From mboxrd@z Thu Jan  1 00:00:00 1970
From: Eric Paris <eparis@parisplace.org>
Subject: Re: [patch 09/33] fs: dcache scale dentry refcount
Date: Sun, 6 Sep 2009 14:01:05 -0400
Message-ID: <7e0fb38c0909061101t6b4f337cvf691814c28bcf50e@mail.gmail.com>
References: <20090904065142.114706411@nick.local0.net>
	 <20090904065535.281876981@nick.local0.net>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: QUOTED-PRINTABLE
Cc: linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org
To: npiggin@suse.de
Return-path: <linux-fsdevel-owner@vger.kernel.org>
Received: from mail-vw0-f195.google.com ([209.85.212.195]:50972 "EHLO
	mail-vw0-f195.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
	with ESMTP id S1758068AbZIFSBD convert rfc822-to-8bit (ORCPT
	<rfc822;linux-fsdevel@vger.kernel.org>);
	Sun, 6 Sep 2009 14:01:03 -0400
In-Reply-To: <20090904065535.281876981@nick.local0.net>
Sender: linux-fsdevel-owner@vger.kernel.org
List-ID: <linux-fsdevel.vger.kernel.org>

On Fri, Sep 4, 2009 at 2:51 AM, <npiggin@suse.de> wrote:
> Make d_count non-atomic and protect it with d_lock. This allows us to
> ensure a 0 refcount dentry remains 0 without dcache_lock. It is also
> fairly natural when we start protecting many other dentry members wit=
h
> d_lock.

> +struct dentry *dget_parent(struct dentry *dentry)
> +{
> + =A0 =A0 =A0 struct dentry *ret;
> +
> +repeat:
> + =A0 =A0 =A0 spin_lock(&dentry->d_lock);
> + =A0 =A0 =A0 ret =3D dentry->d_parent;
> + =A0 =A0 =A0 if (!ret)
> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 goto out;
> + =A0 =A0 =A0 if (!spin_trylock(&ret->d_lock)) {
> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 spin_unlock(&dentry->d_lock);
> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 goto repeat;
> + =A0 =A0 =A0 }
> + =A0 =A0 =A0 BUG_ON(!ret->d_count);
> + =A0 =A0 =A0 ret->d_count++;
> + =A0 =A0 =A0 spin_unlock(&ret->d_lock);
> +out:
> + =A0 =A0 =A0 spin_unlock(&dentry->d_lock);
> + =A0 =A0 =A0 return ret;
> +}
> +EXPORT_SYMBOL(dget_parent);


> Index: linux-2.6/fs/notify/inotify/inotify.c
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> --- linux-2.6.orig/fs/notify/inotify/inotify.c
> +++ linux-2.6/fs/notify/inotify/inotify.c
> @@ -339,18 +339,26 @@ void inotify_dentry_parent_queue_event(s
> =A0 =A0 =A0 =A0if (!(dentry->d_flags & DCACHE_INOTIFY_PARENT_WATCHED)=
)
> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0return;
>
> +again:
> =A0 =A0 =A0 =A0spin_lock(&dentry->d_lock);
> =A0 =A0 =A0 =A0parent =3D dentry->d_parent;
> + =A0 =A0 =A0 if (!spin_trylock(&parent->d_lock)) {
> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 spin_unlock(&dentry->d_lock);
> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 goto again;
> + =A0 =A0 =A0 }
> =A0 =A0 =A0 =A0inode =3D parent->d_inode;
>
> =A0 =A0 =A0 =A0if (inotify_inode_watched(inode)) {
> - =A0 =A0 =A0 =A0 =A0 =A0 =A0 dget(parent);
> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 dget_dlock(parent);
> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0spin_unlock(&dentry->d_lock);
> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 spin_unlock(&parent->d_lock);
> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0inotify_inode_queue_event(inode, mask,=
 cookie, name,
> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0=
 =A0 =A0 =A0dentry->d_inode);
> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0dput(parent);
> - =A0 =A0 =A0 } else
> + =A0 =A0 =A0 } else {
> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0spin_unlock(&dentry->d_lock);
> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 spin_unlock(&parent->d_lock);
> + =A0 =A0 =A0 }

I don't think I understand why in both of these cases you don't need
to check for dentry->d_parent =3D=3D dentry

> Index: linux-2.6/fs/notify/fsnotify.c
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> --- linux-2.6.orig/fs/notify/fsnotify.c
> +++ linux-2.6/fs/notify/fsnotify.c
> @@ -87,13 +87,18 @@ void __fsnotify_parent(struct dentry *de
> =A0 =A0 =A0 =A0if (!(dentry->d_flags & DCACHE_FSNOTIFY_PARENT_WATCHED=
))
> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0return;
>
> +again:
> =A0 =A0 =A0 =A0spin_lock(&dentry->d_lock);
> =A0 =A0 =A0 =A0parent =3D dentry->d_parent;
> + =A0 =A0 =A0 if (parent !=3D dentry && !spin_trylock(&parent->d_lock=
)) {
> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 spin_unlock(&dentry->d_lock);
> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 goto again;
> + =A0 =A0 =A0 }
> =A0 =A0 =A0 =A0p_inode =3D parent->d_inode;
>
> =A0 =A0 =A0 =A0if (fsnotify_inode_watches_children(p_inode)) {
> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0if (p_inode->i_fsnotify_mask & mask) {
> - =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 dget(parent);
> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 dget_dlock(parent);
> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0send =3D true;
> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0}
> =A0 =A0 =A0 =A0} else {

And yet in this case we do check for dentry->d_parent =3D=3D dentry.  (=
my
unknowing self thinks we'd want to check in all places)

-Eric
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel=
" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html