[PATCH next v2] sbitmap: fix lockup while swapping

From: Hugh Dickins
Date: Wed Sep 28 2022 - 00:08:06 EST


Commit 4acb83417cad ("sbitmap: fix batched wait_cnt accounting")
is a big improvement: without it, I had to revert to before commit
040b83fcecfb ("sbitmap: fix possible io hung due to lost wakeup")
to avoid the high system time and freezes which that had introduced.

Now okay on the NVME laptop, but 4acb83417cad is a disaster for heavy
swapping (kernel builds in low memory) on another: soon locking up in
sbitmap_queue_wake_up() (into which __sbq_wake_up() is inlined), cycling
around with waitqueue_active() but wait_cnt 0 . Here is a backtrace,
showing the common pattern of outer sbitmap_queue_wake_up() interrupted
before setting wait_cnt 0 back to wake_batch (in some cases other CPUs
are idle, in other cases they're spinning for a lock in dd_bio_merge()):

sbitmap_queue_wake_up < sbitmap_queue_clear < blk_mq_put_tag <
__blk_mq_free_request < blk_mq_free_request < __blk_mq_end_request <
scsi_end_request < scsi_io_completion < scsi_finish_command <
scsi_complete < blk_complete_reqs < blk_done_softirq < __do_softirq <
__irq_exit_rcu < irq_exit_rcu < common_interrupt < asm_common_interrupt <
_raw_spin_unlock_irqrestore < __wake_up_common_lock < __wake_up <
sbitmap_queue_wake_up < sbitmap_queue_clear < blk_mq_put_tag <
__blk_mq_free_request < blk_mq_free_request < dd_bio_merge <
blk_mq_sched_bio_merge < blk_mq_attempt_bio_merge < blk_mq_submit_bio <
__submit_bio < submit_bio_noacct_nocheck < submit_bio_noacct <
submit_bio < __swap_writepage < swap_writepage < pageout <
shrink_folio_list < evict_folios < lru_gen_shrink_lruvec <
shrink_lruvec < shrink_node < do_try_to_free_pages < try_to_free_pages <
__alloc_pages_slowpath < __alloc_pages < folio_alloc < vma_alloc_folio <
do_anonymous_page < __handle_mm_fault < handle_mm_fault <
do_user_addr_fault < exc_page_fault < asm_exc_page_fault

See how the process-context sbitmap_queue_wake_up() has been interrupted,
after bringing wait_cnt down to 0 (and in this example, after doing its
wakeups), before advancing wake_index and refilling wake_cnt: an
interrupt-context sbitmap_queue_wake_up() of the same sbq gets stuck.

I have almost no grasp of all the possible sbitmap races, and their
consequences: but __sbq_wake_up() can do nothing useful while wait_cnt 0,
so it is better if sbq_wake_ptr() skips on to the next ws in that case:
which fixes the lockup and shows no adverse consequence for me.

Signed-off-by: Hugh Dickins <hughd@xxxxxxxxxx>
---
v2: - v1 to __sbq_wake_up() broke out when this happens, but
v2 to sbq_wake_ptr() does better by skipping on to the next.
- added more comment and deleted dubious Fixes attribution.
- and apologies to Mr Axboe and all for my axbod typo

lib/sbitmap.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

--- a/lib/sbitmap.c
+++ b/lib/sbitmap.c
@@ -587,7 +587,7 @@ static struct sbq_wait_state *sbq_wake_p
for (i = 0; i < SBQ_WAIT_QUEUES; i++) {
struct sbq_wait_state *ws = &sbq->ws[wake_index];

- if (waitqueue_active(&ws->wait)) {
+ if (waitqueue_active(&ws->wait) && atomic_read(&ws->wait_cnt)) {
if (wake_index != atomic_read(&sbq->wake_index))
atomic_set(&sbq->wake_index, wake_index);
return ws;