch only the codepaths
where we are currently guaranteed to leak and all things equal that'd be
an argument in their favour.  OTOH, the minimal alternative for e.g. gfs2
would be to make buffer allocation unconditional, use nd_set_link() instead
of vfs_follow_link() and leave freeing to put_link().  Which can be followed
by killing special gfs2 ->readlink(), since now generic_readlink() will
work just fine (and gfs2_readlinki() can be folded into gfs2_follow_link(),
simplifying things even more).  In other words, long-term variant is not
much more intrusive and it's just as safe.  So that argument in favour
of posted variant is not particulary strong, especially wrt 2.6.33.

So basically it boils down to doing truly minimal fix vs. doing the same
thing (almost as trivial) we'll do past .33.

FWIW, it may be as simple as "is posted gfs2 patch already in a published
gfs2 tree and if it is, how inconvenient for gfs2 folks would it be to
replace it?"; I'm really ambivalent about that one...
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/