Re: [PATCH net-next v3] af_unix: Fix use-after-free crashes

[Tim Chen <tim.c.chen@linux.intel.com>]

On Fri, 2011-09-09 at 08:51 +0200, Eric Dumazet wrote:
> Le jeudi 08 septembre 2011 à 01:37 -0700, Tim Chen a écrit :
> > On Thu, 2011-09-08 at 15:21 +0200, Eric Dumazet wrote:
> > > Le jeudi 08 septembre 2011 à 11:59 +0200, Sedat Dilek a écrit :
> > > 
> > > > I have tested this fixup patch on i386.
> > > > Can we have a separate patch with corrected descriptive text?
> > > > 
> > > > Thanks to all involved people.
> > > 
> > > Here it is :
> > > 
> > > [PATCH net-next v3] af_unix: Fix use-after-free crashes
> > > 
> > > Commit 0856a30409 (Scm: Remove unnecessary pid & credential references
> > > in Unix socket's send and receive path) introduced an use-after-free
> > > bug.
> > > 
> > > We are allowed to steal the references to pid/cred only in the last skb
> > > sent from unix_stream_sendmsg(), because first skbs might be consumed by
> > > the receiver before we finish our sendmsg() call.
> > > 
> > > Remove scm_release() helper, since its cleaner to clear pid/cred fields
> > > in unix_scm_to_skb() when we steal them.
> > > 
> > > Based on prior patches from Yan Zheng and Tim Chen
> > > 
> > > Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
> > > Reported-by: Jiri Slaby <jirislaby@gmail.com>
> > > Tested-by: Sedat Dilek <sedat.dilek@googlemail.com>
> > > Tested-by: Valdis Kletnieks <Valdis.Kletnieks@vt.edu>
> > > ---
> > 
> > Thanks.
> > 
> > Acked-by: Tim Chen <tim.c.chen@linux.intel.com>
> > 
> 
> Now we have to fix a bug in unix_stream_recvmsg() as well.
> 

Thanks for your careful review to catch the bugs.  Wish I had thought
about the stream msg case more in my original patch.

> consume_skb() call actually releases pid/cred references, and we can use
> them after their eventual freeing.
> 
> Keep also in mind that receiver can provides a too short user buffer,
> and skb can be put back to head of sk_receive_queue
> 
> Tim, your 0856a304091b33 commit introduced a lot of bugs, I was right
> asking a revert.
> 
> If we revert your patch, my litle patch (af_unix: dont send
> SCM_CREDENTIALS by default) is enough to solve performance problems.
> 

But the regression where we do send SCM_CREDENTIALS is not addressed,
right?  I don't mind reverting my patch if there's a better solution.

Tim