From c0f091a0f8bd8ba9038197032408eab50fe26d91 Mon Sep 17 00:00:00 2001
From: Tim Chase <tim@onlight.com>
Date: Tue, 5 Jan 2016 13:25:23 -0600
Subject: [PATCH] Update arc_c under a mutex

The arc_c value can be updated concurrently by multiple threads including
the arc reclaim thread, kswapd, kthreadd and user processes.  This patch
updates it under a mutex to close the race between the checking of its
value and subsequent updates.

Reverts: 935434ef
Fixes: #3904
Fixes: #4161
---
 module/zfs/arc.c | 14 +++++++++++---
 1 file changed, 11 insertions(+), 3 deletions(-)

diff --git a/module/zfs/arc.c b/module/zfs/arc.c
index 0741633696b2..9726fcfbafd4 100644
--- a/module/zfs/arc.c
+++ b/module/zfs/arc.c
@@ -649,6 +649,7 @@ static kmutex_t arc_prune_mtx;
 static taskq_t *arc_prune_taskq;
 static arc_buf_t *arc_eviction_list;
 static arc_buf_hdr_t arc_eviction_hdr;
+static kmutex_t arc_c_mtx;
 
 #define	GHOST_STATE(state)	\
 	((state) == arc_mru_ghost || (state) == arc_mfu_ghost ||	\
@@ -3179,6 +3180,7 @@ arc_flush(spa_t *spa, boolean_t retry)
 void
 arc_shrink(int64_t to_free)
 {
+	mutex_enter(&arc_c_mtx);
 	if (arc_c > arc_c_min) {
 
 		if (arc_c > arc_c_min + to_free)
@@ -3195,8 +3197,11 @@ arc_shrink(int64_t to_free)
 		ASSERT((int64_t)arc_p >= 0);
 	}
 
-	if (arc_size > arc_c)
+	if (arc_size > arc_c) {
+		mutex_exit(&arc_c_mtx);
 		(void) arc_adjust();
+	} else
+		mutex_exit(&arc_c_mtx);
 }
 
 typedef enum free_memory_reason_t {
@@ -3762,8 +3767,8 @@ arc_adapt(int bytes, arc_state_t *state)
 	 * If we're within (2 * maxblocksize) bytes of the target
 	 * cache size, increment the target cache size
 	 */
-	ASSERT3U(arc_c, >=, 2ULL << SPA_MAXBLOCKSHIFT);
-	arc_c = MAX(arc_c, 2ULL << SPA_MAXBLOCKSHIFT);
+	VERIFY3U(arc_c, >=, 2ULL << SPA_MAXBLOCKSHIFT);
+	mutex_enter(&arc_c_mtx);
 	if (arc_size >= arc_c - (2ULL << SPA_MAXBLOCKSHIFT)) {
 		atomic_add_64(&arc_c, (int64_t)bytes);
 		if (arc_c > arc_c_max)
@@ -3773,6 +3778,7 @@ arc_adapt(int bytes, arc_state_t *state)
 		if (arc_p > arc_c)
 			arc_p = arc_c;
 	}
+	mutex_exit(&arc_c_mtx);
 	ASSERT((int64_t)arc_p >= 0);
 }
 
@@ -5442,6 +5448,7 @@ arc_init(void)
 	arc_eviction_list = NULL;
 	mutex_init(&arc_prune_mtx, NULL, MUTEX_DEFAULT, NULL);
 	bzero(&arc_eviction_hdr, sizeof (arc_buf_hdr_t));
+	mutex_init(&arc_c_mtx, NULL, MUTEX_DEFAULT, NULL);
 
 	arc_prune_taskq = taskq_create("arc_prune", max_ncpus, defclsyspri,
 	    max_ncpus, INT_MAX, TASKQ_PREPOPULATE | TASKQ_DYNAMIC);
@@ -5542,6 +5549,7 @@ arc_fini(void)
 	list_destroy(&arc_prune_list);
 	mutex_destroy(&arc_prune_mtx);
 	mutex_destroy(&arc_reclaim_lock);
+	mutex_destroy(&arc_c_mtx);
 	cv_destroy(&arc_reclaim_thread_cv);
 	cv_destroy(&arc_reclaim_waiters_cv);