OpenZFS 6736 - ZFS per-vdev ZAPs

6736 ZFS per-vdev ZAPs
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: John Kennedy <john.kennedy@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Don Brady <don.brady@intel.com>
Reviewed by: Dan McDonald <danmcd@omniti.com>

References:
  https://www.illumos.org/issues/6736
  https://github.com/openzfs/openzfs/commit/215198a

Ported-by: Don Brady <don.brady@intel.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #4515

include/sys/dmu.h

@@ -315,6 +315,7 @@ typedef struct dmu_buf {
 #define	DMU_POOL_FREE_BPOBJ		"free_bpobj"
 #define	DMU_POOL_BPTREE_OBJ		"bptree_obj"
 #define	DMU_POOL_EMPTY_BPOBJ		"empty_bpobj"
+#define	DMU_POOL_VDEV_ZAP_MAP		"com.delphix:vdev_zap_map"
 
 /*
  * Allocate an object from this objset.  The range of object numbers

include/sys/fs/zfs.h

@@ -566,6 +566,9 @@ typedef struct zpool_rewind_policy {
 #define	ZPOOL_CONFIG_FEATURES_FOR_READ	"features_for_read"
 #define	ZPOOL_CONFIG_FEATURE_STATS	"feature_stats"	/* not stored on disk */
 #define	ZPOOL_CONFIG_ERRATA		"errata"	/* not stored on disk */
+#define	ZPOOL_CONFIG_VDEV_TOP_ZAP	"com.delphix:vdev_zap_top"
+#define	ZPOOL_CONFIG_VDEV_LEAF_ZAP	"com.delphix:vdev_zap_leaf"
+#define	ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS	"com.delphix:has_per_vdev_zaps"
 /*
  * The persistent vdev state is stored as separate values rather than a single
  * 'vdev_state' entry.  This is because a device can be in multiple states, such

include/sys/spa_impl.h

@@ -116,6 +116,12 @@ typedef struct spa_taskqs {
 	taskq_t **stqs_taskq;
 } spa_taskqs_t;
 
+typedef enum spa_all_vdev_zap_action {
+	AVZ_ACTION_NONE = 0,
+	AVZ_ACTION_DESTROY,	/* Destroy all per-vdev ZAPs and the AVZ. */
+	AVZ_ACTION_REBUILD	/* Populate the new AVZ, see spa_avz_rebuild */
+} spa_avz_action_t;
+
 struct spa {
 	/*
 	 * Fields protected by spa_namespace_lock.
@@ -251,6 +257,8 @@ struct spa {
 	uint64_t	spa_deadman_calls;	/* number of deadman calls */
 	hrtime_t	spa_sync_starttime;	/* starting time of spa_sync */
 	uint64_t	spa_deadman_synctime;	/* deadman expiration timer */
+	uint64_t	spa_all_vdev_zaps;	/* ZAP of per-vd ZAP obj #s */
+	spa_avz_action_t	spa_avz_action;	/* destroy/rebuild AVZ? */
 	uint64_t	spa_errata;		/* errata issues detected */
 	spa_stats_t	spa_stats;		/* assorted spa statistics */
 	hrtime_t	spa_ccw_fail_time;	/* Conf cache write fail time */

include/sys/vdev.h

@@ -70,6 +70,10 @@ extern void vdev_dtl_reassess(vdev_t *vd, uint64_t txg, uint64_t scrub_txg,
 extern boolean_t vdev_dtl_required(vdev_t *vd);
 extern boolean_t vdev_resilver_needed(vdev_t *vd,
     uint64_t *minp, uint64_t *maxp);
+extern void vdev_destroy_unlink_zap(vdev_t *vd, uint64_t zapobj,
+    dmu_tx_t *tx);
+extern uint64_t vdev_create_link_zap(vdev_t *vd, dmu_tx_t *tx);
+extern void vdev_construct_zaps(vdev_t *vd, dmu_tx_t *tx);
 
 extern void vdev_hold(vdev_t *);
 extern void vdev_rele(vdev_t *);
@@ -133,7 +137,8 @@ extern void vdev_state_clean(vdev_t *vd);
 typedef enum vdev_config_flag {
 	VDEV_CONFIG_SPARE = 1 << 0,
 	VDEV_CONFIG_L2CACHE = 1 << 1,
-	VDEV_CONFIG_REMOVING = 1 << 2
+	VDEV_CONFIG_REMOVING = 1 << 2,
+	VDEV_CONFIG_MOS = 1 << 3
 } vdev_config_flag_t;
 
 extern void vdev_top_config_generate(spa_t *spa, nvlist_t *config);

include/sys/vdev_impl.h

@@ -177,6 +177,7 @@ struct vdev {
 	uint64_t	vdev_islog;	/* is an intent log device	*/
 	uint64_t	vdev_removing;	/* device is being removed?	*/
 	boolean_t	vdev_ishole;	/* is a hole in the namespace 	*/
+	uint64_t	vdev_top_zap;
 
 	/*
 	 * Leaf vdev state.
@@ -215,6 +216,7 @@ struct vdev {
 	spa_aux_vdev_t	*vdev_aux;	/* for l2cache and spares vdevs	*/
 	zio_t		*vdev_probe_zio; /* root of current probe	*/
 	vdev_aux_t	vdev_label_aux;	/* on-disk aux state		*/
+	uint64_t	vdev_leaf_zap;
 
 	/*
 	 * For DTrace to work in userland (libzpool) context, these fields must