assertion failed in arc_wait_for_eviction()

If the system is very low on memory (specifically,
`arc_free_memory() < arc_sys_free/2`, i.e. less than 1/16th of RAM
free), `arc_evict_state_impl()` will defer wakups.  In this case, the
arc_evict_waiter_t's remain on the list, even though `arc_evict_count`
has been incremented past their `aew_count`.

The problem is that `arc_wait_for_eviction()` assumes that if there are
waiters on the list, the count they are waiting for has not yet been
reached.  However, the deferred wakeups may violate this, causing
`ASSERT(last->aew_count > arc_evict_count)` to fail.

This commit resolves the issue by having new waiters use the greater of
`arc_evict_count` and the last `aew_count`.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: George Wilson <gwilson@delphix.com>
Reviewed-by: George Amanakis <gamanakis@gmail.com>
Signed-off-by: Matthew Ahrens <mahrens@delphix.com>
Closes #11285
Closes #11397

diff --git a/module/zfs/arc.c b/module/zfs/arc.c
index af6f52e97c90823bdc3aea1dae92b80cc55b08ea..0d776832cd8457658206fe3f4e132d5021f0322b 100644 (file)
--- a/module/zfs/arc.c
+++ b/module/zfs/arc.c
@@ -4050,7 +4050,7 @@ arc_evict_state_impl(multilist_t *ml, int idx, arc_buf_hdr_t *marker,
        mutex_enter(&arc_evict_lock);
        arc_evict_count += bytes_evicted;
 
-       if ((int64_t)(arc_free_memory() - arc_sys_free / 2) > 0) {
+       if (arc_free_memory() > arc_sys_free / 2) {
                arc_evict_waiter_t *aw;
                while ((aw = list_head(&arc_evict_waiters)) != NULL &&
                    aw->aew_count <= arc_evict_count) {
@@ -5136,14 +5136,20 @@ arc_wait_for_eviction(uint64_t amount)
                        list_link_init(&aw.aew_node);
                        cv_init(&aw.aew_cv, NULL, CV_DEFAULT, NULL);
 
-                       arc_evict_waiter_t *last =
-                           list_tail(&arc_evict_waiters);
-                       if (last != NULL) {
-                               ASSERT3U(last->aew_count, >, arc_evict_count);
-                               aw.aew_count = last->aew_count + amount;
-                       } else {
-                               aw.aew_count = arc_evict_count + amount;
+                       uint64_t last_count = 0;
+                       if (!list_is_empty(&arc_evict_waiters)) {
+                               arc_evict_waiter_t *last =
+                                   list_tail(&arc_evict_waiters);
+                               last_count = last->aew_count;
                        }
+                       /*
+                        * Note, the last waiter's count may be less than
+                        * arc_evict_count if we are low on memory in which
+                        * case arc_evict_state_impl() may have deferred
+                        * wakeups (but still incremented arc_evict_count).
+                        */
+                       aw.aew_count =
+                           MAX(last_count, arc_evict_count) + amount;
 
                        list_insert_tail(&arc_evict_waiters, &aw);