From mboxrd@z Thu Jan  1 00:00:00 1970
From: ebiederm@xmission.com (Eric W. Biederman)
Subject: Re: [PATCH] proc: make proc_fd_permission() thread-friendly
Date: Mon, 26 Aug 2013 12:10:16 -0700
Message-ID: <8738pwp8tj.fsf@xmission.com>
References: <20130824212432.GA9299@1wt.eu>
	<20130825052317.GZ27005@ZenIV.linux.org.uk>
	<20130825065039.GB9299@1wt.eu>
	<CA+55aFwFeCRcwn4qo570Y15mKrN3jEdN757PcZhrXs7QYi8MYA@mail.gmail.com>
	<20130825194844.GA16717@redhat.com>
	<CA+55aFzbY3zYD1i5x02Z-6K3nHKzJj6227ZBYRLqij9y0eXRqQ@mail.gmail.com>
	<20130826153301.GA15890@redhat.com>
	<20130826163704.GA21763@redhat.com>
	<20130826175441.GA28856@redhat.com> <87ioyspaju.fsf@xmission.com>
	<20130826184631.GA30917@redhat.com>
Mime-Version: 1.0
Content-Type: text/plain
Cc: Andrew Morton <akpm@linux-foundation.org>,
	Willy Tarreau <w@1wt.eu>, Al Viro <viro@zeniv.linux.org.uk>,
	Andy Lutomirski <luto@amacapital.net>,
	Ingo Molnar <mingo@kernel.org>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	Linux FS Devel <linux-fsdevel@vger.kernel.org>,
	Brad Spengler <spender@grsecurity.net>,
	Linus Torvalds <torvalds@linux-foundation.org>
To: Oleg Nesterov <oleg@redhat.com>
Return-path: <linux-kernel-owner@vger.kernel.org>
In-Reply-To: <20130826184631.GA30917@redhat.com> (Oleg Nesterov's message of
	"Mon, 26 Aug 2013 20:46:31 +0200")
Sender: linux-kernel-owner@vger.kernel.org
List-Id: linux-fsdevel.vger.kernel.org

Oleg Nesterov <oleg@redhat.com> writes:

> On 08/26, Eric W. Biederman wrote:
>>
>> Oleg Nesterov <oleg@redhat.com> writes:
>>
>> > proc_fd_permission() says "process can still access /proc/self/fd
>> > after it has executed a setuid()", but the "task_pid() = proc_pid()
>> > check only helps if the task is group leader, /proc/self points to
>> > /proc/leader-pid.
>> >
>> > Change this check to use task_tgid() so that the whole process can
>> > access /proc/self/fd.
>>
>> There is at least a semantic goofiness here.
>>
>> There is /proc/<tgid>/fd and /proc/<tgid>/task/<pid>/fd, and the same
>> permission check is used by both.
>
> Yes, and we have /proc/<tid>/ which includes fd as well.
>
>> We might just want to have a /proc/thread symlink as well so people
>> don't have this issue.
>
> Yes! I agree.
>
> In particular, from the changelog:
>
> 	Note: CLONE_THREAD doesn't require CLONE_FILES so task->files can
> 	differ,
>
> so /proc/self/fd doesn't necessarily mean current->files, this can confuse
> the application.
>
> And I also assume that you agree with this change ;)

I don't disagree.  Comparing tgid to pids is goofy and my brain is
elsewhere so I have no thought through the implications.

Actually thinking I think the check should really be.  In which case we
are comparing what we really care about.

int proc_fd_permission(struct inode *inode, int mask)
{
	int rv = generic_permission(inode, mask);
	if (rv == 0)
		return 0;

        rcu_read_lock();
        struct task *task = pid_task(proc_pid(inode));
        if (task && (current->files == task->files))
		rv = 0;
        rcu_read_unlock();

	return rv;
}

Eric