linux-kernel - Re: [3.10.4] NFS locking panic, plus persisting NFS shutdown panic from 3.9.*

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 for Android: free password hash cracker in your pocket

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

Message-ID: <20130805143311.4ae59067@tlielax.poochiereds.net>
Date:	Mon, 5 Aug 2013 14:33:11 -0400
From:	Jeff Layton <jlayton@...hat.com>
To:	"Myklebust, Trond" <Trond.Myklebust@...app.com>
Cc:	Nix <nix@...eri.org.uk>,
	Toralf Förster <toralf.foerster@....de>,
	Oleg Nesterov <oleg@...hat.com>,
	NFS list <linux-nfs@...r.kernel.org>,
	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
	"dhowells@...hat.com" <dhowells@...hat.com>
Subject: Re: [3.10.4] NFS locking panic, plus persisting NFS shutdown panic
 from 3.9.*

On Mon, 5 Aug 2013 18:18:03 +0000
"Myklebust, Trond" <Trond.Myklebust@...app.com> wrote:

> On Mon, 2013-08-05 at 13:37 -0400, Jeff Layton wrote:
> > On Mon, 5 Aug 2013 16:15:01 +0000
> > "Myklebust, Trond" <Trond.Myklebust@...app.com> wrote:
> > 
> > > From 3c50ba80105464a28d456d9a1e0f1d81d4af92a8 Mon Sep 17 00:00:00 2001
> > > From: Trond Myklebust <Trond.Myklebust@...app.com>
> > > Date: Mon, 5 Aug 2013 12:06:12 -0400
> > > Subject: [PATCH] LOCKD: Don't call utsname()->nodename from
> > >  nlmclnt_setlockargs
> > > MIME-Version: 1.0
> > > Content-Type: text/plain; charset=UTF-8
> > > Content-Transfer-Encoding: 8bit
> > > 
> > > Firstly, nlmclnt_setlockargs can be called from a reclaimer thread, in
> > > which case we're in entirely the wrong namespace.
> > > Secondly, commit 8aac62706adaaf0fab02c4327761561c8bda9448 (move
> > > exit_task_namespaces() outside of exit_notify()) now means that
> > > exit_task_work() is called after exit_task_namespaces(), which
> > > triggers an Oops when we're freeing up the locks.
> > > 
> > > Signed-off-by: Trond Myklebust <Trond.Myklebust@...app.com>
> > > Cc: Toralf Förster <toralf.foerster@....de>
> > > Cc: Oleg Nesterov <oleg@...hat.com>
> > > Cc: Nix <nix@...eri.org.uk>
> > > Cc: Jeff Layton <jlayton@...hat.com>
> > > ---
> > >  fs/lockd/clntproc.c | 5 +++--
> > >  1 file changed, 3 insertions(+), 2 deletions(-)
> > > 
> > > diff --git a/fs/lockd/clntproc.c b/fs/lockd/clntproc.c
> > > index 9760ecb..acd3947 100644
> > > --- a/fs/lockd/clntproc.c
> > > +++ b/fs/lockd/clntproc.c
> > > @@ -125,14 +125,15 @@ static void nlmclnt_setlockargs(struct nlm_rqst *req, struct file_lock *fl)
> > >  {
> > >  	struct nlm_args	*argp = &req->a_args;
> > >  	struct nlm_lock	*lock = &argp->lock;
> > > +	char *nodename = req->a_host->h_rpcclnt->cl_nodename;
> > >  
> > >  	nlmclnt_next_cookie(&argp->cookie);
> > >  	memcpy(&lock->fh, NFS_FH(file_inode(fl->fl_file)), sizeof(struct nfs_fh));
> > > -	lock->caller  = utsname()->nodename;
> > > +	lock->caller  = nodename;
> > >  	lock->oh.data = req->a_owner;
> > >  	lock->oh.len  = snprintf(req->a_owner, sizeof(req->a_owner), "%u@%s",
> > >  				(unsigned int)fl->fl_u.nfs_fl.owner->pid,
> > > -				utsname()->nodename);
> > > +				nodename);
> > >  	lock->svid = fl->fl_u.nfs_fl.owner->pid;
> > >  	lock->fl.fl_start = fl->fl_start;
> > >  	lock->fl.fl_end = fl->fl_end;
> > 
> > Looks good to me...
> > 
> > Reviewed-by: Jeff Layton <jlayton@...hat.com>
> > 
> > Trond, any thoughts on the other oops that Nix posted? The issue there
> > seems to be that we're trying to do the pathwalk to the rpcbind unix
> > socket from exit_task_work(), but that's happening after we've already
> > called exit_fs().
> > 
> > The trivial answer seems to be to simply call exit_task_work() before
> > exit_fs() there, but it seems like we ought to be doing the upcall to
> > rpcbind in a mount namespace from which we know we can reach the
> > socket...
> 
> Isn't it enough to just do the same thing as we did for gss proxy? i.e.
> set the RPC_CLNT_CREATE_NO_IDLE_TIMEOUT flag.
> 
> See attachment.

Yeah, that looks like a reasonable thing to do...

OTOH, Is there any other way for a unix socket to end up disconnected
other than if we were to close it? Maybe if rpcbind stopped, the socket
unlinked and recreated and then started again?

If so then you still could potentially end up in this situation even if
you didn't autoclose it.

-- 
Jeff Layton <jlayton@...hat.com>
--
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/