Date: Mon, 21 Sep 2009 15:38:57 +0100
From: Al Viro <viro@ZenIV.linux.org.uk>
To: Mike Frysinger <vapier.adi@gmail.com>
Cc: Miklos Szeredi <miklos@szeredi.hu>, akpm@linux-foundation.org,
       linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org,
       Valdis.Kletnieks@vt.edu, agruen@suse.de, hch@lst.de,
       hugh.dickins@tiscali.co.uk, matthew@wil.cx
Subject: Re: [PATCH 2/2] vfs: fix d_path() for unreachable paths
Message-ID: <20090921143857.GE14381@ZenIV.linux.org.uk>
References: <E1MpiJo-0005fv-2y@pomaz-ex.szeredi.hu> <E1MpiMo-0005h9-DG@pomaz-ex.szeredi.hu> <20090921140220.GD14381@ZenIV.linux.org.uk> <8bd0f97a0909210710h5bb75bcdwb666b51a9155a70a@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <8bd0f97a0909210710h5bb75bcdwb666b51a9155a70a@mail.gmail.com>
User-Agent: Mutt/1.5.18 (2008-05-17)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1519
Lines: 39

On Mon, Sep 21, 2009 at 10:10:17AM -0400, Mike Frysinger wrote:

> it works without having to copy & paste the same exact structures over
> and over.  a suggestion as how to do it cleanly without bloating the
> code is certainly welcome.  it doesnt really matter that it's on the
> stack as the usage is small and d_path() is given the size of the
> buffer, so it isnt going to overflow.

Umm...  Surely, you can put a CPU number + one bit into PDE->data?  As in
	proc_create_data("icplb", ....., (void *)(cpu * 2))
	proc_create_data("dcplb", ....., (void *)(cpu * 2 + 1))
and
	struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
	unsigned long n = (unsigned long) pde->data;
	...
	cpu = n / 2;
	is_D = n & 1;

> > ??For another... locking in that loop
> > over processes and VMAs of each looks very suspicios, while we are at it.
> 
> we've reviewed it several times and exercised it in multiple ways.  so
> unless you have a real idea of something being wrong, the code has
> been vetted heavily.

write_lock_irq tasklist_lock
loop over processes
	get_task_mm [now VM is pinned down, but not locked]
	walk the VMA list without any locking

Something on another CPU (you have SMP systems, judging by the previously
discussed code, right?) does munmap().

Boom.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/