Re: printk() + memory offline deadlock (WAS Re: page_alloc.shuffle=1 + CONFIG_PROVE_LOCKING=y = arm64 hang)

From: Sergey Senozhatsky
Date: Wed Sep 18 2019 - 11:51:06 EST


On (09/18/19 10:39), Qian Cai wrote:
> > Perhaps for a quick fix (and a comment that says this needs to be fixed
> > properly). I think the changes to printk() that was discussed at
> > Plumbers may also solve this properly.
>
> I assume that the new printk() stuff will also fix this deadlock between
> printk() and memory offline.

Mother chicken...

Do you actually see a deadlock? I'd rather expect a lockdep splat, but
anyway...

> [  317.337595] WARNING: possible circular locking dependency detected
> [  317.337596] 5.3.0-next-20190917+ #9 Not tainted
> [  317.337597] ------------------------------------------------------
> [  317.337597] test.sh/8738 is trying to acquire lock:
> [  317.337598] ffffffffb33a4978 ((console_sem).lock){-.-.}, at:> down_trylock+0x16/0x50
>
> [  317.337602] but task is already holding lock:
> [  317.337602] ffff88883fff4318 (&(&zone->lock)->rlock){-.-.}, at:> start_isolate_page_range+0x1f7/0x570
>
> [  317.337606] which lock already depends on the new lock.
>
> [  317.337608] the existing dependency chain (in reverse order) is:
>
> [  317.337609] -> #3 (&(&zone->lock)->rlock){-.-.}:
> [  317.337612]        __lock_acquire+0x5b3/0xb40
> [  317.337613]        lock_acquire+0x126/0x280
> [  317.337613]        _raw_spin_lock+0x2f/0x40
> [  317.337614]        rmqueue_bulk.constprop.21+0xb6/0x1160
> [  317.337615]        get_page_from_freelist+0x898/0x22c0
> [  317.337616]        __alloc_pages_nodemask+0x2f3/0x1cd0
> [  317.337617]        alloc_page_interleave+0x18/0x130
> [  317.337618]        alloc_pages_current+0xf6/0x110
> [  317.337619]        allocate_slab+0x4c6/0x19c0
> [  317.337620]        new_slab+0x46/0x70
> [  317.337621]        ___slab_alloc+0x58b/0x960
> [  317.337621]        __slab_alloc+0x43/0x70
> [  317.337622]        kmem_cache_alloc+0x354/0x460
> [  317.337623]        fill_pool+0x272/0x4b0
> [  317.337624]        __debug_object_init+0x86/0x790
> [  317.337624]        debug_object_init+0x16/0x20
> [  317.337625]        hrtimer_init+0x27/0x1e0
> [  317.337626]        init_dl_task_timer+0x20/0x40
> [  317.337627]        __sched_fork+0x10b/0x1f0
> [  317.337627]        init_idle+0xac/0x520
> [  317.337628]        idle_thread_get+0x7c/0xc0
> [  317.337629]        bringup_cpu+0x1a/0x1e0
> [  317.337630]        cpuhp_invoke_callback+0x197/0x1120
> [  317.337630]        _cpu_up+0x171/0x280
> [  317.337631]        do_cpu_up+0xb1/0x120
> [  317.337632]        cpu_up+0x13/0x20
> [  317.337632]        smp_init+0xa4/0x12d
> [  317.337633]        kernel_init_freeable+0x37e/0x76e
> [  317.337634]        kernel_init+0x11/0x12f
> [  317.337635]        ret_from_fork+0x3a/0x50

So you have debug objects enabled. Right? This thing does not behave
when it comes to printing. debug_objects are slightly problematic.

This thing does

rq->lock --> zone->lock

It takes rq->lock and then calls into __sched_fork()->hrtimer_init()->debug_objects()->MM

This doesn't look very right - a dive into MM under rq->lock.

Peter, Thomas am I wrong?

> [  317.337635] -> #2 (&rq->lock){-.-.}:
> [  317.337638]        __lock_acquire+0x5b3/0xb40
> [  317.337639]        lock_acquire+0x126/0x280
> [  317.337639]        _raw_spin_lock+0x2f/0x40
> [  317.337640]        task_fork_fair+0x43/0x200
> [  317.337641]        sched_fork+0x29b/0x420
> [  317.337642]        copy_process+0xf3c/0x2fd0
> [  317.337642]        _do_fork+0xef/0x950
> [  317.337643]        kernel_thread+0xa8/0xe0
> [  317.337644]        rest_init+0x28/0x311
> [  317.337645]        arch_call_rest_init+0xe/0x1b
> [  317.337645]        start_kernel+0x6eb/0x724
> [  317.337646]        x86_64_start_reservations+0x24/0x26
> [  317.337647]        x86_64_start_kernel+0xf4/0xfb
> [  317.337648]        secondary_startup_64+0xb6/0xc0

pi_lock --> rq->lock

> [  317.337649] -> #1 (&p->pi_lock){-.-.}:
> [  317.337651]        __lock_acquire+0x5b3/0xb40
> [  317.337652]        lock_acquire+0x126/0x280
> [  317.337653]        _raw_spin_lock_irqsave+0x3a/0x50
> [  317.337653]        try_to_wake_up+0xb4/0x1030
> [  317.337654]        wake_up_process+0x15/0x20
> [  317.337655]        __up+0xaa/0xc0
> [  317.337655]        up+0x55/0x60
> [  317.337656]        __up_console_sem+0x37/0x60
> [  317.337657]        console_unlock+0x3a0/0x750
> [  317.337658]        vprintk_emit+0x10d/0x340
> [  317.337658]        vprintk_default+0x1f/0x30
> [  317.337659]        vprintk_func+0x44/0xd4
> [  317.337660]        printk+0x9f/0xc5
> [  317.337660]        crng_reseed+0x3cc/0x440
> [  317.337661]        credit_entropy_bits+0x3e8/0x4f0
> [  317.337662]        random_ioctl+0x1eb/0x250
> [  317.337663]        do_vfs_ioctl+0x13e/0xa70
> [  317.337663]        ksys_ioctl+0x41/0x80
> [  317.337664]        __x64_sys_ioctl+0x43/0x4c
> [  317.337665]        do_syscall_64+0xcc/0x76c
> [  317.337666]        entry_SYSCALL_64_after_hwframe+0x49/0xbe

console_sem->lock --> pi_lock

This also covers console_sem->lock --> rq->lock, and maintains
pi_lock --> rq->lock

So we have

console_sem->lock --> pi_lock --> rq->lock

> [  317.337667] -> #0 ((console_sem).lock){-.-.}:
> [  317.337669]        check_prev_add+0x107/0xea0
> [  317.337670]        validate_chain+0x8fc/0x1200
> [  317.337671]        __lock_acquire+0x5b3/0xb40
> [  317.337671]        lock_acquire+0x126/0x280
> [  317.337672]        _raw_spin_lock_irqsave+0x3a/0x50
> [  317.337673]        down_trylock+0x16/0x50
> [  317.337674]        __down_trylock_console_sem+0x2b/0xa0
> [  317.337675]        console_trylock+0x16/0x60
> [  317.337676]        vprintk_emit+0x100/0x340
> [  317.337677]        vprintk_default+0x1f/0x30
> [  317.337678]        vprintk_func+0x44/0xd4
> [  317.337678]        printk+0x9f/0xc5
> [  317.337679]        __dump_page.cold.2+0x73/0x210
> [  317.337680]        dump_page+0x12/0x50
> [  317.337680]        has_unmovable_pages+0x3e9/0x4b0
> [  317.337681]        start_isolate_page_range+0x3b4/0x570
> [  317.337682]        __offline_pages+0x1ad/0xa10
> [  317.337683]        offline_pages+0x11/0x20
> [  317.337683]        memory_subsys_offline+0x7e/0xc0
> [  317.337684]        device_offline+0xd5/0x110
> [  317.337685]        state_store+0xc6/0xe0
> [  317.337686]        dev_attr_store+0x3f/0x60
> [  317.337686]        sysfs_kf_write+0x89/0xb0
> [  317.337687]        kernfs_fop_write+0x188/0x240
> [  317.337688]        __vfs_write+0x50/0xa0
> [  317.337688]        vfs_write+0x105/0x290
> [  317.337689]        ksys_write+0xc6/0x160
> [  317.337690]        __x64_sys_write+0x43/0x50
> [  317.337691]        do_syscall_64+0xcc/0x76c
> [  317.337691]        entry_SYSCALL_64_after_hwframe+0x49/0xbe

zone->lock --> console_sem->lock

So then we have

zone->lock --> console_sem->lock --> pi_lock --> rq->lock

vs. the reverse chain

rq->lock --> console_sem->lock

If I get this right.

> [  317.337693] other info that might help us debug this:
>
> [  317.337694] Chain exists of:
> [  317.337694]   (console_sem).lock --> &rq->lock --> &(&zone->lock)->rlock
>
> [  317.337699]  Possible unsafe locking scenario:
>
> [  317.337700]        CPU0                    CPU1
> [  317.337701]        ----                    ----
> [  317.337701]   lock(&(&zone->lock)->rlock);
> [  317.337703]                                lock(&rq->lock);
> [  317.337705]                                lock(&(&zone->lock)->rlock);
> [  317.337706]   lock((console_sem).lock);
>
> [  317.337708]  *** DEADLOCK ***
>
> [  317.337710] 8 locks held by test.sh/8738:
> [  317.337710]  #0: ffff8883940b5408 (sb_writers#4){.+.+}, at: vfs_write+0x25f/0x290
> [  317.337713]  #1: ffff889fce310280 (&of->mutex){+.+.}, at: kernfs_fop_write+0x128/0x240
> [  317.337716]  #2: ffff889feb6d4830 (kn->count#115){.+.+}, at: kernfs_fop_write+0x138/0x240
> [  317.337720]  #3: ffffffffb3762d40 (device_hotplug_lock){+.+.}, at: lock_device_hotplug_sysfs+0x16/0x50
> [  317.337723]  #4: ffff88981f0dc990 (&dev->mutex){....}, at: device_offline+0x70/0x110
> [  317.337726]  #5: ffffffffb3315250 (cpu_hotplug_lock.rw_sem){++++}, at: __offline_pages+0xbf/0xa10
> [  317.337729]  #6: ffffffffb35408b0 (mem_hotplug_lock.rw_sem){++++}, at: percpu_down_write+0x87/0x2f0
> [  317.337732]  #7: ffff88883fff4318 (&(&zone->lock)->rlock){-.-.}, at: start_isolate_page_range+0x1f7/0x570
> [  317.337736] stack backtrace:
> [  317.337737] CPU: 58 PID: 8738 Comm: test.sh Not tainted 5.3.0-next-20190917+ #9
> [  317.337738] Hardware name: HPE ProLiant DL560 Gen10/ProLiant DL560 Gen10, BIOS U34 05/21/2019
> [  317.337739] Call Trace:
> [  317.337739]  dump_stack+0x86/0xca
> [  317.337740]  print_circular_bug.cold.31+0x243/0x26e
> [  317.337741]  check_noncircular+0x29e/0x2e0
> [  317.337742]  ? debug_lockdep_rcu_enabled+0x4b/0x60
> [  317.337742]  ? print_circular_bug+0x120/0x120
> [  317.337743]  ? is_ftrace_trampoline+0x9/0x20
> [  317.337744]  ? kernel_text_address+0x59/0xc0
> [  317.337744]  ? __kernel_text_address+0x12/0x40
> [  317.337745]  check_prev_add+0x107/0xea0
> [  317.337746]  validate_chain+0x8fc/0x1200
> [  317.337746]  ? check_prev_add+0xea0/0xea0
> [  317.337747]  ? format_decode+0xd6/0x600
> [  317.337748]  ? file_dentry_name+0xe0/0xe0
> [  317.337749]  __lock_acquire+0x5b3/0xb40
> [  317.337749]  lock_acquire+0x126/0x280
> [  317.337750]  ? down_trylock+0x16/0x50
> [  317.337751]  ? vprintk_emit+0x100/0x340
> [  317.337752]  _raw_spin_lock_irqsave+0x3a/0x50
> [  317.337753]  ? down_trylock+0x16/0x50
> [  317.337753]  down_trylock+0x16/0x50
> [  317.337754]  ? vprintk_emit+0x100/0x340
> [  317.337755]  __down_trylock_console_sem+0x2b/0xa0
> [  317.337756]  console_trylock+0x16/0x60
> [  317.337756]  vprintk_emit+0x100/0x340
> [  317.337757]  vprintk_default+0x1f/0x30
> [  317.337758]  vprintk_func+0x44/0xd4
> [  317.337758]  printk+0x9f/0xc5
> [  317.337759]  ? kmsg_dump_rewind_nolock+0x64/0x64
> [  317.337760]  ? __dump_page+0x1d7/0x430
> [  317.337760]  __dump_page.cold.2+0x73/0x210
> [  317.337761]  dump_page+0x12/0x50
> [  317.337762]  has_unmovable_pages+0x3e9/0x4b0
> [  317.337763]  start_isolate_page_range+0x3b4/0x570
> [  317.337763]  ? unset_migratetype_isolate+0x280/0x280
> [  317.337764]  ? rcu_read_lock_bh_held+0xc0/0xc0
> [  317.337765]  __offline_pages+0x1ad/0xa10
> [  317.337765]  ? lock_acquire+0x126/0x280
> [  317.337766]  ? __add_memory+0xc0/0xc0
> [  317.337767]  ? __kasan_check_write+0x14/0x20
> [  317.337767]  ? __mutex_lock+0x344/0xcd0
> [  317.337768]  ? _raw_spin_unlock_irqrestore+0x49/0x50
> [  317.337769]  ? device_offline+0x70/0x110
> [  317.337770]  ? klist_next+0x1c1/0x1e0
> [  317.337770]  ? __mutex_add_waiter+0xc0/0xc0
> [  317.337771]  ? klist_next+0x10b/0x1e0
> [  317.337772]  ? klist_iter_exit+0x16/0x40
> [  317.337772]  ? device_for_each_child+0xd0/0x110
> [  317.337773]  offline_pages+0x11/0x20
> [  317.337774]  memory_subsys_offline+0x7e/0xc0
> [  317.337774]  device_offline+0xd5/0x110
> [  317.337775]  ? auto_online_blocks_show+0x70/0x70
> [  317.337776]  state_store+0xc6/0xe0
> [  317.337776]  dev_attr_store+0x3f/0x60
> [  317.337777]  ? device_match_name+0x40/0x40
> [  317.337778]  sysfs_kf_write+0x89/0xb0
> [  317.337778]  ? sysfs_file_ops+0xa0/0xa0
> [  317.337779]  kernfs_fop_write+0x188/0x240
> [  317.337780]  __vfs_write+0x50/0xa0
> [  317.337780]  vfs_write+0x105/0x290
> [  317.337781]  ksys_write+0xc6/0x160
> [  317.337782]  ? __x64_sys_read+0x50/0x50
> [  317.337782]  ? do_syscall_64+0x79/0x76c
> [  317.337783]  ? do_syscall_64+0x79/0x76c
> [  317.337784]  __x64_sys_write+0x43/0x50
> [  317.337784]  do_syscall_64+0xcc/0x76c
> [  317.337785]  ? trace_hardirqs_on_thunk+0x1a/0x20
> [  317.337786]  ? syscall_return_slowpath+0x210/0x210
> [  317.337787]  ? entry_SYSCALL_64_after_hwframe+0x3e/0xbe
> [  317.337787]  ? trace_hardirqs_off_caller+0x3a/0x150
> [  317.337788]  ? trace_hardirqs_off_thunk+0x1a/0x20
> [  317.337789]  entry_SYSCALL_64_after_hwframe+0x49/0xbe

Lovely.

-ss