xfs: allow queued AG intents to drain before scrubbingscrub-drain-intents_2021-12-15

Currently, online scrub isn't sufficiently careful about quiescing
allocation groups before checking them.  While scrub does take the AG
header locks, it doesn't serialize against chains of AG update intents
that are being processed concurrently.  If there's a collision,
cross-referencing between data structures (e.g. rmapbt and refcountbt)
can yield false corruption events; if repair is running, this results in
incorrect repairs.

Fix this by adding to the perag structure the count of active intents
and make scrub wait until there aren't any to continue.  This is a
little stupid since transactions can queue intents without taking buffer
locks, but we'll also wait for those transactions.

XXX: should have instead a per-ag rwsem that gets taken as soon as the
AG[IF] are locked and stays held until the transaction commits or moves
on to the next AG?  would we rather have a six lock so that intents can
take an ix lock, and not have to upgrade to x until we actually want to
make changes to that ag?  is that how those even work??

Signed-off-by: Darrick J. Wong <djwong@kernel.org>

25 files changed, 599 insertions, 38 deletions

diff --git a/fs/xfs/Kconfig b/fs/xfs/Kconfig
index 7d08a8e92c92..5ad64e173672 100644
--- a/fs/xfs/Kconfig
+++ b/fs/xfs/Kconfig
@@ -96,6 +96,9 @@ config XFS_RT
 config XFS_LIVE_HOOKS
 	bool
 
+config XFS_DRAIN_INTENTS
+	bool
+
 config XFS_ONLINE_SCRUB
 	bool "XFS online metadata check support"
 	default n
@@ -103,6 +106,7 @@ config XFS_ONLINE_SCRUB
 	depends on TMPFS && SHMEM
 	depends on SRCU
 	select XFS_LIVE_HOOKS
+	select XFS_DRAIN_INTENTS
 	help
 	  If you say Y here you will be able to check metadata on a
 	  mounted XFS filesystem.  This feature is intended to reduce
diff --git a/fs/xfs/libxfs/xfs_ag.c b/fs/xfs/libxfs/xfs_ag.c
index 30ded0376db2..2da6af33c826 100644
--- a/fs/xfs/libxfs/xfs_ag.c
+++ b/fs/xfs/libxfs/xfs_ag.c
@@ -193,6 +193,7 @@ xfs_free_perag(
 		spin_unlock(&mp->m_perag_lock);
 		ASSERT(pag);
 		ASSERT(atomic_read(&pag->pag_ref) == 0);
+		xfs_drain_free(&pag->pag_intents);
 
 		cancel_delayed_work_sync(&pag->pag_blockgc_work);
 		xfs_iunlink_destroy(pag);
@@ -255,6 +256,7 @@ xfs_initialize_perag(
 		spin_lock_init(&pag->pag_state_lock);
 		INIT_DELAYED_WORK(&pag->pag_blockgc_work, xfs_blockgc_worker);
 		INIT_RADIX_TREE(&pag->pag_ici_root, GFP_ATOMIC);
+		xfs_drain_init(&pag->pag_intents);
 		init_waitqueue_head(&pag->pagb_wait);
 		pag->pagb_count = 0;
 		pag->pagb_tree = RB_ROOT;
@@ -283,6 +285,7 @@ xfs_initialize_perag(
 
 out_hash_destroy:
 	xfs_buf_hash_destroy(pag);
+	xfs_drain_free(&pag->pag_intents);
 out_remove_pag:
 	radix_tree_delete(&mp->m_perag_tree, index);
 out_free_pag:
@@ -294,6 +297,7 @@ out_unwind_new_pags:
 		if (!pag)
 			break;
 		xfs_buf_hash_destroy(pag);
+		xfs_drain_free(&pag->pag_intents);
 		xfs_iunlink_destroy(pag);
 		kmem_free(pag);
 	}
diff --git a/fs/xfs/libxfs/xfs_ag.h b/fs/xfs/libxfs/xfs_ag.h
index aae98fda12f6..2edd4c4ff07c 100644
--- a/fs/xfs/libxfs/xfs_ag.h
+++ b/fs/xfs/libxfs/xfs_ag.h
@@ -104,6 +104,15 @@ struct xfs_perag {
 	 * or have some other means to control concurrency.
 	 */
 	struct rhashtable	pagi_unlinked_hash;
+
+	/*
+	 * We use xfs_drain to track the number of deferred log intent items
+	 * that have been queued (but not yet processed) so that waiters (e.g.
+	 * scrub) will not lock resources when other threads are in the middle
+	 * of processing a chain of intent items only to find momentary
+	 * inconsistencies.
+	 */
+	struct xfs_drain	pag_intents;
 #endif /* __KERNEL__ */
 };
 
diff --git a/fs/xfs/libxfs/xfs_defer.c b/fs/xfs/libxfs/xfs_defer.c
index ccb834a82e2e..59f8a2af39d5 100644
--- a/fs/xfs/libxfs/xfs_defer.c
+++ b/fs/xfs/libxfs/xfs_defer.c
@@ -370,7 +370,8 @@ xfs_defer_cancel_list(
 		list_for_each_safe(pwi, n, &dfp->dfp_work) {
 			list_del(pwi);
 			dfp->dfp_count--;
-			ops->cancel_item(pwi);
+			trace_xfs_defer_cancel_item(mp, dfp, pwi);
+			ops->cancel_item(mp, pwi);
 		}
 		ASSERT(dfp->dfp_count == 0);
 		kmem_cache_free(xfs_defer_pending_cache, dfp);
@@ -449,6 +450,7 @@ xfs_defer_finish_one(
 	list_for_each_safe(li, n, &dfp->dfp_work) {
 		list_del(li);
 		dfp->dfp_count--;
+		trace_xfs_defer_finish_item(tp->t_mountp, dfp, li);
 		error = ops->finish_item(tp, dfp->dfp_done, li, &state);
 		if (error == -EAGAIN) {
 			/*
@@ -592,7 +594,7 @@ xfs_defer_add(
 	struct list_head		*li)
 {
 	struct xfs_defer_pending	*dfp = NULL;
-	const struct xfs_defer_op_type	*ops;
+	const struct xfs_defer_op_type	*ops = defer_op_types[type];
 
 	ASSERT(tp->t_flags & XFS_TRANS_PERM_LOG_RES);
 	BUILD_BUG_ON(ARRAY_SIZE(defer_op_types) != XFS_DEFER_OPS_TYPE_MAX);
@@ -605,7 +607,6 @@ xfs_defer_add(
 	if (!list_empty(&tp->t_dfops)) {
 		dfp = list_last_entry(&tp->t_dfops,
 				struct xfs_defer_pending, dfp_list);
-		ops = defer_op_types[dfp->dfp_type];
 		if (dfp->dfp_type != type ||
 		    (ops->max_items && dfp->dfp_count >= ops->max_items))
 			dfp = NULL;
@@ -623,6 +624,8 @@ xfs_defer_add(
 	}
 
 	list_add_tail(li, &dfp->dfp_work);
+	trace_xfs_defer_add_item(tp->t_mountp, dfp, li);
+	ops->add_item(tp->t_mountp, li);
 	dfp->dfp_count++;
 }
 
diff --git a/fs/xfs/libxfs/xfs_defer.h b/fs/xfs/libxfs/xfs_defer.h
index 89237d8fdedd..6c575b7aacfb 100644
--- a/fs/xfs/libxfs/xfs_defer.h
+++ b/fs/xfs/libxfs/xfs_defer.h
@@ -55,7 +55,8 @@ struct xfs_defer_op_type {
 			struct list_head *item, struct xfs_btree_cur **state);
 	void (*finish_cleanup)(struct xfs_trans *tp,
 			struct xfs_btree_cur *state, int error);
-	void (*cancel_item)(struct list_head *item);
+	void (*cancel_item)(struct xfs_mount *mp, struct list_head *item);
+	void (*add_item)(struct xfs_mount *mp, const struct list_head *item);
 	unsigned int		max_items;
 };
 
diff --git a/fs/xfs/scrub/bmap.c b/fs/xfs/scrub/bmap.c
index 9ac01b24c9b0..dc472aa6f548 100644
--- a/fs/xfs/scrub/bmap.c
+++ b/fs/xfs/scrub/bmap.c
@@ -318,7 +318,12 @@ xchk_bmap_rt_iextent_xref(
 	struct xchk_bmap_info	*info,
 	struct xfs_bmbt_irec	*irec)
 {
-	xchk_rt_init(info->sc, &info->sc->sr);
+	int			error;
+
+	error = xchk_rt_init(info->sc, &info->sc->sr);
+	if (!xchk_fblock_process_error(info->sc, info->whichfork,
+			irec->br_startoff, &error))
+		goto out_free;
 
 	xchk_xref_is_used_rt_space(info->sc, irec->br_startblock,
 			irec->br_blockcount);
@@ -336,6 +341,7 @@ xchk_bmap_rt_iextent_xref(
 		break;
 	}
 
+out_free:
 	xchk_rt_btcur_free(&info->sc->sr);
 	xchk_rt_unlock(info->sc, &info->sc->sr);
 }
diff --git a/fs/xfs/scrub/bmap_repair.c b/fs/xfs/scrub/bmap_repair.c
index 2256b77e20fc..ad6f00c4767f 100644
--- a/fs/xfs/scrub/bmap_repair.c
+++ b/fs/xfs/scrub/bmap_repair.c
@@ -339,7 +339,9 @@ xrep_bmap_scan_rt(
 	if (xrep_is_rtmeta_ino(sc, sc->ip->i_ino))
 		return 0;
 
-	xchk_rt_lock(sc, &sc->sr);
+	error = xchk_rt_lock(sc, &sc->sr);
+	if (error)
+		return error;
 	xrep_rt_btcur_init(sc, &sc->sr);
 	error = xfs_rmap_query_all(sc->sr.rmap_cur, xrep_bmap_walk_rtrmap, rb);
 	xchk_rt_btcur_free(&sc->sr);
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index 246d33ac46b9..4987aebca74e 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -484,7 +484,35 @@ want_ag_read_header_failure(
  *
  * The headers should be released by xchk_ag_free, but as a fail safe we attach
  * all the buffers we grab to the scrub transaction so they'll all be freed
- * when we cancel it.  Returns ENOENT if we can't grab the perag structure.
+ * when we cancel it.
+ */
+static inline int
+__xchk_ag_read_headers(
+	struct xfs_scrub	*sc,
+	xfs_agnumber_t		agno,
+	struct xchk_ag		*sa)
+{
+	struct xfs_mount	*mp = sc->mp;
+	int			error;
+
+	error = xfs_ialloc_read_agi(mp, sc->tp, agno, &sa->agi_bp);
+	if (error && want_ag_read_header_failure(sc, XFS_SCRUB_TYPE_AGI))
+		return error;
+
+	error = xfs_alloc_read_agf(mp, sc->tp, agno, 0, &sa->agf_bp);
+	if (error && want_ag_read_header_failure(sc, XFS_SCRUB_TYPE_AGF))
+		return error;
+
+	error = xfs_alloc_read_agfl(mp, sc->tp, agno, &sa->agfl_bp);
+	if (error && want_ag_read_header_failure(sc, XFS_SCRUB_TYPE_AGFL))
+		return error;
+
+	return 0;
+}
+
+/*
+ * Grab all the headers for an AG, and wait until there aren't any pending
+ * intents.  Returns -ENOENT if we can't grab the perag structure.
  */
 int
 xchk_ag_read_headers(
@@ -502,29 +530,72 @@ xchk_ag_read_headers(
 	return xchk_ag_lock(sc);
 }
 
-/* Lock the AG headers. */
+/* Lock the AG headers, waiting for pending intents to drain. */
 int
 xchk_ag_lock(
 	struct xfs_scrub	*sc)
 {
-	struct xfs_mount	*mp = sc->mp;
 	struct xchk_ag		*sa = &sc->sa;
-	xfs_agnumber_t		agno = sa->pag->pag_agno;
-	int			error;
+	int			error = 0;
 
-	error = xfs_ialloc_read_agi(mp, sc->tp, agno, &sa->agi_bp);
-	if (error && want_ag_read_header_failure(sc, XFS_SCRUB_TYPE_AGI))
-		return error;
+	ASSERT(sa->pag != NULL);
+	ASSERT(sa->agi_bp == NULL);
+	ASSERT(sa->agf_bp == NULL);
+	ASSERT(sa->agfl_bp == NULL);
 
-	error = xfs_alloc_read_agf(mp, sc->tp, agno, 0, &sa->agf_bp);
-	if (error && want_ag_read_header_failure(sc, XFS_SCRUB_TYPE_AGF))
-		return error;
+	do {
+		if (xchk_should_terminate(sc, &error))
+			return error;
 
-	error = xfs_alloc_read_agfl(mp, sc->tp, agno, &sa->agfl_bp);
-	if (error && want_ag_read_header_failure(sc, XFS_SCRUB_TYPE_AGFL))
-		return error;
+		error = __xchk_ag_read_headers(sc, sa->pag->pag_agno, sa);
+		if (error)
+			return error;
 
-	return 0;
+		/*
+		 * Decide if this AG is quiet enough for all metadata to be
+		 * consistent with each other.  XFS allows the AG header buffer
+		 * locks to cycle across transaction rolls while processing
+		 * chains of deferred ops, which means that there could be
+		 * other threads in the middle of processing a chain of
+		 * deferred ops.  For regular operations we are careful about
+		 * ordering operations to prevent collisions between threads
+		 * (which is why we don't need a per-AG lock), but scrub and
+		 * repair have to serialize against chained operations.
+		 *
+		 * We just locked all the AG headers buffers; now take a look
+		 * to see if there are any intents in progress.  If there are,
+		 * drop the AG headers and wait for the intents to drain.
+		 * Since we hold all the AG header locks for the duration of
+		 * the scrub, this is the only time we have to sample the
+		 * intents counter; any threads increasing it after this point
+		 * can't possibly be in the middle of a chain of AG metadata
+		 * updates.
+		 *
+		 * Obviously, this should be slanted against scrub and in favor
+		 * of runtime threads.
+		 */
+		if (!xfs_drain_busy(&sa->pag->pag_intents))
+			return 0;
+
+		if (sa->agfl_bp) {
+			xfs_trans_brelse(sc->tp, sa->agfl_bp);
+			sa->agfl_bp = NULL;
+		}
+
+		if (sa->agf_bp) {
+			xfs_trans_brelse(sc->tp, sa->agf_bp);
+			sa->agf_bp = NULL;
+		}
+
+		if (sa->agi_bp) {
+			xfs_trans_brelse(sc->tp, sa->agi_bp);
+			sa->agi_bp = NULL;
+		}
+
+		error = xfs_perag_drain_intents(sa->pag);
+	} while (!error);
+
+	return error;
 }
 
 /* Release all the AG btree cursors. */
@@ -653,15 +724,63 @@ xchk_ag_init(
 	return 0;
 }
 
-/* Lock everything we need to work on realtime metadata. */
-void
+#ifdef CONFIG_XFS_RT
+/* Lock everything we need to work on realtime metadata and wait for intents. */
+int
+xchk_rt_lock(
+	struct xfs_scrub	*sc,
+	struct xchk_rt		*sr)
+{
+	int			error = 0;
+
+	do {
+		if (xchk_should_terminate(sc, &error))
+			return error;
+
+		xfs_rtlock(NULL, sc->mp, XFS_RTLOCK_ALL);
+
+		/*
+		 * Decide if the RT volume is quiet enough for all metadata to
+		 * be consistent with each other.  Regular file IO doesn't get
+		 * to lock all the rt inodes at the same time, which means that
+		 * there could be other threads in the middle of processing a
+		 * chain of deferred ops.
+		 *
+		 * We just locked all the rt inodes; now take a look to see if
+		 * there are any rt intents in progress.  If there are, drop
+		 * the rt inode locks and wait for the intents to drain.  Since
+		 * we hold the rt inode locks for the duration of the scrub,
+		 * this is the only time we have to sample the intents counter;
+		 * any threads increasing it after this point can't possibly be
+		 * in the middle of a chain of rt metadata updates.
+		 *
+		 * Obviously, this should be slanted against scrub and in favor
+		 * of runtime threads.
+		 */
+		if (!xfs_drain_busy(&sc->mp->m_rt_intents)) {
+			sr->locked = true;
+			return 0;
+		}
+
+		xfs_rtunlock(sc->mp, XFS_RTLOCK_ALL);
+
+		error = xfs_rt_drain_intents(sc->mp);
+	} while (!error);
+
+	return error;
+}
+#else
+/* Lock everything we need to work on realtime metadata and wait for intents. */
+int
 xchk_rt_lock(
 	struct xfs_scrub	*sc,
 	struct xchk_rt		*sr)
 {
 	xfs_rtlock(NULL, sc->mp, XFS_RTLOCK_ALL);
 	sr->locked = true;
+	return 0;
 }
+#endif /* CONFIG_XFS_RT */
 
 /*
  * For scrubbing a realtime file, grab all the in-core resources we'll need to
@@ -669,14 +788,17 @@ xchk_rt_lock(
  * metadata inodes.  Callers must not join these inodes to the transaction
  * with non-zero lockflags or concurrency problems will result.
  */
-void
+int
 xchk_rt_init(
 	struct xfs_scrub	*sc,
 	struct xchk_rt		*sr)
 {
 	struct xfs_mount	*mp = sc->mp;
+	int			error;
 
-	xchk_rt_lock(sc, sr);
+	error = xchk_rt_lock(sc, sr);
+	if (error)
+		return error;
 
 	if (xfs_has_rtrmapbt(mp))
 		sr->rmap_cur = xfs_rtrmapbt_init_cursor(mp, sc->tp,
@@ -685,6 +807,8 @@ xchk_rt_init(
 	if (xfs_has_reflink(mp))
 		sr->refc_cur = xfs_rtrefcountbt_init_cursor(mp, sc->tp,
 				mp->m_rrefcountip);
+
+	return 0;
 }
 
 /*
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index f2e1cf719c06..36a0be71cfbf 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -155,9 +155,9 @@ xchk_ag_init_existing(
 	return error == -ENOENT ? -EFSCORRUPTED : error;
 }
 
-void xchk_rt_init(struct xfs_scrub *sc, struct xchk_rt *sr);
+int xchk_rt_init(struct xfs_scrub *sc, struct xchk_rt *sr);
 void xchk_rt_btcur_free(struct xchk_rt *sr);
-void xchk_rt_lock(struct xfs_scrub *sc, struct xchk_rt *sr);
+int xchk_rt_lock(struct xfs_scrub *sc, struct xchk_rt *sr);
 void xchk_rt_unlock(struct xfs_scrub *sc, struct xchk_rt *sr);
 int xchk_ag_read_headers(struct xfs_scrub *sc, xfs_agnumber_t agno,
 		struct xchk_ag *sa);
diff --git a/fs/xfs/scrub/inode_repair.c b/fs/xfs/scrub/inode_repair.c
index 97605313f097..8eec28aa1a95 100644
--- a/fs/xfs/scrub/inode_repair.c
+++ b/fs/xfs/scrub/inode_repair.c
@@ -597,7 +597,9 @@ xrep_dinode_count_rt_rmaps(
 	    xrep_is_rtmeta_ino(sc, sc->sm->sm_ino))
 		return 0;
 
-	xchk_rt_lock(sc, &sc->sr);
+	error = xchk_rt_lock(sc, &sc->sr);
+	if (error)
+		return error;
 	xrep_rt_btcur_init(sc, &sc->sr);
 	error = xfs_rmap_query_all(sc->sr.rmap_cur, xrep_dinode_walk_rtrmap,
 			dis);
diff --git a/fs/xfs/scrub/repair.c b/fs/xfs/scrub/repair.c
index c44283b0a502..8b4951ebe2cc 100644
--- a/fs/xfs/scrub/repair.c
+++ b/fs/xfs/scrub/repair.c
@@ -487,6 +487,7 @@ xrep_newbt_schedule_reap(
 
 	INIT_LIST_HEAD(&efi_item.xefi_list);
 	list_add(&efi_item.xefi_list, &items);
+	xfs_fs_bump_intents(xnr->sc->mp, false, resv->fsbno);
 	resv->efi = xfs_extent_free_defer_type.create_intent(xnr->sc->tp,
 			&items, 1, false);
 }
@@ -708,6 +709,7 @@ xrep_newbt_destroy(
 			goto junkit;
 
 		list_del(&resv->list);
+		xfs_fs_drop_intents(sc->mp, false, resv->fsbno);
 		kmem_free(resv);
 	}
 
@@ -720,6 +722,7 @@ junkit:
 	list_for_each_entry_safe(resv, n, &xnr->resv_list, list) {
 		xfs_extent_free_defer_type.abort_intent(resv->efi);
 		list_del(&resv->list);
+		xfs_fs_drop_intents(sc->mp, false, resv->fsbno);
 		kmem_free(resv);
 	}
 
diff --git a/fs/xfs/scrub/rtbitmap.c b/fs/xfs/scrub/rtbitmap.c
index 8f354d27e3ea..1f3e3354749d 100644
--- a/fs/xfs/scrub/rtbitmap.c
+++ b/fs/xfs/scrub/rtbitmap.c
@@ -46,8 +46,7 @@ xchk_setup_rtbitmap(
 	if (error)
 		return error;
 
-	xchk_rt_init(sc, &sc->sr);
-	return 0;
+	return xchk_rt_init(sc, &sc->sr);
 }
 
 /* Realtime bitmap. */
diff --git a/fs/xfs/scrub/rtrefcount.c b/fs/xfs/scrub/rtrefcount.c
index 2a7d5268bd43..860b9feff951 100644
--- a/fs/xfs/scrub/rtrefcount.c
+++ b/fs/xfs/scrub/rtrefcount.c
@@ -33,8 +33,7 @@ xchk_setup_rtrefcountbt(
 	if (error)
 		return error;
 
-	xchk_rt_init(sc, &sc->sr);
-	return 0;
+	return xchk_rt_init(sc, &sc->sr);
 }
 
 /* Realtime Reference count btree scrubber. */
diff --git a/fs/xfs/scrub/rtrmap.c b/fs/xfs/scrub/rtrmap.c
index acd37e63c4bd..25d56cb1b506 100644
--- a/fs/xfs/scrub/rtrmap.c
+++ b/fs/xfs/scrub/rtrmap.c
@@ -50,8 +50,7 @@ xchk_setup_rtrmapbt(
 	if (error)
 		return error;
 
-	xchk_rt_init(sc, &sc->sr);
-	return 0;
+	return xchk_rt_init(sc, &sc->sr);
 }
 
 /* Realtime reverse mapping. */
diff --git a/fs/xfs/scrub/rtrmap_repair.c b/fs/xfs/scrub/rtrmap_repair.c
index edcdc407f6cb..b9fd891f4476 100644
--- a/fs/xfs/scrub/rtrmap_repair.c
+++ b/fs/xfs/scrub/rtrmap_repair.c
@@ -583,7 +583,9 @@ xrep_rtrmap_find_rmaps(
 	error = xchk_setup_fs(sc);
 	if (error)
 		return error;
-	xchk_rt_lock(sc, &sc->sr);
+	error = xchk_rt_lock(sc, &sc->sr);
+	if (error)
+		return error;
 
 	/* Scan for old rtrmap blocks. */
 	for_each_perag(sc->mp, agno, pag) {
diff --git a/fs/xfs/scrub/rtsummary.c b/fs/xfs/scrub/rtsummary.c
index be5ec3e8bf12..f4e1a7e1466b 100644
--- a/fs/xfs/scrub/rtsummary.c
+++ b/fs/xfs/scrub/rtsummary.c
@@ -71,8 +71,7 @@ xchk_setup_rtsummary(
 	if (error)
 		return error;
 
-	xchk_rt_init(sc, &sc->sr);
-	return 0;
+	return xchk_rt_init(sc, &sc->sr);
 }
 
 /* Update the summary file to reflect the free extent that we've accumulated. */
diff --git a/fs/xfs/xfs_bmap_item.c b/fs/xfs/xfs_bmap_item.c
index 4c5cb8072fa6..a4cb64b080f7 100644
--- a/fs/xfs/xfs_bmap_item.c
+++ b/fs/xfs/xfs_bmap_item.c
@@ -342,6 +342,21 @@ xfs_bmap_update_create_done(
 	return &xfs_trans_get_bud(tp, BUI_ITEM(intent))->bud_item;
 }
 
+static inline void
+xfs_bmap_drop_intents(
+	struct xfs_mount		*mp,
+	const struct xfs_bmap_intent	*bi,
+	xfs_fsblock_t			orig_startblock)
+{
+	bool				rt;
+
+	if (!xfs_has_rmapbt(mp))
+		return;
+
+	rt = xfs_ifork_is_realtime(bi->bi_owner, bi->bi_whichfork);
+	xfs_fs_drop_intents(mp, rt, orig_startblock);
+}
+
 /* Process a deferred rmap update. */
 STATIC int
 xfs_bmap_update_finish_item(
@@ -351,14 +366,25 @@ xfs_bmap_update_finish_item(
 	struct xfs_btree_cur		**state)
 {
 	struct xfs_bmap_intent		*bi;
+	struct xfs_mount		*mp = tp->t_mountp;
+	xfs_fsblock_t			orig_startblock;
 	int				error;
 
 	bi = container_of(item, struct xfs_bmap_intent, bi_list);
+	orig_startblock = bi->bi_bmap.br_startblock;
 	error = xfs_trans_log_finish_bmap_update(tp, BUD_ITEM(done), bi);
 	if (!error && bi->bi_bmap.br_blockcount > 0) {
 		ASSERT(bi->bi_type == XFS_BMAP_UNMAP);
 		return -EAGAIN;
 	}
+
+	/*
+	 * Drop our intent counter reference now that we've either queued a
+	 * deferred rmap intent or failed.  Be careful to use the original
+	 * startblock since the finishing functions can update the intent
+	 * state.
+	 */
+	xfs_bmap_drop_intents(mp, bi, orig_startblock);
 	kmem_cache_free(xfs_bmap_intent_cache, bi);
 	return error;
 }
@@ -371,17 +397,41 @@ xfs_bmap_update_abort_intent(
 	xfs_bui_release(BUI_ITEM(intent));
 }
 
-/* Cancel a deferred rmap update. */
+/* Cancel a deferred bmap update. */
 STATIC void
 xfs_bmap_update_cancel_item(
+	struct xfs_mount		*mp,
 	struct list_head		*item)
 {
 	struct xfs_bmap_intent		*bi;
 
 	bi = container_of(item, struct xfs_bmap_intent, bi_list);
+	xfs_bmap_drop_intents(mp, bi, bi->bi_bmap.br_startblock);
 	kmem_cache_free(xfs_bmap_intent_cache, bi);
 }
 
+/* Add a deferred bmap update. */
+STATIC void
+xfs_bmap_update_add_item(
+	struct xfs_mount		*mp,
+	const struct list_head		*item)
+{
+	const struct xfs_bmap_intent	*bi;
+	bool				rt;
+
+	bi = container_of(item, struct xfs_bmap_intent, bi_list);
+
+	/*
+	 * Grab an intent counter reference on behalf of the deferred rmap
+	 * intent item that we will queue when we finish this bmap work.
+	 */
+	if (!xfs_has_rmapbt(mp))
+		return;
+
+	rt = xfs_ifork_is_realtime(bi->bi_owner, bi->bi_whichfork);
+	xfs_fs_bump_intents(mp, rt, bi->bi_bmap.br_startblock);
+}
+
 const struct xfs_defer_op_type xfs_bmap_update_defer_type = {
 	.max_items	= XFS_BUI_MAX_FAST_EXTENTS,
 	.create_intent	= xfs_bmap_update_create_intent,
@@ -389,6 +439,7 @@ const struct xfs_defer_op_type xfs_bmap_update_defer_type = {
 	.create_done	= xfs_bmap_update_create_done,
 	.finish_item	= xfs_bmap_update_finish_item,
 	.cancel_item	= xfs_bmap_update_cancel_item,
+	.add_item	= xfs_bmap_update_add_item,
 };
 
 /* Is this recovered BUI ok? */
diff --git a/fs/xfs/xfs_extfree_item.c b/fs/xfs/xfs_extfree_item.c
index 34de89077961..5345eb8efe30 100644
--- a/fs/xfs/xfs_extfree_item.c
+++ b/fs/xfs/xfs_extfree_item.c
@@ -489,6 +489,15 @@ xfs_extent_free_create_done(
 	return &xfs_trans_get_efd(tp, EFI_ITEM(intent), count)->efd_item;
 }
 
+static inline void
+xfs_extent_free_drop_intents(
+	struct xfs_mount			*mp,
+	const struct xfs_extent_free_item	*xefi)
+{
+	xfs_fs_drop_intents(mp, xefi->xefi_flags & XFS_EFI_REALTIME,
+			xefi->xefi_startblock);
+}
+
 /* Process a free extent. */
 STATIC int
 xfs_extent_free_finish_item(
@@ -497,6 +506,7 @@ xfs_extent_free_finish_item(
 	struct list_head		*item,
 	struct xfs_btree_cur		**state)
 {
+	struct xfs_mount		*mp = tp->t_mountp;
 	struct xfs_extent_free_item	*xefi;
 	int				error;
 
@@ -512,6 +522,7 @@ xfs_extent_free_finish_item(
 	}
 
 	error = xfs_trans_free_extent(tp, EFD_ITEM(done), xefi);
+	xfs_extent_free_drop_intents(mp, xefi);
 	kmem_cache_free(xfs_extfree_item_cache, xefi);
 	return error;
 }
@@ -527,14 +538,31 @@ xfs_extent_free_abort_intent(
 /* Cancel a free extent. */
 STATIC void
 xfs_extent_free_cancel_item(
+	struct xfs_mount		*mp,
 	struct list_head		*item)
 {
 	struct xfs_extent_free_item	*xefi;
 
 	xefi = container_of(item, struct xfs_extent_free_item, xefi_list);
+	xfs_extent_free_drop_intents(mp, xefi);
 	kmem_cache_free(xfs_extfree_item_cache, xefi);
 }
 
+/* Add a deferred free extent. */
+STATIC void
+xfs_extent_free_add_item(
+	struct xfs_mount		*mp,
+	const struct list_head		*item)
+{
+	const struct xfs_extent_free_item *xefi;
+
+	xefi = container_of(item, struct xfs_extent_free_item, xefi_list);
+
+	/* Grab an intent counter reference for this intent item. */
+	xfs_fs_bump_intents(mp, xefi->xefi_flags & XFS_EFI_REALTIME,
+			xefi->xefi_startblock);
+}
+
 const struct xfs_defer_op_type xfs_extent_free_defer_type = {
 	.max_items	= XFS_EFI_MAX_FAST_EXTENTS,
 	.create_intent	= xfs_extent_free_create_intent,
@@ -542,6 +570,7 @@ const struct xfs_defer_op_type xfs_extent_free_defer_type = {
 	.create_done	= xfs_extent_free_create_done,
 	.finish_item	= xfs_extent_free_finish_item,
 	.cancel_item	= xfs_extent_free_cancel_item,
+	.add_item	= xfs_extent_free_add_item,
 };
 
 /*
@@ -596,6 +625,7 @@ xfs_agfl_free_finish_item(
 	extp->ext_len = xefi->xefi_blockcount;
 	efdp->efd_next_extent++;
 
+	xfs_extent_free_drop_intents(mp, xefi);
 	kmem_cache_free(xfs_extfree_item_cache, xefi);
 	return error;
 }
@@ -608,6 +638,7 @@ const struct xfs_defer_op_type xfs_agfl_free_defer_type = {
 	.create_done	= xfs_extent_free_create_done,
 	.finish_item	= xfs_agfl_free_finish_item,
 	.cancel_item	= xfs_extent_free_cancel_item,
+	.add_item	= xfs_extent_free_add_item,
 };
 
 /* Is this recovered EFI ok? */
diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c
index 8c4975556a73..fb38e638e1ca 100644
--- a/fs/xfs/xfs_mount.c
+++ b/fs/xfs/xfs_mount.c
@@ -1510,3 +1510,97 @@ xfs_hook_call(
 	return srcu_notifier_call_chain(&chain->head, val, priv);
 }
 #endif /* CONFIG_XFS_LIVE_HOOKS */
+
+#ifdef CONFIG_XFS_DRAIN_INTENTS
+static inline void xfs_drain_bump(struct xfs_drain *dr)
+{
+	atomic_inc(&dr->dr_count);
+}
+
+static inline void xfs_drain_drop(struct xfs_drain *dr)
+{
+	ASSERT(atomic_read(&dr->dr_count) > 0);
+
+	if (atomic_dec_and_test(&dr->dr_count))
+		wake_up(&dr->dr_waiters);
+}
+
+static inline int xfs_drain_wait(struct xfs_drain *dr)
+{
+	return wait_event_killable(dr->dr_waiters,
+			atomic_read(&dr->dr_count) == 0);
+}
+
+#ifdef CONFIG_XFS_RT
+# define xfs_rt_drain_bump(dr)	xfs_drain_bump(dr)
+# define xfs_rt_drain_drop(dr)	xfs_drain_drop(dr)
+
+/*
+ * Wait for the pending intent count for realtime metadata to hit zero.
+ * Callers must not hold any rt metadata inode locks.
+ */
+int
+xfs_rt_drain_intents(
+	struct xfs_mount	*mp)
+{
+	trace_xfs_rt_wait_intents(mp, __return_address);
+	return xfs_drain_wait(&mp->m_rt_intents);
+}
+#else
+# define trace_xfs_rt_bump_intents(...)
+# define trace_xfs_rt_drop_intents(...)
+# define xfs_rt_drain_bump(dr)
+# define xfs_rt_drain_drop(dr)
+#endif /* CONFIG_XFS_RT */
+
+/* Add an item to the pending count. */
+void
+xfs_fs_bump_intents(
+	struct xfs_mount	*mp,
+	bool			isrt,
+	xfs_fsblock_t		fsb)
+{
+	struct xfs_perag	*pag;
+
+	if (isrt) {
+		trace_xfs_rt_bump_intents(mp, __return_address);
+		xfs_rt_drain_bump(&mp->m_rt_intents);
+		return;
+	}
+
+	pag = xfs_perag_get(mp, XFS_FSB_TO_AGNO(mp, fsb));
+	trace_xfs_perag_bump_intents(pag, __return_address);
+	xfs_drain_bump(&pag->pag_intents);
+	xfs_perag_put(pag);
+}
+
+/* Remove an item from the pending count. */
+void
+xfs_fs_drop_intents(struct xfs_mount *mp, bool isrt, xfs_fsblock_t fsb)
+{
+	struct xfs_perag	*pag;
+
+	if (isrt) {
+		trace_xfs_rt_drop_intents(mp, __return_address);
+		xfs_rt_drain_drop(&mp->m_rt_intents);
+		return;
+	}
+
+	pag = xfs_perag_get(mp, XFS_FSB_TO_AGNO(mp, fsb));
+	trace_xfs_perag_drop_intents(pag, __return_address);
+	xfs_drain_drop(&pag->pag_intents);
+	xfs_perag_put(pag);
+}
+
+/*
+ * Wait for the pending intent count for AG metadata to hit zero.
+ * Callers must not hold any AG header buffers.
+ */
+int
+xfs_perag_drain_intents(
+	struct xfs_perag	*pag)
+{
+	trace_xfs_perag_wait_intents(pag, __return_address);
+	return xfs_drain_wait(&pag->pag_intents);
+}
+#endif /* CONFIG_XFS_DRAIN_INTENTS */
diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h
index 5943f8f100a7..79fd3293ca36 100644
--- a/fs/xfs/xfs_mount.h
+++ b/fs/xfs/xfs_mount.h
@@ -74,6 +74,63 @@ int xfs_hook_call(struct xfs_hook_chain *chain, unsigned long val, void *priv);
 #endif
 
 /*
+ * Passive drain mechanism.  This data structure tracks a count of some items
+ * and contains a waitqueue for callers who would like to wake up when the
+ * count hits zero.
+ */
+struct xfs_drain {
+#ifdef CONFIG_XFS_DRAIN_INTENTS
+	/* Number of items pending in some part of the filesystem. */
+	atomic_t		dr_count;
+
+	/* Queue to wait for dri_count to go to zero */
+	struct wait_queue_head	dr_waiters;
+#endif /* CONFIG_XFS_DRAIN_INTENTS */
+};
+
+#ifdef CONFIG_XFS_DRAIN_INTENTS
+# ifdef CONFIG_XFS_RT
+int xfs_rt_drain_intents(struct xfs_mount *mp);
+# else
+#  define xfs_rt_wait_intents(mp)		(-ENOSYS)
+# endif /* CONFIG_XFS_RT */
+
+int xfs_perag_drain_intents(struct xfs_perag *pag);
+
+void xfs_fs_bump_intents(struct xfs_mount *mp, bool isrt, xfs_fsblock_t fsb);
+void xfs_fs_drop_intents(struct xfs_mount *mp, bool isrt, xfs_fsblock_t fsb);
+
+/* Are there work items pending? */
+static inline bool xfs_drain_busy(struct xfs_drain *dr)
+{
+	return atomic_read(&dr->dr_count) > 0;
+}
+
+static inline void xfs_drain_init(struct xfs_drain *dr)
+{
+	atomic_set(&dr->dr_count, 0);
+	init_waitqueue_head(&dr->dr_waiters);
+}
+
+static inline void xfs_drain_free(struct xfs_drain *dr)
+{
+	ASSERT(atomic_read(&dr->dr_count) == 0);
+}
+#else
+static inline void
+xfs_fs_bump_intents(struct xfs_mount *mp, bool isrt, xfs_fsblock_t fsb)
+{
+}
+
+static inline void
+xfs_fs_drop_intents(struct xfs_mount *mp, bool isrt, xfs_fsblock_t fsb)
+{
+}
+# define xfs_drain_init(dr)
+# define xfs_drain_free(dr)
+#endif /* CONFIG_XFS_DRAIN_INTENTS */
+
+/*
  * Per-cpu deferred inode inactivation GC lists.
  */
 struct xfs_inodegc {
@@ -276,6 +333,17 @@ typedef struct xfs_mount {
 
 	/* online file link count check stuff */
 	struct xfs_hook_chain	m_nlink_delta_hooks;
+
+#ifdef CONFIG_XFS_RT
+	/*
+	 * We use xfs_drain to track the number of deferred log intent items
+	 * that have been queued (but not yet processed) so that waiters (e.g.
+	 * scrub) will not lock resources when other threads are in the middle
+	 * of processing a chain of intent items only to find momentary
+	 * inconsistencies.
+	 */
+	struct xfs_drain	m_rt_intents;
+#endif /* CONFIG_XFS_RT */
 } xfs_mount_t;
 
 #define M_IGEO(mp)		(&(mp)->m_ino_geo)
diff --git a/fs/xfs/xfs_refcount_item.c b/fs/xfs/xfs_refcount_item.c
index 8d481cc5ad6a..f4daf76c82d3 100644
--- a/fs/xfs/xfs_refcount_item.c
+++ b/fs/xfs/xfs_refcount_item.c
@@ -364,9 +364,12 @@ xfs_refcount_update_finish_item(
 	struct xfs_btree_cur		**state)
 {
 	struct xfs_refcount_intent	*ri;
+	struct xfs_mount		*mp = tp->t_mountp;
+	xfs_fsblock_t			orig_startblock;
 	int				error;
 
 	ri = container_of(item, struct xfs_refcount_intent, ri_list);
+	orig_startblock = ri->ri_startblock;
 	error = xfs_trans_log_finish_refcount_update(tp, CUD_ITEM(done), ri,
 			state);
 
@@ -376,6 +379,13 @@ xfs_refcount_update_finish_item(
 		       ri->ri_type == XFS_REFCOUNT_DECREASE);
 		return -EAGAIN;
 	}
+
+	/*
+	 * Drop our intent counter reference now that we've finished all the
+	 * work or failed.  Be careful to use the original startblock because
+	 * the finishing functions can update the intent state.
+	 */
+	xfs_fs_drop_intents(mp, ri->ri_realtime, orig_startblock);
 	kmem_cache_free(xfs_refcount_intent_cache, ri);
 	return error;
 }
@@ -391,14 +401,28 @@ xfs_refcount_update_abort_intent(
 /* Cancel a deferred refcount update. */
 STATIC void
 xfs_refcount_update_cancel_item(
+	struct xfs_mount		*mp,
 	struct list_head		*item)
 {
 	struct xfs_refcount_intent	*ri;
 
 	ri = container_of(item, struct xfs_refcount_intent, ri_list);
+	xfs_fs_drop_intents(mp, ri->ri_realtime, ri->ri_startblock);
 	kmem_cache_free(xfs_refcount_intent_cache, ri);
 }
 
+/* Add a deferred refcount update. */
+STATIC void
+xfs_refcount_update_add_item(
+	struct xfs_mount		*mp,
+	const struct list_head		*item)
+{
+	const struct xfs_refcount_intent *ri;
+
+	ri = container_of(item, struct xfs_refcount_intent, ri_list);
+	xfs_fs_bump_intents(mp, ri->ri_realtime, ri->ri_startblock);
+}
+
 const struct xfs_defer_op_type xfs_refcount_update_defer_type = {
 	.max_items	= XFS_CUI_MAX_FAST_EXTENTS,
 	.create_intent	= xfs_refcount_update_create_intent,
@@ -407,6 +431,7 @@ const struct xfs_defer_op_type xfs_refcount_update_defer_type = {
 	.finish_item	= xfs_refcount_update_finish_item,
 	.finish_cleanup = xfs_refcount_finish_one_cleanup,
 	.cancel_item	= xfs_refcount_update_cancel_item,
+	.add_item	= xfs_refcount_update_add_item,
 };
 
 /* Is this recovered CUI ok? */
diff --git a/fs/xfs/xfs_rmap_item.c b/fs/xfs/xfs_rmap_item.c
index bf2e7d37668f..0d660d000fe7 100644
--- a/fs/xfs/xfs_rmap_item.c
+++ b/fs/xfs/xfs_rmap_item.c
@@ -411,11 +411,13 @@ xfs_rmap_update_finish_item(
 	struct xfs_btree_cur		**state)
 {
 	struct xfs_rmap_intent		*ri;
+	struct xfs_mount		*mp = tp->t_mountp;
 	int				error;
 
 	ri = container_of(item, struct xfs_rmap_intent, ri_list);
 	error = xfs_trans_log_finish_rmap_update(tp, RUD_ITEM(done), ri,
 			state);
+	xfs_fs_drop_intents(mp, ri->ri_realtime, ri->ri_bmap.br_startblock);
 	kmem_cache_free(xfs_rmap_intent_cache, ri);
 	return error;
 }
@@ -431,14 +433,28 @@ xfs_rmap_update_abort_intent(
 /* Cancel a deferred rmap update. */
 STATIC void
 xfs_rmap_update_cancel_item(
+	struct xfs_mount		*mp,
 	struct list_head		*item)
 {
 	struct xfs_rmap_intent		*ri;
 
 	ri = container_of(item, struct xfs_rmap_intent, ri_list);
+	xfs_fs_drop_intents(mp, ri->ri_realtime, ri->ri_bmap.br_startblock);
 	kmem_cache_free(xfs_rmap_intent_cache, ri);
 }
 
+/* Add a deferred rmap update. */
+STATIC void
+xfs_rmap_update_add_item(
+	struct xfs_mount		*mp,
+	const struct list_head		*item)
+{
+	const struct xfs_rmap_intent	*ri;
+
+	ri = container_of(item, struct xfs_rmap_intent, ri_list);
+	xfs_fs_bump_intents(mp, ri->ri_realtime, ri->ri_bmap.br_startblock);
+}
+
 const struct xfs_defer_op_type xfs_rmap_update_defer_type = {
 	.max_items	= XFS_RUI_MAX_FAST_EXTENTS,
 	.create_intent	= xfs_rmap_update_create_intent,
@@ -447,6 +463,7 @@ const struct xfs_defer_op_type xfs_rmap_update_defer_type = {
 	.finish_item	= xfs_rmap_update_finish_item,
 	.finish_cleanup = xfs_rmap_finish_one_cleanup,
 	.cancel_item	= xfs_rmap_update_cancel_item,
+	.add_item	= xfs_rmap_update_add_item,
 };
 
 /* Is this recovered RUI ok? */
diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
index 879e56cd8498..224485863c54 100644
--- a/fs/xfs/xfs_super.c
+++ b/fs/xfs/xfs_super.c
@@ -726,6 +726,9 @@ xfs_mount_free(
 
 	ASSERT(!mutex_is_locked(&mp->m_scrub_freeze));
 	mutex_destroy(&mp->m_scrub_freeze);
+#ifdef CONFIG_XFS_RT
+	xfs_drain_free(&mp->m_rt_intents);
+#endif
 
 	kmem_free(mp);
 }
@@ -1968,6 +1971,10 @@ static int xfs_init_fs_context(
 	if (!mp)
 		return -ENOMEM;
 
+#ifdef CONFIG_XFS_RT
+	xfs_drain_init(&mp->m_rt_intents);
+#endif
+
 	spin_lock_init(&mp->m_sb_lock);
 	spin_lock_init(&mp->m_agirotor_lock);
 	INIT_RADIX_TREE(&mp->m_perag_tree, GFP_ATOMIC);
diff --git a/fs/xfs/xfs_swapext_item.c b/fs/xfs/xfs_swapext_item.c
index f51e83e33027..9e084f9996c1 100644
--- a/fs/xfs/xfs_swapext_item.c
+++ b/fs/xfs/xfs_swapext_item.c
@@ -346,6 +346,7 @@ xfs_swapext_abort_intent(
 /* Cancel a deferred swapext update. */
 STATIC void
 xfs_swapext_cancel_item(
+	struct xfs_mount		*mp,
 	struct list_head		*item)
 {
 	struct xfs_swapext_intent	*sxi;
@@ -354,6 +355,14 @@ xfs_swapext_cancel_item(
 	kmem_free(sxi);
 }
 
+/* Add a deferred swapext update. */
+STATIC void
+xfs_swapext_add_item(
+	struct xfs_mount		*mp,
+	const struct list_head		*item)
+{
+}
+
 const struct xfs_defer_op_type xfs_swapext_defer_type = {
 	.max_items	= XFS_SXI_MAX_FAST_EXTENTS,
 	.create_intent	= xfs_swapext_create_intent,
@@ -361,6 +370,7 @@ const struct xfs_defer_op_type xfs_swapext_defer_type = {
 	.create_done	= xfs_swapext_create_done,
 	.finish_item	= xfs_swapext_finish_item,
 	.cancel_item	= xfs_swapext_cancel_item,
+	.add_item	= xfs_swapext_add_item,
 };
 
 /* Is this recovered SXI ok? */
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index 3c1e61e57805..fed0ede02820 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -2767,6 +2767,44 @@ DEFINE_FREE_EXTENT_DEFERRED_EVENT(xfs_agfl_free_deferred);
 DEFINE_FREE_EXTENT_DEFERRED_EVENT(xfs_extent_free_defer);
 DEFINE_FREE_EXTENT_DEFERRED_EVENT(xfs_extent_free_deferred);
 
+DECLARE_EVENT_CLASS(xfs_defer_pending_item_class,
+	TP_PROTO(struct xfs_mount *mp, struct xfs_defer_pending *dfp,
+		 void *item),
+	TP_ARGS(mp, dfp, item),
+	TP_STRUCT__entry(
+		__field(dev_t, dev)
+		__field(int, type)
+		__field(void *, intent)
+		__field(void *, item)
+		__field(char, committed)
+		__field(int, nr)
+	),
+	TP_fast_assign(
+		__entry->dev = mp ? mp->m_super->s_dev : 0;
+		__entry->type = dfp->dfp_type;
+		__entry->intent = dfp->dfp_intent;
+		__entry->item = item;
+		__entry->committed = dfp->dfp_done != NULL;
+		__entry->nr = dfp->dfp_count;
+	),
+	TP_printk("dev %d:%d optype %d intent %p item %p committed %d nr %d",
+		  MAJOR(__entry->dev), MINOR(__entry->dev),
+		  __entry->type,
+		  __entry->intent,
+		  __entry->item,
+		  __entry->committed,
+		  __entry->nr)
+)
+#define DEFINE_DEFER_PENDING_ITEM_EVENT(name) \
+DEFINE_EVENT(xfs_defer_pending_item_class, name, \
+	TP_PROTO(struct xfs_mount *mp, struct xfs_defer_pending *dfp, \
+		 void *item), \
+	TP_ARGS(mp, dfp, item))
+
+DEFINE_DEFER_PENDING_ITEM_EVENT(xfs_defer_add_item);
+DEFINE_DEFER_PENDING_ITEM_EVENT(xfs_defer_cancel_item);
+DEFINE_DEFER_PENDING_ITEM_EVENT(xfs_defer_finish_item);
+
 /* rmap tracepoints */
 DECLARE_EVENT_CLASS(xfs_rmap_class,
 	TP_PROTO(struct xfs_btree_cur *cur,
@@ -4931,6 +4969,70 @@ TRACE_EVENT(xfs_growfs_check_rtgeom,
 		  __entry->min_logfsbs)
 );
 
+#ifdef CONFIG_XFS_DRAIN_INTENTS
+DECLARE_EVENT_CLASS(xfs_perag_intents_class,
+	TP_PROTO(struct xfs_perag *pag, void *caller_ip),
+	TP_ARGS(pag, caller_ip),
+	TP_STRUCT__entry(
+		__field(dev_t, dev)
+		__field(xfs_agnumber_t, agno)
+		__field(long, nr_intents)
+		__field(void *, caller_ip)
+	),
+	TP_fast_assign(
+		__entry->dev = pag->pag_mount->m_super->s_dev;
+		__entry->agno = pag->pag_agno;
+		__entry->nr_intents = atomic_read(&pag->pag_intents.dr_count);
+		__entry->caller_ip = caller_ip;
+	),
+	TP_printk("dev %d:%d agno 0x%x intents %ld caller %pS",
+		  MAJOR(__entry->dev), MINOR(__entry->dev),
+		  __entry->agno,
+		  __entry->nr_intents,
+		  __entry->caller_ip)
+);
+
+#define DEFINE_PERAG_INTENTS_EVENT(name)	\
+DEFINE_EVENT(xfs_perag_intents_class, name,					\
+	TP_PROTO(struct xfs_perag *pag, void *caller_ip), \
+	TP_ARGS(pag, caller_ip))
+DEFINE_PERAG_INTENTS_EVENT(xfs_perag_bump_intents);
+DEFINE_PERAG_INTENTS_EVENT(xfs_perag_drop_intents);
+DEFINE_PERAG_INTENTS_EVENT(xfs_perag_wait_intents);
+
+# ifdef CONFIG_XFS_RT
+DECLARE_EVENT_CLASS(xfs_rt_intents_class,
+	TP_PROTO(struct xfs_mount *mp, void *caller_ip),
+	TP_ARGS(mp, caller_ip),
+	TP_STRUCT__entry(
+		__field(dev_t, dev)
+		__field(dev_t, rtdev)
+		__field(long, nr_intents)
+		__field(void *, caller_ip)
+	),
+	TP_fast_assign(
+		__entry->dev = mp->m_super->s_dev;
+		__entry->rtdev = mp->m_rtdev_targp->bt_dev;
+		__entry->nr_intents = atomic_read(&mp->m_rt_intents.dr_count);
+		__entry->caller_ip = caller_ip;
+	),
+	TP_printk("dev %d:%d rtdev %d:%d intents %ld caller %pS",
+		  MAJOR(__entry->dev), MINOR(__entry->dev),
+		  MAJOR(__entry->rtdev), MINOR(__entry->rtdev),
+		  __entry->nr_intents,
+		  __entry->caller_ip)
+);
+
+#define DEFINE_RT_INTENTS_EVENT(name)	\
+DEFINE_EVENT(xfs_rt_intents_class, name,					\
+	TP_PROTO(struct xfs_mount *mp, void *caller_ip), \
+	TP_ARGS(mp, caller_ip))
+DEFINE_RT_INTENTS_EVENT(xfs_rt_bump_intents);
+DEFINE_RT_INTENTS_EVENT(xfs_rt_drop_intents);
+DEFINE_RT_INTENTS_EVENT(xfs_rt_wait_intents);
+# endif /* CONFIG_XFS_RT */
+#endif /* CONFIG_XFS_DRAIN_INTENTS */
+
 #endif /* _TRACE_XFS_H */
 
 #undef TRACE_INCLUDE_PATH