OS-2437 arc_get_data_buf() blocks in vmem_alloc() instead of evicting20130905release-20130905

4 files changed, 54 insertions, 7 deletions

diff --git a/usr/src/lib/libzpool/common/sys/zfs_context.h b/usr/src/lib/libzpool/common/sys/zfs_context.h
index 3aa58fb05a..b6f6304f59 100644
--- a/usr/src/lib/libzpool/common/sys/zfs_context.h
+++ b/usr/src/lib/libzpool/common/sys/zfs_context.h
@@ -22,7 +22,7 @@
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
  * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
  * Copyright (c) 2013 by Delphix. All rights reserved.
- * Copyright (c) 2012, Joyent, Inc. All rights reserved.
+ * Copyright (c) 2013, Joyent, Inc. All rights reserved.
  */
 
 #ifndef _SYS_ZFS_CONTEXT_H
@@ -305,6 +305,7 @@ extern void kstat_runq_back_to_waitq(kstat_io_t *);
 #define	KM_SLEEP		UMEM_NOFAIL
 #define	KM_PUSHPAGE		KM_SLEEP
 #define	KM_NOSLEEP		UMEM_DEFAULT
+#define	KM_NORMALPRI		0
 #define	KMC_NODEBUG		UMC_NODEBUG
 #define	KMC_NOTOUCH		0	/* not needed for userland caches */
 #define	kmem_alloc(_s, _f)	umem_alloc(_s, _f)
diff --git a/usr/src/uts/common/fs/zfs/arc.c b/usr/src/uts/common/fs/zfs/arc.c
index dfd02dc69f..0b762ebdb9 100644
--- a/usr/src/uts/common/fs/zfs/arc.c
+++ b/usr/src/uts/common/fs/zfs/arc.c
@@ -280,6 +280,7 @@ typedef struct arc_stats {
 	 * not from the spa we're trying to evict from.
 	 */
 	kstat_named_t arcstat_evict_skip;
+	kstat_named_t arcstat_evict_allocfail;
 	kstat_named_t arcstat_evict_l2_cached;
 	kstat_named_t arcstat_evict_l2_eligible;
 	kstat_named_t arcstat_evict_l2_ineligible;
@@ -346,6 +347,7 @@ static arc_stats_t arc_stats = {
 	{ "recycle_miss",		KSTAT_DATA_UINT64 },
 	{ "mutex_miss",			KSTAT_DATA_UINT64 },
 	{ "evict_skip",			KSTAT_DATA_UINT64 },
+	{ "evict_allocfail",		KSTAT_DATA_UINT64 },
 	{ "evict_l2_cached",		KSTAT_DATA_UINT64 },
 	{ "evict_l2_eligible",		KSTAT_DATA_UINT64 },
 	{ "evict_l2_ineligible",	KSTAT_DATA_UINT64 },
@@ -2513,15 +2515,27 @@ arc_get_data_buf(arc_buf_t *buf)
 	 */
 	if (!arc_evict_needed(type)) {
 		if (type == ARC_BUFC_METADATA) {
-			buf->b_data = zio_buf_alloc(size);
-			arc_space_consume(size, ARC_SPACE_DATA);
+			buf->b_data = zio_buf_alloc_canfail(size);
+			if (buf->b_data != NULL) {
+				arc_space_consume(size, ARC_SPACE_DATA);
+				goto out;
+			}
 		} else {
 			ASSERT(type == ARC_BUFC_DATA);
-			buf->b_data = zio_data_buf_alloc(size);
-			ARCSTAT_INCR(arcstat_data_size, size);
-			atomic_add_64(&arc_size, size);
+			buf->b_data = zio_data_buf_alloc_canfail(size);
+			if (buf->b_data != NULL) {
+				ARCSTAT_INCR(arcstat_data_size, size);
+				atomic_add_64(&arc_size, size);
+				goto out;
+			}
 		}
-		goto out;
+
+		/*
+		 * Memory allocation failed, presumably due to excessive
+		 * fragmentation; we'll bump a counter and drop into the ARC
+		 * eviction case.
+		 */
+		ARCSTAT_BUMP(arcstat_evict_allocfail);
 	}
 
 	/*
diff --git a/usr/src/uts/common/fs/zfs/sys/zio.h b/usr/src/uts/common/fs/zfs/sys/zio.h
index eb4b1f8e71..7b383e4a74 100644
--- a/usr/src/uts/common/fs/zfs/sys/zio.h
+++ b/usr/src/uts/common/fs/zfs/sys/zio.h
@@ -504,8 +504,10 @@ extern zio_t *zio_unique_parent(zio_t *cio);
 extern void zio_add_child(zio_t *pio, zio_t *cio);
 
 extern void *zio_buf_alloc(size_t size);
+extern void *zio_buf_alloc_canfail(size_t size);
 extern void zio_buf_free(void *buf, size_t size);
 extern void *zio_data_buf_alloc(size_t size);
+extern void *zio_data_buf_alloc_canfail(size_t size);
 extern void zio_data_buf_free(void *buf, size_t size);
 
 extern void zio_resubmit_stage_async(void *);
diff --git a/usr/src/uts/common/fs/zfs/zio.c b/usr/src/uts/common/fs/zfs/zio.c
index eb06151cee..dfd8fa3f65 100644
--- a/usr/src/uts/common/fs/zfs/zio.c
+++ b/usr/src/uts/common/fs/zfs/zio.c
@@ -22,6 +22,7 @@
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
  * Copyright (c) 2013 by Delphix. All rights reserved.
  * Copyright (c) 2011 Nexenta Systems, Inc. All rights reserved.
+ * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
  */
 
 #include <sys/zfs_context.h>
@@ -228,6 +229,20 @@ zio_buf_alloc(size_t size)
 }
 
 /*
+ * Same as zio_buf_alloc, but won't sleep in case memory cannot be allocated
+ * and will instead return immediately with a failure.
+ */
+void *
+zio_buf_alloc_canfail(size_t size)
+{
+	size_t c = (size - 1) >> SPA_MINBLOCKSHIFT;
+
+	ASSERT(c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);
+
+	return (kmem_cache_alloc(zio_buf_cache[c], KM_NOSLEEP | KM_NORMALPRI));
+}
+
+/*
  * Use zio_data_buf_alloc to allocate data.  The data will not appear in a
  * crashdump if the kernel panics.  This exists so that we will limit the amount
  * of ZFS data that shows up in a kernel crashdump.  (Thus reducing the amount
@@ -243,6 +258,21 @@ zio_data_buf_alloc(size_t size)
 	return (kmem_cache_alloc(zio_data_buf_cache[c], KM_PUSHPAGE));
 }
 
+/*
+ * Same as zio_data_buf_alloc, but won't sleep in case memory cannot be
+ * allocated and will instead return immediately with a failure.
+ */
+void *
+zio_data_buf_alloc_canfail(size_t size)
+{
+	size_t c = (size - 1) >> SPA_MINBLOCKSHIFT;
+
+	ASSERT(c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);
+
+	return (kmem_cache_alloc(zio_data_buf_cache[c],
+	    KM_NOSLEEP | KM_NORMALPRI));
+}
+
 void
 zio_buf_free(void *buf, size_t size)
 {