Skip to contentSkip to conversation

timer: fix a debugobjects warning in del_timer()

PATCH In review Qian Cai 1 participant 2 messages Dec 7, 2019 – Dec 8, 2019 High-resolution timers, timer wheel, clockevents

lore
Share & tools
lore

Qian Cai

author
#0 lore raw

PATCH 1/1 timer: fix a debugobjects warning in del_timer()

Since the commit e33026831bdb ("x86/intel_rdt/mbm: Handle counter overflow"), it will generate a debugobjects warning while offlining CPUs.

ODEBUG: assert_init not available (active state 0) object type:
timer_list hint: 0x0

WARNING: CPU: 143 PID: 789 at lib/debugobjects.c:484
debug_print_object+0xfe/0x140
Hardware name: HP Synergy 680 Gen9/Synergy 680 Gen9 Compute Module, BIOS
I40 05/23/2018
RIP: 0010:debug_print_object+0xfe/0x140
Call Trace:
 debug_object_assert_init+0x1f5/0x240
 del_timer+0x6f/0xf0
 try_to_grab_pending+0x42/0x3c0
 cancel_delayed_work+0x7d/0x150
 resctrl_offline_cpu+0x3c0/0x520
 cpuhp_invoke_callback+0x197/0x1120
 cpuhp_thread_fun+0x252/0x2f0
 smpboot_thread_fn+0x255/0x440
 kthread+0x1e6/0x210
 ret_from_fork+0x3a/0x50

This is because in domain_remove_cpu() when "cpu == d->mbm_work_cpu", it calls cancel_delayed_work(&d->mbm_over) to deactivate the timer, and then mbm_setup_overflow_handler() calls schedule_delayed_work_on() with 0 delay which does not activiate the timer in __queue_delayed_work().

Later, when the last CPU in the same L3 cache goes offline, it calls cancel_delayed_work(&d->mbm_over) again in domain_remove_cpu() and trigger the warning because the timer is still inactive.

Since del_timer() could be called on both active and inactive timers, debug_assert_init() should be called only when there is an active timer.

---
 kernel/time/timer.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
kernel/time/timer.c +2 -2
index 4820823515e9..90a7658dca07 100644
@@ -1193,9 +1193,9 @@ int del_timer(struct timer_list *timer)
 	unsigned long flags;
 	int ret = 0;
 
-	debug_assert_init(timer);
-
 	if (timer_pending(timer)) {
+		debug_assert_init(timer);
+
 		base = lock_timer_base(timer, &flags);
 		ret = detach_if_pending(timer, base, true);
 		raw_spin_unlock_irqrestore(&base->lock, flags);
Signature · 1 line
2.21.0 (Apple Git-122.2)
author
#1 lore raw
Quoted 32 lines · Qian Cai

Since the commit e33026831bdb ("x86/intel_rdt/mbm: Handle counter overflow"), it will generate a debugobjects warning while offlining CPUs.

ODEBUG: assert_init not available (active state 0) object type:
timer_list hint: 0x0

WARNING: CPU: 143 PID: 789 at lib/debugobjects.c:484
debug_print_object+0xfe/0x140
Hardware name: HP Synergy 680 Gen9/Synergy 680 Gen9 Compute Module, BIOS
I40 05/23/2018
RIP: 0010:debug_print_object+0xfe/0x140
Call Trace:
debug_object_assert_init+0x1f5/0x240
del_timer+0x6f/0xf0
try_to_grab_pending+0x42/0x3c0
cancel_delayed_work+0x7d/0x150
resctrl_offline_cpu+0x3c0/0x520
cpuhp_invoke_callback+0x197/0x1120
cpuhp_thread_fun+0x252/0x2f0
smpboot_thread_fn+0x255/0x440
kthread+0x1e6/0x210
ret_from_fork+0x3a/0x50

This is because in domain_remove_cpu() when "cpu == d->mbm_work_cpu", it calls cancel_delayed_work(&d->mbm_over) to deactivate the timer, and then mbm_setup_overflow_handler() calls schedule_delayed_work_on() with 0 delay which does not activiate the timer in __queue_delayed_work().

Quoted 8 lines

Later, when the last CPU in the same L3 cache goes offline, it calls cancel_delayed_work(&d->mbm_over) again in domain_remove_cpu() and trigger the warning because the timer is still inactive.

Since del_timer() could be called on both active and inactive timers, debug_assert_init() should be called only when there is an active timer.

Self-NACK this and I’ll post a more correct patch.