[PATCH] rtc/cmos: Protect rtc_lock from interrupts
From: Chris Wilson
Date: Mon Mar 02 2020 - 08:45:04 EST
cmos_interrrupt() is called directly on resume paths, and by the
irqhandler. It currently assumes that it can only be invoked directly
from a hardirq, and so leads to the lockdep splat:
<4>[ 259.166718] WARNING: inconsistent lock state
<4>[ 259.166725] 5.6.0-rc3-CI-CI_DRM_8038+ #1 Tainted: G U
<4>[ 259.166727] --------------------------------
<4>[ 259.166731] inconsistent {IN-HARDIRQ-W} -> {HARDIRQ-ON-W} usage.
<4>[ 259.166741] rtcwake/4221 [HC0[0]:SC0[0]:HE1:SE1] takes:
<4>[ 259.166745] ffffffff82635198 (rtc_lock){?...}, at: cmos_interrupt+0x18/0x100
<4>[ 259.166768] {IN-HARDIRQ-W} state was registered at:
<4>[ 259.166780] lock_acquire+0xa7/0x1c0
<4>[ 259.166790] _raw_spin_lock+0x2a/0x40
<4>[ 259.166799] cmos_interrupt+0x18/0x100
<4>[ 259.166808] rtc_handler+0x75/0xc0
<4>[ 259.166822] acpi_ev_fixed_event_detect+0xf9/0x132
<4>[ 259.166829] acpi_ev_sci_xrupt_handler+0xb/0x28
<4>[ 259.166838] acpi_irq+0x13/0x30
<4>[ 259.166849] __handle_irq_event_percpu+0x41/0x2c0
<4>[ 259.166859] handle_irq_event_percpu+0x2b/0x70
<4>[ 259.166868] handle_irq_event+0x2f/0x50
<4>[ 259.166875] handle_fasteoi_irq+0x8e/0x150
<4>[ 259.166883] do_IRQ+0x7e/0x160
<4>[ 259.166891] ret_from_intr+0x0/0x35
<4>[ 259.166898] mwait_idle+0x7e/0x200
<4>[ 259.166905] do_idle+0x1bb/0x260
<4>[ 259.166912] cpu_startup_entry+0x14/0x20
<4>[ 259.166921] start_secondary+0x15f/0x1b0
<4>[ 259.166929] secondary_startup_64+0xa4/0xb0
<4>[ 259.167264] irq event stamp: 41593
<4>[ 259.167275] hardirqs last enabled at (41593): [<ffffffff81a394e7>] _raw_spin_unlock_irqrestore+0x47/0x60
<4>[ 259.167285] hardirqs last disabled at (41592): [<ffffffff81a3926d>] _raw_spin_lock_irqsave+0xd/0x50
<4>[ 259.167296] softirqs last enabled at (41568): [<ffffffff81e00385>] __do_softirq+0x385/0x47f
<4>[ 259.167306] softirqs last disabled at (41561): [<ffffffff810babaa>] irq_exit+0xba/0xc0
<4>[ 259.167309]
other info that might help us debug this:
<4>[ 259.167312] Possible unsafe locking scenario:
<4>[ 259.167314] CPU0
<4>[ 259.167316] ----
<4>[ 259.167319] lock(rtc_lock);
<4>[ 259.167324] <Interrupt>
<4>[ 259.167326] lock(rtc_lock);
<4>[ 259.167332]
*** DEADLOCK ***
<4>[ 259.167337] 6 locks held by rtcwake/4221:
<4>[ 259.167665] #0: ffff888175e89408 (sb_writers#5){.+.+}, at: vfs_write+0x1a4/0x1d0
<4>[ 259.167687] #1: ffff88816e112080 (&of->mutex){+.+.}, at: kernfs_fop_write+0xdd/0x1b0
<4>[ 259.167706] #2: ffff888179be85e0 (kn->count#236){.+.+}, at: kernfs_fop_write+0xe6/0x1b0
<4>[ 259.167728] #3: ffffffff82641e00 (system_transition_mutex){+.+.}, at: pm_suspend+0xb3/0x3b0
<4>[ 259.167748] #4: ffffffff826b3ea0 (acpi_scan_lock){+.+.}, at: acpi_suspend_begin+0x47/0x80
<4>[ 259.167763] #5: ffff888178f6b960 (&dev->mutex){....}, at: device_resume+0x92/0x1c0
<4>[ 259.167778]
stack backtrace:
<4>[ 259.167788] CPU: 1 PID: 4221 Comm: rtcwake Tainted: G U 5.6.0-rc3-CI-CI_DRM_8038+ #1
<4>[ 259.168106] Hardware name: Google Soraka/Soraka, BIOS MrChromebox-4.10 08/25/2019
<4>[ 259.168110] Call Trace:
<4>[ 259.168123] dump_stack+0x71/0x9b
<4>[ 259.168133] mark_lock+0x49a/0x500
<4>[ 259.168457] ? print_shortest_lock_dependencies+0x200/0x200
<4>[ 259.168469] __lock_acquire+0x6d4/0x15d0
<4>[ 259.168479] ? __lock_acquire+0x460/0x15d0
<4>[ 259.168490] lock_acquire+0xa7/0x1c0
<4>[ 259.168500] ? cmos_interrupt+0x18/0x100
<4>[ 259.168824] _raw_spin_lock+0x2a/0x40
<4>[ 259.168834] ? cmos_interrupt+0x18/0x100
<4>[ 259.168843] cmos_interrupt+0x18/0x100
<4>[ 259.169159] cmos_resume+0x1fd/0x290
<4>[ 259.169174] ? __acpi_pm_set_device_wakeup+0x24/0x100
<4>[ 259.169498] pnp_bus_resume+0x5e/0x90
<4>[ 259.169509] ? pnp_bus_suspend+0x10/0x10
<4>[ 259.169518] dpm_run_callback+0x64/0x280
<4>[ 259.169530] device_resume+0xd4/0x1c0
<4>[ 259.169540] ? dpm_watchdog_set+0x60/0x60
<4>[ 259.169860] dpm_resume+0x106/0x410
<4>[ 259.169870] ? dpm_resume_early+0x38c/0x3e0
<4>[ 259.169881] dpm_resume_end+0x8/0x10
<4>[ 259.170195] suspend_devices_and_enter+0x16f/0xbe0
<4>[ 259.170211] ? rcu_read_lock_sched_held+0x4d/0x80
<4>[ 259.170528] pm_suspend+0x344/0x3b0
<4>[ 259.170542] state_store+0x78/0xe0
<4>[ 259.170559] kernfs_fop_write+0x112/0x1b0
<4>[ 259.170579] vfs_write+0xb9/0x1d0
<4>[ 259.170896] ksys_write+0x9f/0xe0
<4>[ 259.170907] do_syscall_64+0x4f/0x220
<4>[ 259.170918] entry_SYSCALL_64_after_hwframe+0x49/0xbe
<4>[ 259.171229] RIP: 0033:0x7f9b4f3cb154
<4>[ 259.171240] Code: 89 02 48 c7 c0 ff ff ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 66 90 48 8d 05 b1 07 2e 00 8b 00 85 c0 75 13 b8 01 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 54 f3 c3 66 90 41 54 55 49 89 d4 53 48 89 f5
<4>[ 259.171245] RSP: 002b:00007ffc057ce438 EFLAGS: 00000246 ORIG_RAX: 0000000000000001
<4>[ 259.171253] RAX: ffffffffffffffda RBX: 0000000000000004 RCX: 00007f9b4f3cb154
<4>[ 259.171257] RDX: 0000000000000004 RSI: 000055f4b3d185a0 RDI: 000000000000000a
<4>[ 259.171572] RBP: 000055f4b3d185a0 R08: 000055f4b3d165e0 R09: 00007f9b4fab7740
<4>[ 259.171576] R10: 000055f4b3d14010 R11: 0000000000000246 R12: 000055f4b3d16500
<4>[ 259.171580] R13: 0000000000000004 R14: 00007f9b4f6a32a0 R15: 00007f9b4f6a2760
Fixes: c6d3a278cc12 ("rtc: cmos: acknowledge ACPI driven wake alarms upon resume")
Fixes: 311ee9c151ad ("rtc: cmos: allow using ACPI for RTC alarm instead of HPET")
Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
Cc: Zhang Rui <rui.zhang@xxxxxxxxx>
Cc: Alexandre Belloni <alexandre.belloni@xxxxxxxxxxx>
Cc: Alessandro Zummo <a.zummo@xxxxxxxxxxxx>
Cc: <stable@xxxxxxxxxxxxxxx> # v4.18+
---
drivers/rtc/rtc-cmos.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
diff --git a/drivers/rtc/rtc-cmos.c b/drivers/rtc/rtc-cmos.c
index b795fe4cbd2e..7754225c6f9d 100644
--- a/drivers/rtc/rtc-cmos.c
+++ b/drivers/rtc/rtc-cmos.c
@@ -651,8 +651,9 @@ static irqreturn_t cmos_interrupt(int irq, void *p)
{
u8 irqstat;
u8 rtc_control;
+ unsigned long flags;
- spin_lock(&rtc_lock);
+ spin_lock_irqsave(&rtc_lock, flags);
/* When the HPET interrupt handler calls us, the interrupt
* status is passed as arg1 instead of the irq number. But
@@ -686,7 +687,7 @@ static irqreturn_t cmos_interrupt(int irq, void *p)
hpet_mask_rtc_irq_bit(RTC_AIE);
CMOS_READ(RTC_INTR_FLAGS);
}
- spin_unlock(&rtc_lock);
+ spin_unlock_irqrestore(&rtc_lock, flags);
if (is_intr(irqstat)) {
rtc_update_irq(p, 1, irqstat);
--
2.25.1