Skip dbuf_evict_one() from dbuf_evict_notify() for reclaim thread

Avoid calling dbuf_evict_one() from memory reclaim contexts (e.g. Linux
kswapd, FreeBSD pagedaemon). This prevents deadlock caused by reclaim
threads waiting for the dbuf hash lock in the call sequence:
dbuf_evict_one -> dbuf_destroy -> arc_buf_destroy

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Alexander Motin <alexander.motin@TrueNAS.com>
Signed-off-by: Kaitlin Hoang <kthoang@amazon.com>
Closes #17561
This commit is contained in:
khoang98
2025-08-01 19:47:41 -04:00
committed by GitHub
parent 1aec627c60
commit 0f8a1105ee
6 changed files with 46 additions and 1 deletions
+9 -1
View File
@@ -866,8 +866,16 @@ dbuf_evict_notify(uint64_t size)
* and grabbing the lock results in massive lock contention.
*/
if (size > dbuf_cache_target_bytes()) {
if (size > dbuf_cache_hiwater_bytes())
/*
* Avoid calling dbuf_evict_one() from memory reclaim context
* (e.g. Linux kswapd, FreeBSD pagedaemon) to prevent deadlocks.
* Memory reclaim threads can get stuck waiting for the dbuf
* hash lock.
*/
if (size > dbuf_cache_hiwater_bytes() &&
!current_is_reclaim_thread()) {
dbuf_evict_one();
}
cv_signal(&dbuf_evict_cv);
}
}