svn commit: r275780 - head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs

[Xin LI]

Author: delphij
Date: Mon Dec 15 04:51:36 2014
New Revision: 275780
URL: https://svnweb.freebsd.org/changeset/base/275780

Log:
  MFV r275549:
  
  Add a loader tunable, vfs.zfs.arc_meta_min, which controls how much metadata
  ZFS should keep in ARC at minimum.
  
  In arc_evict(), when doing recycle, take more factors into account by
  applying the following policy:
  
    1. If no evictable data, evict metadata;
    2. If no evictable metadata, evict data;
    3. If we hit arc_meta_limit, evict metadata;
    4. If we haven't hit arc_meta_min, evict data;
    5* (Illumos only, not present in new FreeBSD code, yet) evict the oldest
       cached element from data and metadata.
       (FreeBSD) evict the data type specified by caller, which is the
       existing behavior.
  
  Note that because of our splitted locks (implemented in r205231 to improve
  scalability by reducing lock contention), implementing the fifth Illumos
  behavior will not be cheap, so for now just implement the 1-4 and fall back
  to current behavior for 5.
  
  Illumos issue:
      5368 ARC should cache more metadata
  
  MFC after:	2 months (assuming we didn't found better solution)

Modified:
  head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/arc.c
Directory Properties:
  head/sys/cddl/contrib/opensolaris/   (props changed)

Modified: head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/arc.c
==============================================================================
--- head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/arc.c	Sun Dec 14 22:41:08 2014	(r275779)
+++ head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/arc.c	Mon Dec 15 04:51:36 2014	(r275780)
@@ -197,6 +197,7 @@ static boolean_t arc_warm;
 uint64_t zfs_arc_max;
 uint64_t zfs_arc_min;
 uint64_t zfs_arc_meta_limit = 0;
+uint64_t zfs_arc_meta_min = 0;
 int zfs_arc_grow_retry = 0;
 int zfs_arc_shrink_shift = 0;
 int zfs_arc_p_min_shift = 0;
@@ -218,6 +219,7 @@ SYSINIT(arc_free_target_init, SI_SUB_KTH
     arc_free_target_init, NULL);
 
 TUNABLE_QUAD("vfs.zfs.arc_meta_limit", &zfs_arc_meta_limit);
+TUNABLE_QUAD("vfs.zfs.arc_meta_min", &zfs_arc_meta_min);
 TUNABLE_INT("vfs.zfs.arc_shrink_shift", &zfs_arc_shrink_shift);
 SYSCTL_DECL(_vfs_zfs);
 SYSCTL_UQUAD(_vfs_zfs, OID_AUTO, arc_max, CTLFLAG_RDTUN, &zfs_arc_max, 0,
@@ -426,6 +428,7 @@ typedef struct arc_stats {
 	kstat_named_t arcstat_meta_used;
 	kstat_named_t arcstat_meta_limit;
 	kstat_named_t arcstat_meta_max;
+	kstat_named_t arcstat_meta_min;
 } arc_stats_t;
 
 static arc_stats_t arc_stats = {
@@ -506,7 +509,8 @@ static arc_stats_t arc_stats = {
 	{ "duplicate_reads",		KSTAT_DATA_UINT64 },
 	{ "arc_meta_used",		KSTAT_DATA_UINT64 },
 	{ "arc_meta_limit",		KSTAT_DATA_UINT64 },
-	{ "arc_meta_max",		KSTAT_DATA_UINT64 }
+	{ "arc_meta_max",		KSTAT_DATA_UINT64 },
+	{ "arc_meta_min",		KSTAT_DATA_UINT64 }
 };
 
 #define	ARCSTAT(stat)	(arc_stats.stat.value.ui64)
@@ -569,6 +573,7 @@ static arc_state_t	*arc_l2c_only;
 #define	arc_c_min	ARCSTAT(arcstat_c_min)	/* min target cache size */
 #define	arc_c_max	ARCSTAT(arcstat_c_max)	/* max target cache size */
 #define	arc_meta_limit	ARCSTAT(arcstat_meta_limit) /* max size for metadata */
+#define	arc_meta_min	ARCSTAT(arcstat_meta_min) /* min size for metadata */
 #define	arc_meta_used	ARCSTAT(arcstat_meta_used) /* size of metadata */
 #define	arc_meta_max	ARCSTAT(arcstat_meta_max) /* max size of metadata */
 
@@ -2057,6 +2062,49 @@ arc_evict(arc_state_t *state, uint64_t s
 
 	evicted_state = (state == arc_mru) ? arc_mru_ghost : arc_mfu_ghost;
 
+	/*
+	 * Decide which "type" (data vs metadata) to recycle from.
+	 *
+	 * If we are over the metadata limit, recycle from metadata.
+	 * If we are under the metadata minimum, recycle from data.
+	 * Otherwise, recycle from whichever type has the oldest (least
+	 * recently accessed) header.  This is not yet implemented.
+	 */
+	if (recycle) {
+		arc_buf_contents_t realtype;
+		if (state->arcs_lsize[ARC_BUFC_DATA] == 0) {
+			realtype = ARC_BUFC_METADATA;
+		} else if (state->arcs_lsize[ARC_BUFC_METADATA] == 0) {
+			realtype = ARC_BUFC_DATA;
+		} else if (arc_meta_used >= arc_meta_limit) {
+			realtype = ARC_BUFC_METADATA;
+		} else if (arc_meta_used <= arc_meta_min) {
+			realtype = ARC_BUFC_DATA;
+		} else {
+#ifdef illumos
+			if (data_hdr->b_arc_access <
+			    metadata_hdr->b_arc_access) {
+				realtype = ARC_BUFC_DATA;
+			} else {
+				realtype = ARC_BUFC_METADATA;
+			}
+#else
+			/* TODO */
+			realtype = type;
+#endif
+		}
+		if (realtype != type) {
+			/*
+			 * If we want to evict from a different list,
+			 * we can not recycle, because DATA vs METADATA
+			 * buffers are segregated into different kmem
+			 * caches (and vmem arenas).
+			 */
+			type = realtype;
+			recycle = B_FALSE;
+		}
+	}
+
 	if (type == ARC_BUFC_METADATA) {
 		offset = 0;
 		list_count = ARC_BUFC_NUMMETADATALISTS;
@@ -4198,6 +4246,12 @@ arc_init(void)
 	if (arc_c_min < arc_meta_limit / 2 && zfs_arc_min == 0)
 		arc_c_min = arc_meta_limit / 2;
 
+	if (zfs_arc_meta_min > 0) {
+		arc_meta_min = zfs_arc_meta_min;
+	} else {
+		arc_meta_min = arc_c_min / 2;
+	}
+
 	if (zfs_arc_grow_retry > 0)
 		arc_grow_retry = zfs_arc_grow_retry;