From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-wi0-f180.google.com (mail-wi0-f180.google.com [209.85.212.180]) by kanga.kvack.org (Postfix) with ESMTP id B31E36B0253 for ; Mon, 21 Sep 2015 11:01:39 -0400 (EDT) Received: by wicgb1 with SMTP id gb1so118810444wic.1 for ; Mon, 21 Sep 2015 08:01:39 -0700 (PDT) Received: from mail-wi0-f170.google.com (mail-wi0-f170.google.com. [209.85.212.170]) by mx.google.com with ESMTPS id ir10si31665131wjb.206.2015.09.21.08.01.37 for (version=TLS1_2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128/128); Mon, 21 Sep 2015 08:01:37 -0700 (PDT) Received: by wicfx3 with SMTP id fx3so150409234wic.1 for ; Mon, 21 Sep 2015 08:01:37 -0700 (PDT) Date: Mon, 21 Sep 2015 18:01:35 +0300 From: "Kirill A. Shutemov" Subject: Re: [linux-next] khugepaged inconsistent lock state Message-ID: <20150921150135.GB30755@node.dhcp.inet.fi> References: <20150921044600.GA863@swordfish> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20150921044600.GA863@swordfish> Sender: owner-linux-mm@kvack.org List-ID: To: Sergey Senozhatsky , Michal Hocko , Ebru Akagunduz , Hugh Dickins Cc: Andrew Morton , linux-mm@kvack.org, linux-kernel@vger.kernel.org, Sergey Senozhatsky On Mon, Sep 21, 2015 at 01:46:00PM +0900, Sergey Senozhatsky wrote: > Hi, > > 4.3.0-rc1-next-20150918 > > [18344.236625] ================================= > [18344.236628] [ INFO: inconsistent lock state ] > [18344.236633] 4.3.0-rc1-next-20150918-dbg-00014-ge5128d0-dirty #361 Not tainted > [18344.236636] --------------------------------- > [18344.236640] inconsistent {IN-RECLAIM_FS-W} -> {RECLAIM_FS-ON-W} usage. > [18344.236645] khugepaged/32 [HC0[0]:SC0[0]:HE1:SE1] takes: > [18344.236648] (&anon_vma->rwsem){++++?.}, at: [] khugepaged+0x8b0/0x1987 > [18344.236662] {IN-RECLAIM_FS-W} state was registered at: > [18344.236666] [] __lock_acquire+0x8e2/0x1183 > [18344.236673] [] lock_acquire+0x10b/0x1a6 > [18344.236678] [] down_write+0x3b/0x6a > [18344.236686] [] split_huge_page_to_list+0x5b/0x61f > [18344.236689] [] add_to_swap+0x37/0x78 > [18344.236691] [] shrink_page_list+0x4c2/0xb9a > [18344.236694] [] shrink_inactive_list+0x371/0x5d9 > [18344.236696] [] shrink_lruvec+0x410/0x5ae > [18344.236698] [] shrink_zone+0x57/0x140 > [18344.236700] [] kswapd+0x6a5/0x91b > [18344.236702] [] kthread+0x107/0x10f > [18344.236706] [] ret_from_fork+0x3f/0x70 > [18344.236708] irq event stamp: 6517947 > [18344.236709] hardirqs last enabled at (6517947): [] get_page_from_freelist+0x362/0x59e > [18344.236713] hardirqs last disabled at (6517946): [] _raw_spin_lock_irqsave+0x18/0x51 > [18344.236715] softirqs last enabled at (6507072): [] __do_softirq+0x2df/0x3f5 > [18344.236719] softirqs last disabled at (6507055): [] irq_exit+0x40/0x94 > [18344.236722] > other info that might help us debug this: > [18344.236723] Possible unsafe locking scenario: > > [18344.236724] CPU0 > [18344.236725] ---- > [18344.236726] lock(&anon_vma->rwsem); > [18344.236728] > [18344.236729] lock(&anon_vma->rwsem); > [18344.236731] > *** DEADLOCK *** > > [18344.236733] 2 locks held by khugepaged/32: > [18344.236733] #0: (&mm->mmap_sem){++++++}, at: [] khugepaged+0x5cf/0x1987 > [18344.236738] #1: (&anon_vma->rwsem){++++?.}, at: [] khugepaged+0x8b0/0x1987 > [18344.236741] > stack backtrace: > [18344.236744] CPU: 3 PID: 32 Comm: khugepaged Not tainted 4.3.0-rc1-next-20150918-dbg-00014-ge5128d0-dirty #361 > [18344.236747] 0000000000000000 ffff880132827a00 ffffffff81230867 ffffffff8237ba90 > [18344.236750] ffff880132827a38 ffffffff810ea9b9 000000000000000a ffff8801333b52e0 > [18344.236753] ffff8801333b4c00 ffffffff8107b3ce 000000000000000a ffff880132827a78 > [18344.236755] Call Trace: > [18344.236758] [] dump_stack+0x4e/0x79 > [18344.236761] [] print_usage_bug.part.24+0x259/0x268 > [18344.236763] [] ? print_shortest_lock_dependencies+0x180/0x180 > [18344.236765] [] mark_lock+0x381/0x567 > [18344.236766] [] mark_held_locks+0x5e/0x74 > [18344.236768] [] lockdep_trace_alloc+0xb0/0xb3 > [18344.236771] [] __alloc_pages_nodemask+0x99/0x856 > [18344.236772] [] ? find_get_entry+0x14b/0x17a > [18344.236774] [] ? find_get_entry+0x168/0x17a > [18344.236777] [] __read_swap_cache_async+0x7b/0x1aa > [18344.236778] [] read_swap_cache_async+0x15/0x2d > [18344.236780] [] swapin_readahead+0x11a/0x16a > [18344.236783] [] do_swap_page+0xa7/0x36b > [18344.236784] [] ? do_swap_page+0xa7/0x36b > [18344.236787] [] khugepaged+0x8f9/0x1987 > [18344.236790] [] ? wait_woken+0x88/0x88 > [18344.236792] [] ? maybe_pmd_mkwrite+0x1a/0x1a > [18344.236794] [] kthread+0x107/0x10f > [18344.236797] [] ? kthread_create_on_node+0x1ea/0x1ea > [18344.236799] [] ret_from_fork+0x3f/0x70 > [18344.236801] [] ? kthread_create_on_node+0x1ea/0x1ea Hm. If I read this correctly, we see following scenario: - khugepaged tries to swap in a page under mmap_sem and anon_vma lock; - do_swap_page() calls swapin_readahead() with GFP_HIGHUSER_MOVABLE; - __read_swap_cache_async() tries to allocate the page for swap in; - lockdep_trace_alloc() in __alloc_pages_nodemask() notices that with given gfp_mask we could end up in direct relaim. - Lockdep already knows that reclaim sometimes (e.g. in case of split_huge_page()) wants to take anon_vma lock on its own. Therefore deadlock is possible. I see two ways to fix this: - take anon_vma lock *after* __collapse_huge_page_swapin() in collapse_huge_page(): I don't really see why we need the lock during swapin; - respect FAULT_FLAG_RETRY_NOWAIT in do_swap_page(): add GFP_NOWAIT to gfp_mask for swapin_readahead() in this case. I guess it could be beneficial to do both. Any comments? -- Kirill A. Shutemov -- 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