From: Mel Gorman Subject: Re: [PATCH] mm: migration: Use rcu_dereference_protected when dereferencing the radix tree slot during file page migration Date: Tue, 21 Dec 2010 10:49:08 +0000 Message-ID: <20101221104907.GT13914@csn.ul.ie> References: <20101220152335.GR13914@csn.ul.ie> <20101220170146.GS13914@csn.ul.ie> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-15 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: Andrew Morton , gerald.schaefer@de.ibm.com, KAMEZAWA Hiroyuki , Milton Miller , linux-kernel@vger.kernel.org, linux-mm@kvack.org, linux-ext4@vger.kernel.org, Ted Ts'o , Arun Bhanu , Heiko Carstens , Martin Schwidefsky To: Minchan Kim Return-path: Received: from gir.skynet.ie ([193.1.99.77]:42294 "EHLO gir.skynet.ie" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751058Ab0LUKt2 (ORCPT ); Tue, 21 Dec 2010 05:49:28 -0500 Content-Disposition: inline In-Reply-To: Sender: linux-ext4-owner@vger.kernel.org List-ID: On Tue, Dec 21, 2010 at 08:48:50AM +0900, Minchan Kim wrote: > On Tue, Dec 21, 2010 at 2:01 AM, Mel Gorman wrote: > > On Mon, Dec 20, 2010 at 03:23:36PM +0000, Mel Gorman wrote: > >> migrate_pages() -> unmap_and_move() only calls rcu_read_lock() for= anonymous > >> pages, as introduced by git commit 989f89c57e6361e7d16fbd9572b5da7= d313b073d. > >> The point of the RCU protection there is part of getting a stable = reference > >> to anon_vma and is only held for anon pages as file pages are lock= ed > >> which is sufficient protection against freeing. > >> > >> However, while a file page's mapping is being migrated, the radix > >> tree is double checked to ensure it is the expected page. This use= s > >> radix_tree_deref_slot() -> rcu_dereference() without the RCU lock = held > >> triggering the following warning under CONFIG_PROVE_RCU. > >> > >> [ =A0173.674290] =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 > >> [ =A0173.676016] [ INFO: suspicious rcu_dereference_check() usage.= ] > >> [ =A0173.676016] -------------------------------------------------= -- > >> [ =A0173.676016] include/linux/radix-tree.h:145 invoked rcu_derefe= rence_check() without protection! > >> [ =A0173.676016] > >> [ =A0173.676016] other info that might help us debug this: > >> [ =A0173.676016] > >> [ =A0173.676016] > >> [ =A0173.676016] rcu_scheduler_active =3D 1, debug_locks =3D 0 > >> [ =A0173.676016] 1 lock held by hugeadm/2899: > >> [ =A0173.676016] =A0#0: =A0(&(&inode->i_data.tree_lock)->rlock){..= -.-.}, at: [] migrate_page_move_mapping+0x40/0x1ab > >> [ =A0173.676016] > >> [ =A0173.676016] stack backtrace: > >> [ =A0173.676016] Pid: 2899, comm: hugeadm Not tainted 2.6.37-rc5-a= utobuild > >> [ =A0173.676016] Call Trace: > >> [ =A0173.676016] =A0[] ? printk+0x14/0x1b > >> [ =A0173.676016] =A0[] lockdep_rcu_dereference+0x7d/0x86 > >> [ =A0173.676016] =A0[] migrate_page_move_mapping+0xca/0x= 1ab > >> [ =A0173.676016] =A0[] migrate_page+0x23/0x39 > >> [ =A0173.676016] =A0[] buffer_migrate_page+0x22/0x107 > >> [ =A0173.676016] =A0[] ? buffer_migrate_page+0x0/0x107 > >> [ =A0173.676016] =A0[] move_to_new_page+0x9a/0x1ae > >> [ =A0173.676016] =A0[] migrate_pages+0x1e7/0x2fa > >> > >> This patch introduces radix_tree_deref_slot_protected() which call= s > >> rcu_dereference_protected(). Users of it must pass in the mapping-= >tree_lock > >> that is protecting this dereference. Holding the tree lock protect= s against > >> parallel updaters of the radix tree meaning that rcu_dereference_p= rotected > >> is allowable. > >> > >> Signed-off-by: Mel Gorman > >> --- > >> =A0include/linux/radix-tree.h | =A0 17 +++++++++++++++++ > >> =A0mm/migrate.c =A0 =A0 =A0 =A0 =A0 =A0 =A0 | =A0 =A04 ++-- > >> =A02 files changed, 19 insertions(+), 2 deletions(-) > >> > >> diff --git a/include/linux/radix-tree.h b/include/linux/radix-tree= =2Eh > >> index ab2baa5..a1f1672 100644 > >> --- a/include/linux/radix-tree.h > >> +++ b/include/linux/radix-tree.h > >> @@ -146,6 +146,23 @@ static inline void *radix_tree_deref_slot(voi= d **pslot) > >> =A0} > >> > >> =A0/** > >> + * radix_tree_deref_slot_protected =A0 - dereference a slot witho= ut RCU lock but with tree lock held > >> + * @pslot: =A0 pointer to slot, returned by radix_tree_lookup_slo= t > >> + * Returns: =A0item that was stored in that slot with any direct = pointer flag > >> + * =A0 =A0 =A0 =A0 =A0 removed. > >> + * > >> + * Similar to radix_tree_deref_slot but only used during migratio= n when a pages > >> + * mapping is being moved. The caller does not hold the RCU read = lock but it > >> + * must hold the tree lock to prevent parallel updates. > >> + */ > >> +static inline void *radix_tree_deref_slot_protected(void **pslot, > >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 = =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 spinlock_t *treelock) > >> +{ > >> + =A0 =A0 BUG_ON(rcu_read_lock_held()); >=20 > Hmm.. Why did you add the check? > If rcu_read_lock were already held, we wouldn't need this new API. >=20 Because our earlier discussions assumed that RCU read lock was not held in this path. The check was added to ensure that assumption was correct= , it wasn't. > > > > This was a bad idea. After some extended testing, it was obvious th= at > > this function can be called for swapcache pages with the RCU lock h= eld. > > Paul, is it still permissible to use rcu_dereference_protected() or= must >=20 > I guess has no problem. >=20 > > the RCU read lock not be held? > > >=20 >=20 >=20 > --=20 > Kind regards, > Minchan Kim >=20 --=20 Mel Gorman Part-time Phd Student Linux Technology Center University of Limerick IBM Dublin Software Lab -- To unsubscribe from this list: send the line "unsubscribe linux-ext4" i= n the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html