From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail172.messagelabs.com (mail172.messagelabs.com [216.82.254.3]) by kanga.kvack.org (Postfix) with ESMTP id 579026B004D for ; Wed, 29 Apr 2009 22:13:27 -0400 (EDT) Date: Wed, 29 Apr 2009 19:10:44 -0700 From: Andrew Morton Subject: Re: [PATCH] use GFP_NOFS in kernel_event() Message-Id: <20090429191044.b6fceae2.akpm@linux-foundation.org> In-Reply-To: <20090430020004.GA1898@localhost> References: <20090430020004.GA1898@localhost> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-linux-mm@kvack.org To: Wu Fengguang Cc: LKML , linux-mm@kvack.org, Matt Mackall , Christoph Lameter , Ingo Molnar , Al Viro , Eric Paris List-ID: On Thu, 30 Apr 2009 10:00:04 +0800 Wu Fengguang wrote: > Fix a possible deadlock on inotify_mutex, reported by lockdep. > > inotify_inode_queue_event() => take inotify_mutex => kernel_event() => > kmalloc() => SLOB => alloc_pages_node() => page reclaim => slab reclaim => > dcache reclaim => inotify_inode_is_dead => take inotify_mutex => deadlock > > The actual deadlock may not happen because the inode was grabbed at > inotify_add_watch(). But the GFP_KERNEL here is unsound and not > consistent with the other two GFP_NOFS inside the same function. > > [ 2668.325318] > [ 2668.325322] ================================= > [ 2668.327448] [ INFO: inconsistent lock state ] > [ 2668.327448] 2.6.30-rc2-next-20090417 #203 > [ 2668.327448] --------------------------------- > [ 2668.327448] inconsistent {RECLAIM_FS-ON-W} -> {IN-RECLAIM_FS-W} usage. > [ 2668.327448] kswapd0/380 [HC0[0]:SC0[0]:HE1:SE1] takes: > [ 2668.327448] (&inode->inotify_mutex){+.+.?.}, at: [] inotify_inode_is_dead+0x35/0xb0 > [ 2668.327448] {RECLAIM_FS-ON-W} state was registered at: > [ 2668.327448] [] mark_held_locks+0x68/0x90 > [ 2668.327448] [] lockdep_trace_alloc+0xf5/0x100 > [ 2668.327448] [] __kmalloc_node+0x31/0x1e0 > [ 2668.327448] [] kernel_event+0xe2/0x190 > [ 2668.327448] [] inotify_dev_queue_event+0x126/0x230 > [ 2668.327448] [] inotify_inode_queue_event+0xc6/0x110 > [ 2668.327448] [] vfs_create+0xcd/0x140 > [ 2668.327448] [] do_filp_open+0x88d/0xa20 > [ 2668.327448] [] do_sys_open+0x98/0x140 > [ 2668.327448] [] sys_open+0x20/0x30 > [ 2668.327448] [] system_call_fastpath+0x16/0x1b > [ 2668.327448] [] 0xffffffffffffffff > [ 2668.327448] irq event stamp: 690455 > [ 2668.327448] hardirqs last enabled at (690455): [] _spin_unlock_irqrestore+0x44/0x80 > [ 2668.327448] hardirqs last disabled at (690454): [] _spin_lock_irqsave+0x32/0xa0 > [ 2668.327448] softirqs last enabled at (690178): [] __do_softirq+0x202/0x220 > [ 2668.327448] softirqs last disabled at (690157): [] call_softirq+0x1c/0x50 > [ 2668.327448] > [ 2668.327448] other info that might help us debug this: > [ 2668.327448] 2 locks held by kswapd0/380: > [ 2668.327448] #0: (shrinker_rwsem){++++..}, at: [] shrink_slab+0x37/0x180 > [ 2668.327448] #1: (&type->s_umount_key#17){++++..}, at: [] shrink_dcache_memory+0x11f/0x1e0 > [ 2668.327448] > [ 2668.327448] stack backtrace: > [ 2668.327448] Pid: 380, comm: kswapd0 Not tainted 2.6.30-rc2-next-20090417 #203 > [ 2668.327448] Call Trace: > [ 2668.327448] [] print_usage_bug+0x19f/0x200 > [ 2668.327448] [] ? save_stack_trace+0x2f/0x50 > [ 2668.327448] [] mark_lock+0x4bb/0x6d0 > [ 2668.327448] [] ? check_usage_forwards+0x0/0xc0 > [ 2668.327448] [] __lock_acquire+0xc62/0x1ae0 > [ 2668.327448] [] ? slob_free+0x10c/0x370 > [ 2668.327448] [] lock_acquire+0xe1/0x120 > [ 2668.327448] [] ? inotify_inode_is_dead+0x35/0xb0 > [ 2668.327448] [] mutex_lock_nested+0x63/0x420 > [ 2668.327448] [] ? inotify_inode_is_dead+0x35/0xb0 > [ 2668.327448] [] ? inotify_inode_is_dead+0x35/0xb0 > [ 2668.327448] [] ? sched_clock+0x9/0x10 > [ 2668.327448] [] ? lock_release_holdtime+0x35/0x1c0 > [ 2668.327448] [] inotify_inode_is_dead+0x35/0xb0 > [ 2668.327448] [] dentry_iput+0xbc/0xe0 > [ 2668.327448] [] d_kill+0x33/0x60 > [ 2668.327448] [] __shrink_dcache_sb+0x2d3/0x350 > [ 2668.327448] [] shrink_dcache_memory+0x15a/0x1e0 > [ 2668.327448] [] shrink_slab+0x125/0x180 > [ 2668.327448] [] kswapd+0x560/0x7a0 > [ 2668.327448] [] ? isolate_pages_global+0x0/0x2c0 > [ 2668.327448] [] ? autoremove_wake_function+0x0/0x40 > [ 2668.327448] [] ? trace_hardirqs_on+0xd/0x10 > [ 2668.327448] [] ? kswapd+0x0/0x7a0 > [ 2668.327448] [] kthread+0x5b/0xa0 > [ 2668.327448] [] child_rip+0xa/0x20 > [ 2668.327448] [] ? restore_args+0x0/0x30 > [ 2668.327448] [] ? kthread+0x0/0xa0 > [ 2668.327448] [] ? child_rip+0x0/0x20 > > cc: Al Viro > cc: Matt Mackall > cc: Christoph Lameter > Signed-off-by: Wu Fengguang > --- > fs/notify/inotify/inotify_user.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > --- mm.orig/fs/notify/inotify/inotify_user.c > +++ mm/fs/notify/inotify/inotify_user.c > @@ -220,7 +220,7 @@ static struct inotify_kernel_event * ker > rem = 0; > } > > - kevent->name = kmalloc(len + rem, GFP_KERNEL); > + kevent->name = kmalloc(len + rem, GFP_NOFS); > if (unlikely(!kevent->name)) { > kmem_cache_free(event_cachep, kevent); > return NULL; Somebody was going to fix this for us via lockdep annotation. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: email@kvack.org