Fast Dedup: Dedup Quota

cmd/ztest.c

@@ -8495,17 +8495,24 @@ print_time(hrtime_t t, char *timebuf)
 }
 
 static nvlist_t *
-make_random_props(void)
+make_random_pool_props(void)
 {
 	nvlist_t *props;
 
 	props = fnvlist_alloc();
 
-	if (ztest_random(2) == 0)
-		return (props);
+	/* Twenty percent of the time enable ZPOOL_PROP_DEDUP_TABLE_QUOTA */
+	if (ztest_random(5) == 0) {
+		fnvlist_add_uint64(props,
+		    zpool_prop_to_name(ZPOOL_PROP_DEDUP_TABLE_QUOTA),
+		    2 * 1024 * 1024);
+	}
 
-	fnvlist_add_uint64(props,
-	    zpool_prop_to_name(ZPOOL_PROP_AUTOREPLACE), 1);
+	/* Fifty percent of the time enable ZPOOL_PROP_AUTOREPLACE */
+	if (ztest_random(2) == 0) {
+		fnvlist_add_uint64(props,
+		    zpool_prop_to_name(ZPOOL_PROP_AUTOREPLACE), 1);
+	}
 
 	return (props);
 }
@@ -8537,7 +8544,7 @@ ztest_init(ztest_shared_t *zs)
 	zs->zs_mirrors = ztest_opts.zo_mirrors;
 	nvroot = make_vdev_root(NULL, NULL, NULL, ztest_opts.zo_vdev_size, 0,
 	    NULL, ztest_opts.zo_raid_children, zs->zs_mirrors, 1);
-	props = make_random_props();
+	props = make_random_pool_props();
 
 	/*
 	 * We don't expect the pool to suspend unless maxfaults == 0,

include/sys/ddt.h

@@ -151,7 +151,8 @@ enum ddt_phys_type {
  */
 
 /* State flags for dde_flags */
-#define	DDE_FLAG_LOADED (1 << 0)	/* entry ready for use */
+#define	DDE_FLAG_LOADED		(1 << 0)	/* entry ready for use */
+#define	DDE_FLAG_OVERQUOTA	(1 << 1)	/* entry unusable, no space */
 
 typedef struct {
 	/* key must be first for ddt_key_compare */
@@ -170,6 +171,7 @@ typedef struct {
 
 	uint8_t		dde_flags;	/* load state flags */
 	kcondvar_t	dde_cv;		/* signaled when load completes */
+	uint64_t	dde_waiters;	/* count of waiters on dde_cv */
 
 	avl_node_t	dde_node;	/* ddt_tree node */
 } ddt_entry_t;
@@ -228,6 +230,7 @@ extern void ddt_histogram_add(ddt_histogram_t *dst, const ddt_histogram_t *src);
 extern void ddt_histogram_stat(ddt_stat_t *dds, const ddt_histogram_t *ddh);
 extern boolean_t ddt_histogram_empty(const ddt_histogram_t *ddh);
 extern void ddt_get_dedup_object_stats(spa_t *spa, ddt_object_t *ddo);
+extern uint64_t ddt_get_ddt_dsize(spa_t *spa);
 extern void ddt_get_dedup_histogram(spa_t *spa, ddt_histogram_t *ddh);
 extern void ddt_get_dedup_stats(spa_t *spa, ddt_stat_t *dds_total);
 

include/sys/fs/zfs.h

@@ -258,6 +258,8 @@ typedef enum {
 	ZPOOL_PROP_BCLONEUSED,
 	ZPOOL_PROP_BCLONESAVED,
 	ZPOOL_PROP_BCLONERATIO,
+	ZPOOL_PROP_DEDUP_TABLE_SIZE,
+	ZPOOL_PROP_DEDUP_TABLE_QUOTA,
 	ZPOOL_NUM_PROPS
 } zpool_prop_t;
 

include/sys/spa.h

@@ -1051,6 +1051,7 @@ extern metaslab_class_t *spa_special_class(spa_t *spa);
 extern metaslab_class_t *spa_dedup_class(spa_t *spa);
 extern metaslab_class_t *spa_preferred_class(spa_t *spa, uint64_t size,
     dmu_object_type_t objtype, uint_t level, uint_t special_smallblk);
+extern boolean_t spa_special_has_ddt(spa_t *spa);
 
 extern void spa_evicting_os_register(spa_t *, objset_t *os);
 extern void spa_evicting_os_deregister(spa_t *, objset_t *os);

include/sys/spa_impl.h

@@ -465,6 +465,9 @@ struct spa {
 	boolean_t	spa_waiters_cancel;	/* waiters should return */
 
 	char		*spa_compatibility;	/* compatibility file(s) */
+	uint64_t	spa_dedup_table_quota;	/* property DDT maximum size */
+	uint64_t	spa_dedup_dsize;	/* cached on-disk size of DDT */
+	uint64_t	spa_dedup_class_full_txg; /* txg dedup class was full */
 
 	/*
 	 * spa_refcount & spa_config_lock must be the last elements

lib/libzfs/libzfs.abi

@@ -2921,7 +2921,9 @@
       <enumerator name='ZPOOL_PROP_BCLONEUSED' value='33'/>
       <enumerator name='ZPOOL_PROP_BCLONESAVED' value='34'/>
       <enumerator name='ZPOOL_PROP_BCLONERATIO' value='35'/>
-      <enumerator name='ZPOOL_NUM_PROPS' value='36'/>
+      <enumerator name='ZPOOL_PROP_DEDUP_TABLE_SIZE' value='36'/>
+      <enumerator name='ZPOOL_PROP_DEDUP_TABLE_QUOTA' value='37'/>
+      <enumerator name='ZPOOL_NUM_PROPS' value='38'/>
     </enum-decl>
     <typedef-decl name='zpool_prop_t' type-id='af1ba157' id='5d0c23fb'/>
     <typedef-decl name='regoff_t' type-id='95e97e5e' id='54a2a2a8'/>

lib/libzfs/libzfs_pool.c

@@ -332,6 +332,24 @@ zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf,
 		intval = zpool_get_prop_int(zhp, prop, &src);
 
 		switch (prop) {
+		case ZPOOL_PROP_DEDUP_TABLE_QUOTA:
+			/*
+			 * If dedup quota is 0, we translate this into 'none'
+			 * (unless literal is set). And if it is UINT64_MAX
+			 * we translate that as 'automatic' (limit to size of
+			 * the dedicated dedup VDEV.  Otherwise, fall throught
+			 * into the regular number formating.
+			 */
+			if (intval == 0) {
+				(void) strlcpy(buf, literal ? "0" : "none",
+				    len);
+				break;
+			} else if (intval == UINT64_MAX) {
+				(void) strlcpy(buf, "auto", len);
+				break;
+			}
+			zfs_fallthrough;
+
 		case ZPOOL_PROP_SIZE:
 		case ZPOOL_PROP_ALLOCATED:
 		case ZPOOL_PROP_FREE:
@@ -342,6 +360,7 @@ zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf,
 		case ZPOOL_PROP_MAXDNODESIZE:
 		case ZPOOL_PROP_BCLONESAVED:
 		case ZPOOL_PROP_BCLONEUSED:
+		case ZPOOL_PROP_DEDUP_TABLE_SIZE:
 			if (literal)
 				(void) snprintf(buf, len, "%llu",
 				    (u_longlong_t)intval);

lib/libzfs/libzfs_util.c

@@ -1691,6 +1691,16 @@ zprop_parse_value(libzfs_handle_t *hdl, nvpair_t *elem, int prop,
 			    "use 'none' to disable quota/refquota"));
 			goto error;
 		}
+		/*
+		 * Pool dedup table quota; force use of 'none' instead of 0
+		 */
+		if ((type & ZFS_TYPE_POOL) && *ivalp == 0 &&
+		    (!isnone && !isauto) &&
+		    prop == ZPOOL_PROP_DEDUP_TABLE_QUOTA) {
+			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
+			    "use 'none' to disable ddt table quota"));
+			goto error;
+		}
 
 		/*
 		 * Special handling for "*_limit=none". In this case it's not
@@ -1732,6 +1742,10 @@ zprop_parse_value(libzfs_handle_t *hdl, nvpair_t *elem, int prop,
 				}
 				*ivalp = UINT64_MAX;
 				break;
+			case ZPOOL_PROP_DEDUP_TABLE_QUOTA:
+				ASSERT(type & ZFS_TYPE_POOL);
+				*ivalp = UINT64_MAX;
+				break;
 			default:
 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
 				    "'auto' is invalid value for '%s'"),

man/man7/zpoolprops.7

@@ -28,7 +28,7 @@
 .\" Copyright (c) 2021, Colm Buckley <[email protected]>
 .\" Copyright (c) 2023, Klara Inc.
 .\"
-.Dd January 2, 2024
+.Dd January 14, 2024
 .Dt ZPOOLPROPS 7
 .Os
 .
@@ -73,6 +73,8 @@ The amount of storage used by cloned blocks.
 Percentage of pool space used.
 This property can also be referred to by its shortened column name,
 .Sy cap .
+.It Sy dedup_table_size
+Total on-disk size of the deduplication table.
 .It Sy expandsize
 Amount of uninitialized space within the pool or device that can be used to
 increase the total capacity of the pool.
@@ -348,6 +350,27 @@ See
 and
 .Xr zpool-upgrade 8
 for more information on the operation of compatibility feature sets.
+.It Sy dedup_table_quota Ns = Ns Ar number Ns | Ns Sy none Ns | Ns Sy auto
+This property sets a limit on the on-disk size of the pool's dedup table.
+Entries will not be added to the dedup table once this size is reached;
+if a dedup table already exists, and is larger than this size, they
+will not be removed as part of setting this property.
+Existing entries will still have their reference counts updated.
+.Pp
+The actual size limit of the table may be above or below the quota,
+depending on the actual on-disk size of the entries (which may be
+approximated for purposes of calculating the quota).
+That is, setting a quota size of 1M may result in the maximum size being
+slightly below, or slightly above, that value.
+Set to
+.Sy 'none'
+to disable.
+In automatic mode, which is the default, the size of a dedicated dedup vdev
+is used as the quota limit.
+.Pp
+The
+.Sy dedup_table_quota
+property works for both legacy and fast dedup tables.
 .It Sy dedupditto Ns = Ns Ar number
 This property is deprecated and no longer has any effect.
 .It Sy delegation Ns = Ns Sy on Ns | Ns Sy off

module/zcommon/zpool_prop.c

@@ -23,7 +23,7 @@
  * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
  * Copyright (c) 2012, 2018 by Delphix. All rights reserved.
  * Copyright (c) 2021, Colm Buckley <[email protected]>
- * Copyright (c) 2021, Klara Inc.
+ * Copyright (c) 2021, 2023, Klara Inc.
  */
 
 #include <sys/zio.h>
@@ -125,6 +125,9 @@ zpool_prop_init(void)
 	zprop_register_number(ZPOOL_PROP_BCLONERATIO, "bcloneratio", 0,
 	    PROP_READONLY, ZFS_TYPE_POOL, "<1.00x or higher if cloned>",
 	    "BCLONE_RATIO", B_FALSE, sfeatures);
+	zprop_register_number(ZPOOL_PROP_DEDUP_TABLE_SIZE, "dedup_table_size",
+	    0, PROP_READONLY, ZFS_TYPE_POOL, "<size>", "DDTSIZE", B_FALSE,
+	    sfeatures);
 
 	/* default number properties */
 	zprop_register_number(ZPOOL_PROP_VERSION, "version", SPA_VERSION,
@@ -133,6 +136,9 @@ zpool_prop_init(void)
 	zprop_register_number(ZPOOL_PROP_ASHIFT, "ashift", 0, PROP_DEFAULT,
 	    ZFS_TYPE_POOL, "<ashift, 9-16, or 0=default>", "ASHIFT", B_FALSE,
 	    sfeatures);
+	zprop_register_number(ZPOOL_PROP_DEDUP_TABLE_QUOTA, "dedup_table_quota",
+	    UINT64_MAX, PROP_DEFAULT, ZFS_TYPE_POOL, "<size>", "DDTQUOTA",
+	    B_FALSE, sfeatures);
 
 	/* default index (boolean) properties */
 	zprop_register_index(ZPOOL_PROP_DELEGATION, "delegation", 1,