On 7/3/24 03:28, Yang Yang wrote:
Configuration for sbq:
depth=64, wake_batch=6, shift=6, map_nr=1
1. There are 64 requests in progress:
map->word = 0xFFFFFFFFFFFFFFFF
2. After all the 64 requests complete, and no more requests come:
map->word = 0xFFFFFFFFFFFFFFFF, map->cleared = 0xFFFFFFFFFFFFFFFF
3. Now two tasks try to allocate requests:
T1: T2:
__blk_mq_get_tag .
__sbitmap_queue_get .
sbitmap_get .
sbitmap_find_bit .
sbitmap_find_bit_in_word .
__sbitmap_get_word -> nr=-1 __blk_mq_get_tag
sbitmap_deferred_clear __sbitmap_queue_get
/* map->cleared=0xFFFFFFFFFFFFFFFF */ sbitmap_find_bit
if (!READ_ONCE(map->cleared)) sbitmap_find_bit_in_word
return false; __sbitmap_get_word -> nr=-1
mask = xchg(&map->cleared, 0) sbitmap_deferred_clear
atomic_long_andnot() /* map->cleared=0 */
if (!(map->cleared))
return false;
/*
* map->cleared is cleared by T1
* T2 fail to acquire the tag
*/
4. T2 is the sole tag waiter. When T1 puts the tag, T2 cannot be woken
up due to the wake_batch being set at 6. If no more requests come, T1
will wait here indefinitely.
This patch achieves two purposes:
1. Check on ->cleared and update on both ->cleared and ->word need to
be done atomically, and using spinlock could be the simplest solution.
So revert commit 661d4f55a794 ("sbitmap: remove swap_lock"), which
may cause potential race.
2. Add extra check in sbitmap_deferred_clear(), to identify whether
->word has free bits.
Fixes: 661d4f55a794 ("sbitmap: remove swap_lock")
Is it blamed right? Considering that the revert alone doesn't fix
the problem, it sounds like the 2nd step might need to be ported
to kernels even without the blamed commit.
--
Pavel Begunkov
