Enable offlining of log devices.

OpenSolaris revision and Bug IDs:

9701:cc5b64682e64
6803605	should be able to offline log devices
6726045	vdev_deflate_ratio is not set when offlining a log device
6599442	zpool import has faults in the display

Approved by:	delphij (mentor)
Obtained from:	OpenSolaris (Bug ID 6803605, 6726045, 6599442)
MFC after:	3 weeks

6 files changed, 147 insertions, 75 deletions

diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dsl_scrub.c b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dsl_scrub.c
index d11f106..50cc069 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dsl_scrub.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dsl_scrub.c
@@ -351,7 +351,7 @@ traverse_zil_block(zilog_t *zilog, blkptr_t *bp, void *arg, uint64_t claim_txg)
 		return;
 
 	/*
-	 * One block ("stumpy") can be allocated a long time ago; we
+	 * One block ("stubby") can be allocated a long time ago; we
 	 * want to visit that one because it has been allocated
 	 * (on-disk) even if it hasn't been claimed (even though for
 	 * plain scrub there's nothing to do to it).
diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa.c b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa.c
index da030f1..32caec6 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa.c
@@ -1110,6 +1110,33 @@ spa_check_removed(vdev_t *vd)
 }
 
 /*
+ * Load the slog device state from the config object since it's possible
+ * that the label does not contain the most up-to-date information.
+ */
+void
+spa_load_log_state(spa_t *spa)
+{
+	nvlist_t *nv, *nvroot, **child;
+	uint64_t is_log;
+	uint_t children, c;
+	vdev_t *rvd = spa->spa_root_vdev;
+
+	VERIFY(load_nvlist(spa, spa->spa_config_object, &nv) == 0);
+	VERIFY(nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
+	VERIFY(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
+	    &child, &children) == 0);
+
+	for (c = 0; c < children; c++) {
+		vdev_t *tvd = rvd->vdev_child[c];
+
+		if (nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
+		    &is_log) == 0 && is_log)
+			vdev_load_log_state(tvd, child[c]);
+	}
+	nvlist_free(nv);
+}
+
+/*
  * Check for missing log devices
  */
 int
@@ -1125,13 +1152,7 @@ spa_check_logs(spa_t *spa)
 			return (1);
 		}
 		break;
-
-	case SPA_LOG_CLEAR:
-		(void) dmu_objset_find(spa->spa_name, zil_clear_log_chain, NULL,
-		    DS_FIND_CHILDREN);
-		break;
 	}
-	spa->spa_log_state = SPA_LOG_GOOD;
 	return (0);
 }
 
@@ -1455,6 +1476,8 @@ spa_load(spa_t *spa, nvlist_t *config, spa_load_state_t state, int mosconfig)
 		spa_config_exit(spa, SCL_ALL, FTAG);
 	}
 
+	spa_load_log_state(spa);
+
 	if (spa_check_logs(spa)) {
 		vdev_set_state(rvd, B_TRUE, VDEV_STATE_CANT_OPEN,
 		    VDEV_AUX_BAD_LOG);
@@ -1542,6 +1565,7 @@ spa_load(spa_t *spa, nvlist_t *config, spa_load_state_t state, int mosconfig)
 		    zil_claim, tx, DS_FIND_CHILDREN);
 		dmu_tx_commit(tx);
 
+		spa->spa_log_state = SPA_LOG_GOOD;
 		spa->spa_sync_on = B_TRUE;
 		txg_sync_start(spa->spa_dsl_pool);
 
diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/vdev_impl.h b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/vdev_impl.h
index 1406d15..93e4102 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/vdev_impl.h
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/vdev_impl.h
@@ -259,6 +259,7 @@ extern void vdev_remove_parent(vdev_t *cvd);
 /*
  * vdev sync load and sync
  */
+extern void vdev_load_log_state(vdev_t *vd, nvlist_t *nv);
 extern void vdev_load(vdev_t *vd);
 extern void vdev_sync(vdev_t *vd, uint64_t txg);
 extern void vdev_sync_done(vdev_t *vd, uint64_t txg);
diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zil.h b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zil.h
index e992f6a..efbf65e 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zil.h
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zil.h
@@ -366,9 +366,9 @@ extern uint64_t zil_itx_assign(zilog_t *zilog, itx_t *itx, dmu_tx_t *tx);
 
 extern void	zil_commit(zilog_t *zilog, uint64_t seq, uint64_t oid);
 
+extern int	zil_vdev_offline(char *osname, void *txarg);
 extern int	zil_claim(char *osname, void *txarg);
 extern int	zil_check_log_chain(char *osname, void *txarg);
-extern int	zil_clear_log_chain(char *osname, void *txarg);
 extern void	zil_sync(zilog_t *zilog, dmu_tx_t *tx);
 extern void	zil_clean(zilog_t *zilog);
 extern int	zil_is_committed(zilog_t *zilog);
diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev.c b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev.c
index 140deed..82006fc 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev.c
@@ -39,6 +39,7 @@
 #include <sys/zap.h>
 #include <sys/fs/zfs.h>
 #include <sys/arc.h>
+#include <sys/zil.h>
 
 SYSCTL_DECL(_vfs_zfs);
 SYSCTL_NODE(_vfs_zfs, OID_AUTO, vdev, CTLFLAG_RW, 0, "ZFS VDEV");
@@ -765,6 +766,15 @@ vdev_metaslab_init(vdev_t *vd, uint64_t txg)
 	if (vd->vdev_ms_shift == 0)	/* not being allocated from yet */
 		return (0);
 
+	/*
+	 * Compute the raidz-deflation ratio.  Note, we hard-code
+	 * in 128k (1 << 17) because it is the current "typical" blocksize.
+	 * Even if SPA_MAXBLOCKSIZE changes, this algorithm must never change,
+	 * or we will inconsistently account for existing bp's.
+	 */
+	vd->vdev_deflate_ratio = (1 << 17) /
+	    (vdev_psize_to_asize(vd, 1 << 17) >> SPA_MINBLOCKSHIFT);
+
 	ASSERT(oldc <= newc);
 
 	if (vd->vdev_islog)
@@ -998,6 +1008,8 @@ vdev_open(vdev_t *vd)
 	    vd->vdev_state == VDEV_STATE_OFFLINE);
 
 	vd->vdev_stat.vs_aux = VDEV_AUX_NONE;
+	vd->vdev_cant_read = B_FALSE;
+	vd->vdev_cant_write = B_FALSE;
 
 	if (!vd->vdev_removed && vd->vdev_faulted) {
 		ASSERT(vd->vdev_children == 0);
@@ -1113,18 +1125,6 @@ vdev_open(vdev_t *vd)
 	}
 
 	/*
-	 * If this is a top-level vdev, compute the raidz-deflation
-	 * ratio.  Note, we hard-code in 128k (1<<17) because it is the
-	 * current "typical" blocksize.  Even if SPA_MAXBLOCKSIZE
-	 * changes, this algorithm must never change, or we will
-	 * inconsistently account for existing bp's.
-	 */
-	if (vd->vdev_top == vd) {
-		vd->vdev_deflate_ratio = (1<<17) /
-		    (vdev_psize_to_asize(vd, 1<<17) >> SPA_MINBLOCKSHIFT);
-	}
-
-	/*
 	 * If a leaf vdev has a DTL, and seems healthy, then kick off a
 	 * resilver.  But don't do this if we are doing a reopen for a scrub,
 	 * since this would just restart the scrub we are already doing.
@@ -1937,7 +1937,8 @@ vdev_online(spa_t *spa, uint64_t guid, uint64_t flags, vdev_state_t *newstate)
 int
 vdev_offline(spa_t *spa, uint64_t guid, uint64_t flags)
 {
-	vdev_t *vd;
+	vdev_t *vd, *tvd;
+	int error;
 
 	spa_vdev_state_enter(spa);
 
@@ -1947,34 +1948,58 @@ vdev_offline(spa_t *spa, uint64_t guid, uint64_t flags)
 	if (!vd->vdev_ops->vdev_op_leaf)
 		return (spa_vdev_state_exit(spa, NULL, ENOTSUP));
 
+	tvd = vd->vdev_top;
+
 	/*
 	 * If the device isn't already offline, try to offline it.
 	 */
 	if (!vd->vdev_offline) {
 		/*
 		 * If this device has the only valid copy of some data,
-		 * don't allow it to be offlined.
+		 * don't allow it to be offlined. Log devices are always
+		 * expendable.
 		 */
-		if (vd->vdev_aux == NULL && vdev_dtl_required(vd))
+		if (!tvd->vdev_islog && vd->vdev_aux == NULL &&
+		    vdev_dtl_required(vd))
 			return (spa_vdev_state_exit(spa, NULL, EBUSY));
 
 		/*
 		 * Offline this device and reopen its top-level vdev.
-		 * If this action results in the top-level vdev becoming
-		 * unusable, undo it and fail the request.
+		 * If the top-level vdev is a log device then just offline
+		 * it. Otherwise, if this action results in the top-level
+		 * vdev becoming unusable, undo it and fail the request.
 		 */
 		vd->vdev_offline = B_TRUE;
-		vdev_reopen(vd->vdev_top);
-		if (vd->vdev_aux == NULL && vdev_is_dead(vd->vdev_top)) {
+		vdev_reopen(tvd);
+
+		if (!tvd->vdev_islog && vd->vdev_aux == NULL &&
+		    vdev_is_dead(tvd)) {
 			vd->vdev_offline = B_FALSE;
-			vdev_reopen(vd->vdev_top);
+			vdev_reopen(tvd);
 			return (spa_vdev_state_exit(spa, NULL, EBUSY));
 		}
 	}
 
 	vd->vdev_tmpoffline = !!(flags & ZFS_OFFLINE_TEMPORARY);
 
-	return (spa_vdev_state_exit(spa, vd, 0));
+	if (!tvd->vdev_islog || !vdev_is_dead(tvd))
+		return (spa_vdev_state_exit(spa, vd, 0));
+
+	(void) spa_vdev_state_exit(spa, vd, 0);
+
+	error = dmu_objset_find(spa_name(spa), zil_vdev_offline,
+	    NULL, DS_FIND_CHILDREN);
+	if (error) {
+		(void) vdev_online(spa, guid, 0, NULL);
+		return (error);
+	}
+	/*
+	 * If we successfully offlined the log device then we need to
+	 * sync out the current txg so that the "stubby" block can be
+	 * removed by zil_sync().
+	 */
+	txg_wait_synced(spa->spa_dsl_pool, 0);
+	return (0);
 }
 
 /*
@@ -2279,6 +2304,7 @@ vdev_space_update(vdev_t *vd, int64_t space_delta, int64_t alloc_delta,
 	 * childrens', thus not accurate enough for us.
 	 */
 	ASSERT((dspace_delta & (SPA_MINBLOCKSIZE-1)) == 0);
+	ASSERT(vd->vdev_deflate_ratio != 0 || vd->vdev_isl2cache);
 	dspace_delta = (dspace_delta >> SPA_MINBLOCKSHIFT) *
 	    vd->vdev_deflate_ratio;
 
@@ -2631,11 +2657,7 @@ vdev_set_state(vdev_t *vd, boolean_t isopen, vdev_state_t state, vdev_aux_t aux)
 boolean_t
 vdev_is_bootable(vdev_t *vd)
 {
-#ifdef __FreeBSD_version
-	return (B_TRUE);
-#else
-	int c;
-
+#ifdef sun
 	if (!vd->vdev_ops->vdev_op_leaf) {
 		char *vdev_type = vd->vdev_ops->vdev_op_type;
 
@@ -2654,6 +2676,35 @@ vdev_is_bootable(vdev_t *vd)
 		if (!vdev_is_bootable(vd->vdev_child[c]))
 			return (B_FALSE);
 	}
+#endif	/* sun */
 	return (B_TRUE);
-#endif
+}
+
+void
+vdev_load_log_state(vdev_t *vd, nvlist_t *nv)
+{
+	uint_t c, children;
+	nvlist_t **child;
+	uint64_t val;
+	spa_t *spa = vd->vdev_spa;
+
+	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
+	    &child, &children) == 0) {
+		for (c = 0; c < children; c++)
+			vdev_load_log_state(vd->vdev_child[c], child[c]);
+	}
+
+	if (vd->vdev_ops->vdev_op_leaf && nvlist_lookup_uint64(nv,
+	    ZPOOL_CONFIG_OFFLINE, &val) == 0 && val) {
+
+		/*
+		 * It would be nice to call vdev_offline()
+		 * directly but the pool isn't fully loaded and
+		 * the txg threads have not been started yet.
+		 */
+		spa_config_enter(spa, SCL_STATE_ALL, FTAG, RW_WRITER);
+		vd->vdev_offline = val;
+		vdev_reopen(vd->vdev_top);
+		spa_config_exit(spa, SCL_STATE_ALL, FTAG);
+	}
 }
diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zil.c b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zil.c
index 7839713..1dd8ea0 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zil.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zil.c
@@ -25,6 +25,7 @@
 
 #include <sys/zfs_context.h>
 #include <sys/spa.h>
+#include <sys/spa_impl.h>
 #include <sys/dmu.h>
 #include <sys/zap.h>
 #include <sys/arc.h>
@@ -515,6 +516,13 @@ zil_claim(char *osname, void *txarg)
 	zilog = dmu_objset_zil(os);
 	zh = zil_header_in_syncing_context(zilog);
 
+	if (zilog->zl_spa->spa_log_state == SPA_LOG_CLEAR) {
+		if (!BP_IS_HOLE(&zh->zh_log))
+			zio_free_blk(zilog->zl_spa, &zh->zh_log, first_txg);
+		BP_ZERO(&zh->zh_log);
+		dsl_dataset_dirty(dmu_objset_ds(os), tx);
+	}
+
 	/*
 	 * Record here whether the zil has any records to replay.
 	 * If the header block pointer is null or the block points
@@ -527,8 +535,10 @@ zil_claim(char *osname, void *txarg)
 	 * Note, the intent log can be empty but still need the
 	 * stubby to be claimed.
 	 */
-	if (!zil_empty(zilog))
+	if (!zil_empty(zilog)) {
 		zh->zh_flags |= ZIL_REPLAY_NEEDED;
+		dsl_dataset_dirty(dmu_objset_ds(os), tx);
+	}
 
 	/*
 	 * Claim all log blocks if we haven't already done so, and remember
@@ -597,36 +607,6 @@ zil_check_log_chain(char *osname, void *txarg)
 	return (error);
 }
 
-/*
- * Clear a log chain
- */
-/* ARGSUSED */
-int
-zil_clear_log_chain(char *osname, void *txarg)
-{
-	zilog_t *zilog;
-	zil_header_t *zh;
-	objset_t *os;
-	dmu_tx_t *tx;
-	int error;
-
-	error = dmu_objset_open(osname, DMU_OST_ANY, DS_MODE_USER, &os);
-	if (error) {
-		cmn_err(CE_WARN, "can't open objset for %s", osname);
-		return (0);
-	}
-
-	zilog = dmu_objset_zil(os);
-	tx = dmu_tx_create(zilog->zl_os);
-	(void) dmu_tx_assign(tx, TXG_WAIT);
-	zh = zil_header_in_syncing_context(zilog);
-	BP_ZERO(&zh->zh_log);
-	dsl_dataset_dirty(dmu_objset_ds(os), tx);
-	dmu_tx_commit(tx);
-	dmu_objset_close(os);
-	return (0);
-}
-
 static int
 zil_vdev_compare(const void *x1, const void *x2)
 {
@@ -771,9 +751,9 @@ zil_lwb_write_init(zilog_t *zilog, lwb_t *lwb)
 	}
 	if (lwb->lwb_zio == NULL) {
 		lwb->lwb_zio = zio_rewrite(zilog->zl_root_zio, zilog->zl_spa,
-		    0, &lwb->lwb_blk, lwb->lwb_buf,
-		    lwb->lwb_sz, zil_lwb_write_done, lwb,
-		    ZIO_PRIORITY_LOG_WRITE, ZIO_FLAG_CANFAIL, &zb);
+		    0, &lwb->lwb_blk, lwb->lwb_buf, lwb->lwb_sz,
+		    zil_lwb_write_done, lwb, ZIO_PRIORITY_LOG_WRITE,
+		    ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE, &zb);
 	}
 }
 
@@ -1270,12 +1250,7 @@ zil_sync(zilog_t *zilog, dmu_tx_t *tx)
 		}
 	}
 
-	for (;;) {
-		lwb = list_head(&zilog->zl_lwb_list);
-		if (lwb == NULL) {
-			mutex_exit(&zilog->zl_lock);
-			return;
-		}
+	while ((lwb = list_head(&zilog->zl_lwb_list)) != NULL) {
 		zh->zh_log = lwb->lwb_blk;
 		if (lwb->lwb_buf != NULL || lwb->lwb_max_txg > txg)
 			break;
@@ -1692,3 +1667,24 @@ out:
 	mutex_exit(&zilog->zl_lock);
 	return (ret);
 }
+
+/* ARGSUSED */
+int
+zil_vdev_offline(char *osname, void *arg)
+{
+	objset_t *os;
+	zilog_t *zilog;
+	int error;
+
+	error = dmu_objset_open(osname, DMU_OST_ANY, DS_MODE_USER, &os);
+	if (error)
+		return (error);
+
+	zilog = dmu_objset_zil(os);
+	if (zil_suspend(zilog) != 0)
+		error = EEXIST;
+	else
+		zil_resume(zilog);
+	dmu_objset_close(os);
+	return (error);
+}