From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: linux-nfs-owner@vger.kernel.org
Received: from fieldses.org ([174.143.236.118]:49105 "EHLO fieldses.org"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S1756511Ab2FYUoC (ORCPT <rfc822;linux-nfs@vger.kernel.org>);
	Mon, 25 Jun 2012 16:44:02 -0400
Date: Mon, 25 Jun 2012 16:44:00 -0400
From: "J. Bruce Fields" <bfields@fieldses.org>
To: Greg KH <gregkh@linuxfoundation.org>
Cc: "J. Bruce Fields" <bfields@redhat.com>, stable@vger.kernel.org,
        linux-nfs@vger.kernel.org,
        Stanislav Kinsbursky <skinsbursky@parallels.com>
Subject: Re: [PATCH 1/3] SUNRPC: new svc_bind() routine introduced
Message-ID: <20120625204400.GB12807@fieldses.org>
References: <1340050049-12880-1-git-send-email-bfields@redhat.com>
 <1340050049-12880-2-git-send-email-bfields@redhat.com>
 <20120620163427.GA6695@kroah.com>
 <20120620180928.GB24777@pad.fieldses.org>
 <20120620181835.GA8356@kroah.com>
 <20120625203332.GA12807@fieldses.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
In-Reply-To: <20120625203332.GA12807@fieldses.org>
Sender: linux-nfs-owner@vger.kernel.org
List-ID: <linux-nfs.vger.kernel.org>

On Mon, Jun 25, 2012 at 04:33:32PM -0400, bfields wrote:
> On Wed, Jun 20, 2012 at 11:18:35AM -0700, Greg KH wrote:
> > On Wed, Jun 20, 2012 at 02:09:29PM -0400, J. Bruce Fields wrote:
> > > On Wed, Jun 20, 2012 at 09:34:27AM -0700, Greg KH wrote:
> > > > On Mon, Jun 18, 2012 at 04:07:27PM -0400, J. Bruce Fields wrote:
> > > > > From: Stanislav Kinsbursky <skinsbursky@parallels.com>
> > > > > 
> > > > > commit cd96891d48a945ca2011fbeceda73813d6286195 upstream.
> > > > 
> > > > Um, no, that's not what this commit id is at all.
> > > 
> > > Whoops, sorry!  Should have been 9793f7c88.  Upstream commits on #2 and
> > > #3 are correct.
> > > 
> > > > Where did you get that from?  I kind of don't trust this series now,
> > > > care to fix these up with the correct git commit ids and resend them?
> > > 
> > > Sure, I'll resend once I get the "4/3" patch sorted out.
> > > 
> > > (That got a form rejection--I assume it was the missing upstream commit
> > > ID?  That's because there isn't one--3.4 requires a different fix.  But
> > > I can point to where the bug got fixed upstream even if it isn't the
> > > same.)
> > 
> > You need to be really specific when sending a patch that is not
> > upstream, detailing why it isn't, and what different fix went in
> > upstream, if possible.
> 
> OK, looking at it closer I notice we can actually just take a different
> fix directly from upstream, and then there's no problem.  But it took me
> a few days to get around to setting up a test-case to make sure that
> still did the job....

(Oh, and my reproducer in case anyone wonders: build the following, run

	./namespace-exec /bin/bash

and

	mount -tproc proc /proc
	ps ax

just to verify you're in a new pid/mount namespace, then

	mount -tnfs yourserver:/somepath /mnt

then ^D out of the shell, and it should crash.)

--b.


#define _GNU_SOURCE
#include <unistd.h>
#include <signal.h>
#include <sched.h>
#include <sys/syscall.h>
#include <err.h>

int main(int argc, char *argv[])
{
	int pid;
	int flags = SIGCHLD;

	if (argc < 2)
		errx(1, "usage: %s command arg1 arg2 ...", argv[0]);
	flags |= CLONE_NEWNS|CLONE_NEWPID;
	pid = syscall(__NR_clone, flags, NULL);
	if (pid < 0)
		err(1, "clone failed");
	if (pid == 0) {
		/* child */
		int ret;

		ret = execv(argv[1], argv+1);
		if (ret)
			err(1,"failed to exec %s", argv[1]);
	}
	wait();
}