Skip to contentSkip to conversation

flex_proportions: Make fprop_new_period() hardirq safe

PATCH Applied Jan Kara 2 participants 2 messages Jan 21

Subsystems (4)
lore
Share & tools
lore

Jan Kara

author reviewer
#0 lore raw

PATCH v2 1/1 flex_proportions: Make fprop_new_period() hardirq safe

Bernd has reported a lockdep splat from flexible proportions code that is essentially complaining about the following race:

<timer fires>
run_timer_softirq - we are in softirq context
call_timer_fn

    writeout_period
      fprop_new_period
        write_seqcount_begin(&p->sequence);

        <hardirq is raised>
        ...
        blk_mq_end_request()
	  blk_update_request()
	    ext4_end_bio()
	      folio_end_writeback()
		__wb_writeout_add()
		  __fprop_add_percpu_max()
		    if (unlikely(max_frac < FPROP_FRAC_BASE)) {
		      fprop_fraction_percpu()
			seq = read_seqcount_begin(&p->sequence);
			  - sees odd sequence so loops indefinitely

Note that a deadlock like this is only possible if the bdi has configured maximum fraction of writeout throughput which is very rare in general but frequent for example for FUSE bdis. To fix this problem we have to make sure write section of the sequence counter is irqsafe.

Fixes a91befde3503 ("lib/flex_proportions.c: remove local_irq_ops in fprop_new_period()")
---
 lib/flex_proportions.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

Sending v2 because I've messed up Andrew's address and forgot some tags in the first posting...

lib/flex_proportions.c +3 -2
index 84ecccddc771..012d5614efb9 100644
@@ -64,13 +64,14 @@ void fprop_global_destroy(struct fprop_global *p)
 bool fprop_new_period(struct fprop_global *p, int periods)
 {
 	s64 events = percpu_counter_sum(&p->events);
+	unsigned long flags;
 
 	/*
 	 * Don't do anything if there are no events.
 	 */
 	if (events <= 1)
 		return false;
-	preempt_disable_nested();
+	local_irq_save(flags);
 	write_seqcount_begin(&p->sequence);
 	if (periods < 64)
 		events -= events >> periods;
@@ -78,7 +79,7 @@ bool fprop_new_period(struct fprop_global *p, int periods)
 	percpu_counter_add(&p->events, -events);
 	p->period += periods;
 	write_seqcount_end(&p->sequence);
-	preempt_enable_nested();
+	local_irq_restore(flags);
 
 	return true;
 }
Signature · 1 line
2.51.0
Reviewed-by on v2 1/1
#1 lore raw
Quoted 2 lines · Jan Kara

Bernd has reported a lockdep splat from flexible proportions code that is essentially complaining about the following race:

[...]

Quoted 4 lines

Note that a deadlock like this is only possible if the bdi has configured maximum fraction of writeout throughput which is very rare in general but frequent for example for FUSE bdis. To fix this problem we have to make sure write section of the sequence counter is irqsafe.

Ah, that's why we haven't seen it reported before.

Quoted 5 lines
Fixes a91befde3503 ("lib/flex_proportions.c: remove local_irq_ops in fprop_new_period()")