Improved zpool status output, list all the dataset affected by errlog.

include/sys/spa.h

@@ -1175,11 +1175,14 @@ extern void zfs_post_remove(spa_t *spa, vdev_t *vd);
 extern void zfs_post_state_change(spa_t *spa, vdev_t *vd, uint64_t laststate);
 extern void zfs_post_autoreplace(spa_t *spa, vdev_t *vd);
 extern uint64_t spa_get_errlog_size(spa_t *spa);
-extern int spa_get_errlog(spa_t *spa, void *uaddr, size_t *count);
+extern int spa_get_errlog(spa_t *spa, void *uaddr, uint64_t *count);
 extern void spa_errlog_rotate(spa_t *spa);
 extern void spa_errlog_drain(spa_t *spa);
 extern void spa_errlog_sync(spa_t *spa, uint64_t txg);
 extern void spa_get_errlists(spa_t *spa, avl_tree_t *last, avl_tree_t *scrub);
+extern void spa_delete_dataset_errlog(spa_t *spa, uint64_t ds, dmu_tx_t *tx);
+extern void spa_swap_errlog(spa_t *spa, uint64_t new_head_ds,
+    uint64_t old_head_ds, dmu_tx_t *tx);
 
 /* vdev cache */
 extern void vdev_cache_stat_init(void);

include/sys/zio.h

@@ -305,6 +305,13 @@ struct zbookmark_phys {
 	uint64_t	zb_blkid;
 };
 
+typedef struct zbookmark_err_phys {
+	uint64_t	zb_object;
+	int64_t		zb_level;
+	uint64_t	zb_blkid;
+	uint64_t	zb_birth;
+} zbookmark_err_phys_t;
+
 #define	SET_BOOKMARK(zb, objset, object, level, blkid)  \
 {                                                       \
 	(zb)->zb_objset = objset;                       \

include/zfeature_common.h

@@ -72,6 +72,7 @@ typedef enum spa_feature {
 	SPA_FEATURE_BOOKMARK_WRITTEN,
 	SPA_FEATURE_LOG_SPACEMAP,
 	SPA_FEATURE_LIVELIST,
+	SPA_FEATURE_HEAD_ERRLOG,
 	SPA_FEATURES
 } spa_feature_t;
 

man/man5/zpool-features.5

@@ -925,5 +925,27 @@ is used to checkpoint the pool.
 The feature will only return back to being \fBenabled\fR when the pool
 is rewound or the checkpoint has been discarded.
 
+
+.sp
+.ne 2
+.na
+\fBhead_errlog\fR
+.ad
+.RS 4n
+.TS
+l l .
+GUID	com.delphix:head_errlog
+READ\-ONLY COMPATIBLE	no
+DEPENDENCIES	none
+.TE
+
+This feature enables the upgraded version of errlog. This required on disk
+error format change. Now, errorlog of each head dataset is stored
+seprately in zap object and primarily keyed by this head Id.
+By this feature every dataset affected by an error block is listed under
+\fBzpool status\fR command.
+
+The feature can only be set at the time of pool creation.
+
 .SH "SEE ALSO"
 zpool(8)

module/zcommon/zfeature_common.c

@@ -553,6 +553,13 @@ zpool_feature_init(void)
 	    "com.datto:resilver_defer", "resilver_defer",
 	    "Support for defering new resilvers when one is already running.",
 	    ZFEATURE_FLAG_READONLY_COMPAT, ZFEATURE_TYPE_BOOLEAN, NULL);
+
+	{
+	zfeature_register(SPA_FEATURE_HEAD_ERRLOG,
+	    "com.delphix:head_errlog", "head_errlog",
+	    "Improved zpool status command output.",
+	    ZFEATURE_FLAG_ACTIVATE_ON_ENABLE, ZFEATURE_TYPE_BOOLEAN, NULL);
+	}
 }
 
 #if defined(_KERNEL)

module/zfs/dsl_dataset.c

@@ -3726,6 +3726,15 @@ dsl_dataset_promote_sync(void *arg, dmu_tx_t *tx)
 
 	dsl_dir_rele(odd, FTAG);
 	promote_rele(ddpa, FTAG);
+
+	/*
+	 * Transfer common error blocks from old head to new head.
+	 */
+	if (spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_HEAD_ERRLOG)) {
+		uint64_t old_head = origin_head->ds_object;
+		uint64_t new_head = hds->ds_object;
+		spa_swap_errlog(dp->dp_spa, new_head, old_head, tx);
+	}
 }
 
 /*

module/zfs/dsl_destroy.c

@@ -1139,6 +1139,9 @@ dsl_destroy_head_sync_impl(dsl_dataset_t *ds, dmu_tx_t *tx)
 		dsl_destroy_snapshot_sync_impl(prev, B_FALSE, tx);
 		dsl_dataset_rele(prev, FTAG);
 	}
+	/* Delete errlog. */
+	if (spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_HEAD_ERRLOG))
+		spa_delete_dataset_errlog(dp->dp_spa, ds->ds_object, tx);
 }
 
 void