Deadlock at reboot?

From: Maxime Ripard
Date: Fri Mar 04 2022 - 05:49:55 EST

Hi,

I'm not entirely sure who I'm supposed to send this to, so apologies if
I forgot somebody.

I've had an issue when rebooting on and off for a couple of monthes, but
I got a stacktrace from lockdep today:

[71721.872904] reboot: Restarting system
[71721.876743] ------------[ cut here ]------------
[71721.877081]
[71721.877084] ======================================================
[71721.877086] WARNING: possible circular locking dependency detected
[71721.877088] 5.17.0-rc6-next-20220303-v8+ #10 Not tainted
[71721.877092] ------------------------------------------------------
[71721.877094] swapper/1/0 is trying to acquire lock:
[71721.877097] ffffffc009610900 (console_owner){-.-.}-{0:0}, at: console_unlock+0x194/0x4e0
[71721.877117]
[71721.877117] but task is already holding lock:
[71721.877118] ffffff80fb7974d8 (&rq->__lock){-.-.}-{2:2}, at: raw_spin_rq_lock_nested+0x30/0x50
[71721.877131]
[71721.877131] which lock already depends on the new lock.
[71721.877131]
[71721.877132]
[71721.877132] the existing dependency chain (in reverse order) is:
[71721.877133]
[71721.877133] -> #4 (&rq->__lock){-.-.}-{2:2}:
[71721.877139] __lock_acquire+0x458/0xab0
[71721.877146] lock_acquire.part.0+0x9c/0x1e0
[71721.877152] lock_acquire+0x68/0x84
[71721.877157] _raw_spin_lock_nested+0x6c/0x84
[71721.877164] raw_spin_rq_lock_nested+0x30/0x50
[71721.877169] task_fork_fair+0x48/0x164
[71721.877175] sched_cgroup_fork+0xd4/0x130
[71721.877181] copy_process+0xd88/0x1380
[71721.877188] kernel_clone+0x98/0x364
[71721.877193] kernel_thread+0x78/0xb0
[71721.877198] rest_init+0x38/0x2c0
[71721.877202] arch_call_rest_init+0x1c/0x28
[71721.877207] start_kernel+0x644/0x684
[71721.877211] __primary_switched+0xa4/0xac
[71721.877214]
[71721.877214] -> #3 (&p->pi_lock){-.-.}-{2:2}:
[71721.877220] __lock_acquire+0x458/0xab0
[71721.877225] lock_acquire.part.0+0x9c/0x1e0
[71721.877230] lock_acquire+0x68/0x84
[71721.877234] _raw_spin_lock_irqsave+0x90/0xd0
[71721.877241] try_to_wake_up+0x60/0x460
[71721.877247] default_wake_function+0x28/0x50
[71721.877253] woken_wake_function+0x2c/0x40
[71721.877257] __wake_up_common+0xa0/0x180
[71721.877261] __wake_up_common_lock+0x80/0xd0
[71721.877265] __wake_up+0x24/0x30
[71721.877268] tty_wakeup+0x40/0x80
[71721.877274] tty_port_default_wakeup+0x2c/0x4c
[71721.877278] tty_port_tty_wakeup+0x24/0x30
[71721.877282] uart_write_wakeup+0x24/0x34
[71721.877288] serial8250_tx_chars+0x110/0x250
[71721.877293] serial8250_handle_irq.part.0+0x11c/0x170
[71721.877297] serial8250_default_handle_irq+0x58/0x9c
[71721.877301] serial8250_interrupt+0x5c/0xc0
[71721.877305] __handle_irq_event_percpu+0x9c/0x1b4
[71721.877310] handle_irq_event+0x58/0x130
[71721.877314] handle_fasteoi_irq+0xb0/0x1f0
[71721.877321] generic_handle_domain_irq+0x48/0x6c
[71721.877325] gic_handle_irq+0x88/0xc0
[71721.877329] call_on_irq_stack+0x2c/0x58
[71721.877334] do_interrupt_handler+0x88/0xa0
[71721.877339] el1_interrupt+0x34/0x50
[71721.877342] el1h_64_irq_handler+0x18/0x24
[71721.877346] el1h_64_irq+0x64/0x68
[71721.877349] arch_cpu_idle+0x18/0x2c
[71721.877353] default_idle_call+0x6c/0x104
[71721.877358] cpuidle_idle_call+0x158/0x190
[71721.877362] do_idle+0xdc/0x114
[71721.877365] cpu_startup_entry+0x34/0xa0
[71721.877369] rest_init+0x1bc/0x2c0
[71721.877373] arch_call_rest_init+0x1c/0x28
[71721.877376] start_kernel+0x644/0x684
[71721.877379] __primary_switched+0xa4/0xac
[71721.877383]
[71721.877383] -> #2 (&tty->write_wait){-.-.}-{2:2}:
[71721.877389] __lock_acquire+0x458/0xab0
[71721.877393] lock_acquire.part.0+0x9c/0x1e0
[71721.877398] lock_acquire+0x68/0x84
[71721.877403] _raw_spin_lock_irqsave+0x90/0xd0
[71721.877408] __wake_up_common_lock+0x60/0xd0
[71721.877412] __wake_up+0x24/0x30
[71721.877416] tty_wakeup+0x40/0x80
[71721.877421] tty_port_default_wakeup+0x2c/0x4c
[71721.877425] tty_port_tty_wakeup+0x24/0x30
[71721.877428] uart_write_wakeup+0x24/0x34
[71721.877433] serial8250_tx_chars+0x110/0x250
[71721.877438] serial8250_handle_irq.part.0+0x11c/0x170
[71721.877442] serial8250_default_handle_irq+0x58/0x9c
[71721.877446] serial8250_interrupt+0x5c/0xc0
[71721.877450] __handle_irq_event_percpu+0x9c/0x1b4
[71721.877454] handle_irq_event+0x58/0x130
[71721.877459] handle_fasteoi_irq+0xb0/0x1f0
[71721.877464] generic_handle_domain_irq+0x48/0x6c
[71721.877468] gic_handle_irq+0x88/0xc0
[71721.877472] call_on_irq_stack+0x2c/0x58
[71721.877476] do_interrupt_handler+0x88/0xa0
[71721.877481] el1_interrupt+0x34/0x50
[71721.877484] el1h_64_irq_handler+0x18/0x24
[71721.877487] el1h_64_irq+0x64/0x68
[71721.877491] arch_cpu_idle+0x18/0x2c
[71721.877494] default_idle_call+0x6c/0x104
[71721.877499] cpuidle_idle_call+0x158/0x190
[71721.877503] do_idle+0xdc/0x114
[71721.877507] cpu_startup_entry+0x34/0xa0
[71721.877510] rest_init+0x1bc/0x2c0
[71721.877514] arch_call_rest_init+0x1c/0x28
[71721.877518] start_kernel+0x644/0x684
[71721.877521] __primary_switched+0xa4/0xac
[71721.877524]
[71721.877524] -> #1 (&port_lock_key){-.-.}-{2:2}:
[71721.877529] __lock_acquire+0x458/0xab0
[71721.877534] lock_acquire.part.0+0x9c/0x1e0
[71721.877539] lock_acquire+0x68/0x84
[71721.877543] _raw_spin_lock_irqsave+0x90/0xd0
[71721.877549] serial8250_console_write+0x1b4/0x3a0
[71721.877554] univ8250_console_write+0x30/0x40
[71721.877557] call_console_drivers.constprop.0+0x100/0x1fc
[71721.877561] console_unlock+0x1ec/0x4e0
[71721.877564] register_console+0x13c/0x2b0
[71721.877567] uart_add_one_port+0x33c/0x370
[71721.877573] serial8250_register_8250_port+0x328/0x420
[71721.877576] bcm2835aux_serial_probe+0x1c0/0x290
[71721.877585] platform_probe+0x74/0xf0
[71721.877591] really_probe.part.0+0xa4/0x300
[71721.877596] __driver_probe_device+0xa0/0x150
[71721.877600] driver_probe_device+0xb4/0x150
[71721.877604] __driver_attach+0x110/0x1b0
[71721.877608] bus_for_each_dev+0x7c/0xe0
[71721.877612] driver_attach+0x30/0x3c
[71721.877616] bus_add_driver+0x15c/0x200
[71721.877620] driver_register+0x84/0x140
[71721.877624] __platform_driver_register+0x34/0x40
[71721.877630] bcm2835aux_serial_driver_init+0x28/0x34
[71721.877634] do_one_initcall+0x54/0x1a0
[71721.877638] kernel_init_freeable+0x1dc/0x25c
[71721.877642] kernel_init+0x30/0x140
[71721.877645] ret_from_fork+0x10/0x20
[71721.877649]
[71721.877649] -> #0 (console_owner){-.-.}-{0:0}:
[71721.877654] check_prev_add+0xac/0xbcc
[71721.877659] validate_chain+0x420/0x5d0
[71721.877664] __lock_acquire+0x458/0xab0
[71721.877668] lock_acquire.part.0+0x9c/0x1e0
[71721.877673] lock_acquire+0x68/0x84
[71721.877678] console_unlock+0x1dc/0x4e0
[71721.877681] vprintk_emit+0x1c8/0x1ec
[71721.877684] vprintk_default+0x44/0x50
[71721.877688] vprintk+0xc4/0x110
[71721.877691] _printk+0x68/0x90
[71721.877697] report_bug+0x104/0x130
[71721.877704] bug_handler+0x2c/0x80
[71721.877707] call_break_hook+0x70/0x84
[71721.877711] brk_handler+0x28/0x70
[71721.877714] do_debug_exception+0xa4/0x130
[71721.877721] el1_dbg+0x70/0x8c
[71721.877724] el1h_64_sync_handler+0x94/0xd0
[71721.877728] el1h_64_sync+0x64/0x68
[71721.877731] set_task_cpu+0x1b4/0x1f0
[71721.877736] detach_tasks+0x148/0x380
[71721.877741] load_balance+0x240/0x750
[71721.877746] rebalance_domains+0x3c0/0x4d0
[71721.877751] _nohz_idle_balance.constprop.0.isra.0+0x1ec/0x380
[71721.877757] run_rebalance_domains+0x70/0x80
[71721.877762] __do_softirq+0x1a0/0x37c
[71721.877765] __irq_exit_rcu+0x120/0x160
[71721.877770] irq_exit_rcu+0x1c/0x44
[71721.877774] el1_interrupt+0x38/0x50
[71721.877777] el1h_64_irq_handler+0x18/0x24
[71721.877780] el1h_64_irq+0x64/0x68
[71721.877783] arch_cpu_idle+0x18/0x2c
[71721.877787] default_idle_call+0x6c/0x104
[71721.877792] cpuidle_idle_call+0x158/0x190
[71721.877796] do_idle+0xdc/0x114
[71721.877799] cpu_startup_entry+0x30/0xa0
[71721.877803] secondary_start_kernel+0xdc/0x100
[71721.877809] __secondary_switched+0xa0/0xa4
[71721.877814]
[71721.877814] other info that might help us debug this:
[71721.877814]
[71721.877815] Chain exists of:
[71721.877815] console_owner --> &p->pi_lock --> &rq->__lock
[71721.877815]
[71721.877821] Possible unsafe locking scenario:
[71721.877821]
[71721.877823] CPU0 CPU1
[71721.877824] ---- ----
[71721.877825] lock(&rq->__lock);
[71721.877827] lock(&p->pi_lock);
[71721.877830] lock(&rq->__lock);
[71721.877832] lock(console_owner);
[71721.877835]
[71721.877835] *** DEADLOCK ***
[71721.877835]
[71721.877836] 3 locks held by swapper/1/0:
[71721.877839] #0: ffffffc009681fe8 (rcu_read_lock){....}-{1:2}, at: rebalance_domains+0x8c/0x4d0
[71721.877850] #1: ffffff80fb7974d8 (&rq->__lock){-.-.}-{2:2}, at: raw_spin_rq_lock_nested+0x30/0x50
[71721.877861] #2: ffffffc009610790 (console_lock){+.+.}-{0:0}, at: console_trylock_spinning+0x20/0x1a4
[71721.877870]
[71721.877870] stack backtrace:
[71721.877872] CPU: 1 PID: 0 Comm: swapper/1 Not tainted 5.17.0-rc6-next-20220303-v8+ #10
[71721.877877] Hardware name: Raspberry Pi 4 Model B Rev 1.1 (DT)
[71721.877880] Call trace:
[71721.877881] dump_backtrace.part.0+0xd4/0xe0
[71721.877884] show_stack+0x24/0x80
[71721.877888] dump_stack_lvl+0x8c/0xb8
[71721.877893] dump_stack+0x18/0x34
[71721.877896] print_circular_bug+0x1f8/0x200
[71721.877901] check_noncircular+0x130/0x144
[71721.877906] check_prev_add+0xac/0xbcc
[71721.877910] validate_chain+0x420/0x5d0
[71721.877915] __lock_acquire+0x458/0xab0
[71721.877920] lock_acquire.part.0+0x9c/0x1e0
[71721.877924] lock_acquire+0x68/0x84
[71721.877929] console_unlock+0x1dc/0x4e0
[71721.877932] vprintk_emit+0x1c8/0x1ec
[71721.877936] vprintk_default+0x44/0x50
[71721.877939] vprintk+0xc4/0x110
[71721.877943] _printk+0x68/0x90
[71721.877948] report_bug+0x104/0x130
[71721.877953] bug_handler+0x2c/0x80
[71721.877956] call_break_hook+0x70/0x84
[71721.877960] brk_handler+0x28/0x70
[71721.877964] do_debug_exception+0xa4/0x130
[71721.877970] el1_dbg+0x70/0x8c
[71721.877973] el1h_64_sync_handler+0x94/0xd0
[71721.877976] el1h_64_sync+0x64/0x68
[71721.877980] set_task_cpu+0x1b4/0x1f0
[71721.877985] detach_tasks+0x148/0x380
[71721.877990] load_balance+0x240/0x750
[71721.877995] rebalance_domains+0x3c0/0x4d0
[71721.878000] _nohz_idle_balance.constprop.0.isra.0+0x1ec/0x380
[71721.878006] run_rebalance_domains+0x70/0x80
[71721.878011] __do_softirq+0x1a0/0x37c
[71721.878014] __irq_exit_rcu+0x120/0x160
[71721.878018] irq_exit_rcu+0x1c/0x44
[71721.878022] el1_interrupt+0x38/0x50
[71721.878026] el1h_64_irq_handler+0x18/0x24
[71721.878029] el1h_64_irq+0x64/0x68
[71721.878032] arch_cpu_idle+0x18/0x2c
[71721.878035] default_idle_call+0x6c/0x104
[71721.878041] cpuidle_idle_call+0x158/0x190
[71721.878045] do_idle+0xdc/0x114
[71721.878048] cpu_startup_entry+0x30/0xa0
[71721.878052] secondary_start_kernel+0xdc/0x100
[71721.878057] __secondary_switched+0xa0/0xa4
[71722.926930] WARNING: CPU: 1 PID: 0 at kernel/sched/core.c:3052 set_task_cpu+0x1b4/0x1f0
[71722.935049] Modules linked in:
[71722.938139] CPU: 1 PID: 0 Comm: swapper/1 Not tainted 5.17.0-rc6-next-20220303-v8+ #10
[71722.946164] Hardware name: Raspberry Pi 4 Model B Rev 1.1 (DT)
[71722.950095] SMP: failed to stop secondary CPUs 0-1

The board would then require a power cycle to boot again (the changes
were some registers changes in KMS, nothing that would affect this part
of the code)

I'm not entirely sure how to fix it. Some versions seems to be more
affected than others. next-20220303 seems to trigger it 100% of the
time, some older versions would reboot fine. Like I said, it's been on
and off, so I think the source has been there for some time, whether it
works or not depends probably on subtle timings or something like that.

Maxime

Attachment: signature.asc
Description: PGP signature