| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <560D448F.9050507@oracle.com>
Date: Thu, 01 Oct 2015 10:34:55 -0400
From: Sasha Levin <sasha.levin@...cle.com>
To: Hugh Dickins <hughd@...gle.com>
CC: LKML <linux-kernel@...r.kernel.org>,
"linux-mm@...ck.org" <linux-mm@...ck.org>,
Andrew Morton <akpm@...ux-foundation.org>
Subject: mm: ksm: deadlock in oom killing process while breaking ksm pages
Hi Hugh,
I've hit this (actual) lockup during testing. It seems that we were trying to allocate
a new page to break KSM on an existing page, ended up in the oom killer who killed our
process, and locked up in __ksm_exit() trying to get a write lock while already holding
a read lock.
A very similar scenario is presented in the patch that introduced this behaviour
(9ba6929480 ("ksm: fix oom deadlock")):
There's a now-obvious deadlock in KSM's out-of-memory handling:
imagine ksmd or KSM_RUN_UNMERGE handling, holding ksm_thread_mutex,
trying to allocate a page to break KSM in an mm which becomes the
OOM victim (quite likely in the unmerge case): it's killed and goes
to exit, and hangs there waiting to acquire ksm_thread_mutex.
So I'm guessing that the solution is incomplete for the slow path.
[3201844.610523] =============================================
[3201844.610988] [ INFO: possible recursive locking detected ]
[3201844.611405] 4.3.0-rc3-next-20150930-sasha-00077-g3434920 #4 Not tainted
[3201844.611907] ---------------------------------------------
[3201844.612373] ksm02/28830 is trying to acquire lock:
[3201844.612749] (&mm->mmap_sem){++++++}, at: __ksm_exit (mm/ksm.c:1821)
[3201844.613472] RWsem: count: 1 owner: None
[3201844.613782]
[3201844.613782] but task is already holding lock:
[3201844.614248] (&mm->mmap_sem){++++++}, at: run_store (mm/ksm.c:769 mm/ksm.c:2124)
[3201844.614904] RWsem: count: 1 owner: None
[3201844.615212]
[3201844.615212] other info that might help us debug this:
[3201844.615727] Possible unsafe locking scenario:
[3201844.615727]
[3201844.616240] CPU0
[3201844.616446] ----
[3201844.616650] lock(&mm->mmap_sem);
[3201844.616952] lock(&mm->mmap_sem);
[3201844.617252]
[3201844.617252] *** DEADLOCK ***
[3201844.617252]
[3201844.617733] May be due to missing lock nesting notation
[3201844.617733]
[3201844.618265] 6 locks held by ksm02/28830:
[3201844.618576] #0: (sb_writers#5){.+.+.+}, at: __sb_start_write (fs/super.c:1176)
[3201844.619327] RWsem: count: 0 owner: None
[3201844.619633] #1: (&of->mutex){+.+.+.}, at: kernfs_fop_write (fs/kernfs/file.c:298)
[3201844.624648] Mutex: counter: 0 owner: ksm02
[3201844.624978] #2: (s_active#448){.+.+.+}, at: kernfs_fop_write (fs/kernfs/file.c:298)
[3201844.625733] #3: (ksm_thread_mutex){+.+.+.}, at: run_store (mm/ksm.c:2120)
[3201844.626448] Mutex: counter: -1 owner: ksm02
[3201844.626786] #4: (&mm->mmap_sem){++++++}, at: run_store (mm/ksm.c:769 mm/ksm.c:2124)
[3201844.627486] RWsem: count: 1 owner: None
[3201844.627792] #5: (oom_lock){+.+...}, at: __alloc_pages_nodemask (mm/page_alloc.c:2779 mm/page_alloc.c:3213 mm/page_alloc.c:3300)
[3201844.628594] Mutex: counter: 0 owner: ksm02
[3201844.628919]
[3201844.628919] stack backtrace:
[3201844.629276] CPU: 0 PID: 28830 Comm: ksm02 Not tainted 4.3.0-rc3-next-20150930-sasha-00077-g3434920 #4
[3201844.629970] ffffffffaf41d680 00000000b8d5e1f1 ffff88065e42eec0 ffffffffa1d454c8
[3201844.630663] ffffffffaf41d680 ffff88065e42f080 ffffffffa04269ee ffff88065e42f088
[3201844.631292] ffffffffa0427746 ffff882c88b24008 ffff8806845b8e10 ffffffffafb842c0
[3201844.631952] Call Trace:
[3201844.632204] dump_stack (lib/dump_stack.c:52)
[3201844.636449] __lock_acquire (kernel/locking/lockdep.c:1776 kernel/locking/lockdep.c:1820 kernel/locking/lockdep.c:2152 kernel/locking/lockdep.c:3239)
[3201844.639909] lock_acquire (kernel/locking/lockdep.c:3620)
[3201844.640997] down_write (./arch/x86/include/asm/rwsem.h:130 kernel/locking/rwsem.c:51)
[3201844.642011] __ksm_exit (mm/ksm.c:1821)
[3201844.642501] mmput (./arch/x86/include/asm/bitops.h:311 include/linux/khugepaged.h:35 kernel/fork.c:701)
[3201844.642920] oom_kill_process (mm/oom_kill.c:604)
[3201844.644528] out_of_memory (mm/oom_kill.c:700)
[3201844.646626] __alloc_pages_nodemask (mm/page_alloc.c:2822 mm/page_alloc.c:3213 mm/page_alloc.c:3300)
[3201844.649972] alloc_pages_vma (mm/mempolicy.c:2044)
[3201844.650462] ? wp_page_copy.isra.36 (mm/memory.c:2074)
[3201844.651000] wp_page_copy.isra.36 (mm/memory.c:2074)
[3201844.652544] do_wp_page (mm/memory.c:2349)
[3201844.654048] handle_mm_fault (mm/memory.c:3310 mm/memory.c:3404 mm/memory.c:3433)
[3201844.657519] break_ksm (mm/ksm.c:374)
[3201844.659348] unmerge_ksm_pages (mm/ksm.c:673)
[3201844.659831] run_store (mm/ksm.c:776 mm/ksm.c:2124)
[3201844.661837] kobj_attr_store (lib/kobject.c:792)
[3201844.662743] sysfs_kf_write (fs/sysfs/file.c:131)
[3201844.663656] kernfs_fop_write (fs/kernfs/file.c:312)
[3201844.664154] __vfs_write (fs/read_write.c:489)
[3201844.666502] vfs_write (fs/read_write.c:539)
[3201844.666935] SyS_write (fs/read_write.c:586 fs/read_write.c:577)
[3201844.668965] tracesys_phase2 (arch/x86/entry/entry_64.S:270)
Thanks,
Sasha
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists