md/raid5: fix 'out of memory' during raid cache recovery

This fixes the case when md array assembly fails because of raid cache recovery
unable to allocate a stripe, despite attempts to replay stripes and increase
cache size. This happens because stripes released by r5c_recovery_replay_stripes
and raid5_set_cache_size don't become available for allocation immediately.
Released stripes first are placed on conf->released_stripes list and require
md thread to merge them on conf->inactive_list before they can be allocated.

Patch allows final allocation attempt during cache recovery to wait for
new stripes to become availabe for allocation.

Cc: linux-raid@vger.kernel.org
Cc: Shaohua Li <shli@kernel.org>
Cc: linux-stable <stable@vger.kernel.org> # 4.10+
Fixes: b4c625c6 ("md/r5cache: r5cache recovery: part 1")
Signed-off-by: Alexei Naberezhnov <anaberezhnov@fb.com>
Signed-off-by: Song Liu <songliubraving@fb.com>

drivers/md/raid5-cache.c

@@ -1935,12 +1935,14 @@ r5l_recovery_replay_one_stripe(struct r5conf *conf,
 }
 
 static struct stripe_head *
-r5c_recovery_alloc_stripe(struct r5conf *conf,
-			  sector_t stripe_sect)
+r5c_recovery_alloc_stripe(
+		struct r5conf *conf,
+		sector_t stripe_sect,
+		int noblock)
 {
 	struct stripe_head *sh;
 
-	sh = raid5_get_active_stripe(conf, stripe_sect, 0, 1, 0);
+	sh = raid5_get_active_stripe(conf, stripe_sect, 0, noblock, 0);
 	if (!sh)
 		return NULL;  /* no more stripe available */
 
@@ -2150,7 +2152,7 @@ r5c_recovery_analyze_meta_block(struct r5l_log *log,
 						stripe_sect);
 
 		if (!sh) {
-			sh = r5c_recovery_alloc_stripe(conf, stripe_sect);
+			sh = r5c_recovery_alloc_stripe(conf, stripe_sect, 1);
 			/*
 			 * cannot get stripe from raid5_get_active_stripe
 			 * try replay some stripes
@@ -2159,16 +2161,25 @@ r5c_recovery_analyze_meta_block(struct r5l_log *log,
 				r5c_recovery_replay_stripes(
 					cached_stripe_list, ctx);
 				sh = r5c_recovery_alloc_stripe(
-					conf, stripe_sect);
+					conf, stripe_sect, 1);
 			}
 			if (!sh) {
+				int new_size = conf->min_nr_stripes * 2;
 				pr_debug("md/raid:%s: Increasing stripe cache size to %d to recovery data on journal.\n",
 					mdname(mddev),
-					conf->min_nr_stripes * 2);
-				raid5_set_cache_size(mddev,
-						     conf->min_nr_stripes * 2);
-				sh = r5c_recovery_alloc_stripe(conf,
-							       stripe_sect);
+					new_size);
+				ret = raid5_set_cache_size(mddev, new_size);
+				if (conf->min_nr_stripes <= new_size / 2) {
+					pr_err("md/raid:%s: Cannot increase cache size, ret=%d, new_size=%d, min_nr_stripes=%d, max_nr_stripes=%d\n",
+						mdname(mddev),
+						ret,
+						new_size,
+						conf->min_nr_stripes,
+						conf->max_nr_stripes);
+					return -ENOMEM;
+				}
+				sh = r5c_recovery_alloc_stripe(
+					conf, stripe_sect, 0);
 			}
 			if (!sh) {
 				pr_err("md/raid:%s: Cannot get enough stripes due to memory pressure. Recovery failed.\n",

drivers/md/raid5.c

@@ -6369,6 +6369,7 @@ raid5_show_stripe_cache_size(struct mddev *mddev, char *page)
 int
 raid5_set_cache_size(struct mddev *mddev, int size)
 {
+	int result = 0;
 	struct r5conf *conf = mddev->private;
 
 	if (size <= 16 || size > 32768)
@@ -6385,11 +6386,14 @@ raid5_set_cache_size(struct mddev *mddev, int size)
 
 	mutex_lock(&conf->cache_size_mutex);
 	while (size > conf->max_nr_stripes)
-		if (!grow_one_stripe(conf, GFP_KERNEL))
+		if (!grow_one_stripe(conf, GFP_KERNEL)) {
+			conf->min_nr_stripes = conf->max_nr_stripes;
+			result = -ENOMEM;
 			break;
+		}
 	mutex_unlock(&conf->cache_size_mutex);
 
-	return 0;
+	return result;
 }
 EXPORT_SYMBOL(raid5_set_cache_size);