--- a/module/zfs/dbuf.c	2026-04-24 14:20:12.565390796 -0400
+++ b/module/zfs/dbuf.c	2026-04-24 14:20:40.711198653 -0400
@@ -864,23 +864,16 @@
 dbuf_evict_notify(uint64_t size)
 {
 	/*
-	 * We check if we should evict without holding the dbuf_evict_lock,
-	 * because it's OK to occasionally make the wrong decision here,
-	 * and grabbing the lock results in massive lock contention.
+	 * Wake the dedicated eviction thread when the cache
+	 * exceeds target. Inline eviction from write-completion
+	 * context is intentionally avoided: it adds a second
+	 * multilist sublist lock to every write, and when
+	 * kmem_cache_free is slow (__free_frozen_pages zone lock
+	 * contention on 6.14+), the cascade stalls txg_sync.
+	 * See openzfs#18426.
 	 */
-	if (size > dbuf_cache_target_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();
-		}
+	if (size > dbuf_cache_target_bytes())
 		cv_signal(&dbuf_evict_cv);
-	}
 }
 
 /*