xfs: allow queued AG intents to drain before scrubbing

When a writer thread executes a chain of log intent items, the AG header
buffer locks will cycle during a transaction roll to get from one intent
item to the next in a chain.  Although scrub takes all AG header buffer
locks, this isn't sufficient to guard against scrub checking an AG while
that writer thread is in the middle of finishing a chain because there's
no higher level locking primitive guarding allocation groups.

When there's a collision, cross-referencing between data structures
(e.g. rmapbt and refcountbt) yields false corruption events; if repair
is running, this results in incorrect repairs, which is catastrophic.

Fix this by adding to the perag structure the count of active intents
and make scrub wait until it has both AG header buffer locks and the
intent counter reaches zero.

One quirk of the drain code is that deferred bmap updates also bump and
drop the intent counter.  A fundamental decision made during the design
phase of the reverse mapping feature is that updates to the rmapbt
records are always made by the same code that updates the primary
metadata.  In other words, callers of bmapi functions expect that the
bmapi functions will queue deferred rmap updates.

Some parts of the reflink code queue deferred refcount (CUI) and bmap
(BUI) updates in the same head transaction, but the deferred work
manager completely finishes the CUI before the BUI work is started.  As
a result, the CUI drops the intent count long before the deferred rmap
(RUI) update even has a chance to bump the intent count.  The only way
to keep the intent count elevated between the CUI and RUI is for the BUI
to bump the counter until the RUI has been created.

A second quirk of the intent drain code is that deferred work items must
increment the intent counter as soon as the work item is added to the
transaction.  When a BUI completes and queues an RUI, the RUI must
increment the counter before the BUI decrements it.  The only way to
accomplish this is to require that the counter be bumped as soon as the
deferred work item is created in memory.

In the next patches we'll improve on this facility, but this patch
provides the basic functionality.

Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>

1 files changed, 84 insertions, 0 deletions

diff --git a/fs/xfs/xfs_drain.h b/fs/xfs/xfs_drain.h
new file mode 100644
index 000000000000..9b16df3cc7dc
--- /dev/null
+++ b/fs/xfs/xfs_drain.h
@@ -0,0 +1,84 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright (C) 2022-2023 Oracle.  All Rights Reserved.
+ * Author: Darrick J. Wong <djwong@kernel.org>
+ */
+#ifndef XFS_DRAIN_H_
+#define XFS_DRAIN_H_
+
+struct xfs_perag;
+
+#ifdef CONFIG_XFS_DRAIN_INTENTS
+/*
+ * 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_defer_drain {
+	/* 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;
+};
+
+void xfs_defer_drain_init(struct xfs_defer_drain *dr);
+void xfs_defer_drain_free(struct xfs_defer_drain *dr);
+
+/*
+ * Deferred Work Intent Drains
+ * ===========================
+ *
+ * When a writer thread executes a chain of log intent items, the AG header
+ * buffer locks will cycle during a transaction roll to get from one intent
+ * item to the next in a chain.  Although scrub takes all AG header buffer
+ * locks, this isn't sufficient to guard against scrub checking an AG while
+ * that writer thread is in the middle of finishing a chain because there's no
+ * higher level locking primitive guarding allocation groups.
+ *
+ * When there's a collision, cross-referencing between data structures (e.g.
+ * rmapbt and refcountbt) yields false corruption events; if repair is running,
+ * this results in incorrect repairs, which is catastrophic.
+ *
+ * The solution is to the perag structure the count of active intents and make
+ * scrub wait until it has both AG header buffer locks and the intent counter
+ * reaches zero.  It is therefore critical that deferred work threads hold the
+ * AGI or AGF buffers when decrementing the intent counter.
+ *
+ * Given a list of deferred work items, the deferred work manager will complete
+ * a work item and all the sub-items that the parent item creates before moving
+ * on to the next work item in the list.  This is also true for all levels of
+ * sub-items.  Writer threads are permitted to queue multiple work items
+ * targetting the same AG, so a deferred work item (such as a BUI) that creates
+ * sub-items (such as RUIs) must bump the intent counter and maintain it until
+ * the sub-items can themselves bump the intent counter.
+ *
+ * Therefore, the intent count tracks entire lifetimes of deferred work items.
+ * All functions that create work items must increment the intent counter as
+ * soon as the item is added to the transaction and cannot drop the counter
+ * until the item is finished or cancelled.
+ */
+struct xfs_perag *xfs_perag_intent_get(struct xfs_mount *mp,
+		xfs_agnumber_t agno);
+void xfs_perag_intent_put(struct xfs_perag *pag);
+
+void xfs_perag_intent_hold(struct xfs_perag *pag);
+void xfs_perag_intent_rele(struct xfs_perag *pag);
+
+int xfs_perag_intent_drain(struct xfs_perag *pag);
+bool xfs_perag_intent_busy(struct xfs_perag *pag);
+#else
+struct xfs_defer_drain { /* empty */ };
+
+#define xfs_defer_drain_free(dr)		((void)0)
+#define xfs_defer_drain_init(dr)		((void)0)
+
+#define xfs_perag_intent_get(mp, agno)		xfs_perag_get((mp), (agno))
+#define xfs_perag_intent_put(pag)		xfs_perag_put(pag)
+
+static inline void xfs_perag_intent_hold(struct xfs_perag *pag) { }
+static inline void xfs_perag_intent_rele(struct xfs_perag *pag) { }
+
+#endif /* CONFIG_XFS_DRAIN_INTENTS */
+
+#endif /* XFS_DRAIN_H_ */