xfs: connect refcount adjust functions to upper layers

Plumb in the upper level interface to schedule and finish deferred
refcount operations via the deferred ops mechanism.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>

fs/xfs/libxfs/xfs_defer.h

@@ -51,6 +51,7 @@ struct xfs_defer_pending {
  * find all the space it needs.
  */
 enum xfs_defer_ops_type {
+	XFS_DEFER_OPS_TYPE_REFCOUNT,
 	XFS_DEFER_OPS_TYPE_RMAP,
 	XFS_DEFER_OPS_TYPE_FREE,
 	XFS_DEFER_OPS_TYPE_MAX,

fs/xfs/libxfs/xfs_refcount.c

@@ -989,3 +989,177 @@ out_error:
 			error, _RET_IP_);
 	return error;
 }
+
+/* Clean up after calling xfs_refcount_finish_one. */
+void
+xfs_refcount_finish_one_cleanup(
+	struct xfs_trans	*tp,
+	struct xfs_btree_cur	*rcur,
+	int			error)
+{
+	struct xfs_buf		*agbp;
+
+	if (rcur == NULL)
+		return;
+	agbp = rcur->bc_private.a.agbp;
+	xfs_btree_del_cursor(rcur, error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
+	if (error)
+		xfs_trans_brelse(tp, agbp);
+}
+
+/*
+ * Process one of the deferred refcount operations.  We pass back the
+ * btree cursor to maintain our lock on the btree between calls.
+ * This saves time and eliminates a buffer deadlock between the
+ * superblock and the AGF because we'll always grab them in the same
+ * order.
+ */
+int
+xfs_refcount_finish_one(
+	struct xfs_trans		*tp,
+	struct xfs_defer_ops		*dfops,
+	enum xfs_refcount_intent_type	type,
+	xfs_fsblock_t			startblock,
+	xfs_extlen_t			blockcount,
+	xfs_fsblock_t			*new_fsb,
+	xfs_extlen_t			*new_len,
+	struct xfs_btree_cur		**pcur)
+{
+	struct xfs_mount		*mp = tp->t_mountp;
+	struct xfs_btree_cur		*rcur;
+	struct xfs_buf			*agbp = NULL;
+	int				error = 0;
+	xfs_agnumber_t			agno;
+	xfs_agblock_t			bno;
+	xfs_agblock_t			new_agbno;
+	unsigned long			nr_ops = 0;
+	int				shape_changes = 0;
+
+	agno = XFS_FSB_TO_AGNO(mp, startblock);
+	ASSERT(agno != NULLAGNUMBER);
+	bno = XFS_FSB_TO_AGBNO(mp, startblock);
+
+	trace_xfs_refcount_deferred(mp, XFS_FSB_TO_AGNO(mp, startblock),
+			type, XFS_FSB_TO_AGBNO(mp, startblock),
+			blockcount);
+
+	if (XFS_TEST_ERROR(false, mp,
+			XFS_ERRTAG_REFCOUNT_FINISH_ONE,
+			XFS_RANDOM_REFCOUNT_FINISH_ONE))
+		return -EIO;
+
+	/*
+	 * If we haven't gotten a cursor or the cursor AG doesn't match
+	 * the startblock, get one now.
+	 */
+	rcur = *pcur;
+	if (rcur != NULL && rcur->bc_private.a.agno != agno) {
+		nr_ops = rcur->bc_private.a.priv.refc.nr_ops;
+		shape_changes = rcur->bc_private.a.priv.refc.shape_changes;
+		xfs_refcount_finish_one_cleanup(tp, rcur, 0);
+		rcur = NULL;
+		*pcur = NULL;
+	}
+	if (rcur == NULL) {
+		error = xfs_alloc_read_agf(tp->t_mountp, tp, agno,
+				XFS_ALLOC_FLAG_FREEING, &agbp);
+		if (error)
+			return error;
+		if (!agbp)
+			return -EFSCORRUPTED;
+
+		rcur = xfs_refcountbt_init_cursor(mp, tp, agbp, agno, dfops);
+		if (!rcur) {
+			error = -ENOMEM;
+			goto out_cur;
+		}
+		rcur->bc_private.a.priv.refc.nr_ops = nr_ops;
+		rcur->bc_private.a.priv.refc.shape_changes = shape_changes;
+	}
+	*pcur = rcur;
+
+	switch (type) {
+	case XFS_REFCOUNT_INCREASE:
+		error = xfs_refcount_adjust(rcur, bno, blockcount, &new_agbno,
+			new_len, XFS_REFCOUNT_ADJUST_INCREASE, dfops, NULL);
+		*new_fsb = XFS_AGB_TO_FSB(mp, agno, new_agbno);
+		break;
+	case XFS_REFCOUNT_DECREASE:
+		error = xfs_refcount_adjust(rcur, bno, blockcount, &new_agbno,
+			new_len, XFS_REFCOUNT_ADJUST_DECREASE, dfops, NULL);
+		*new_fsb = XFS_AGB_TO_FSB(mp, agno, new_agbno);
+		break;
+	default:
+		ASSERT(0);
+		error = -EFSCORRUPTED;
+	}
+	if (!error && *new_len > 0)
+		trace_xfs_refcount_finish_one_leftover(mp, agno, type,
+				bno, blockcount, new_agbno, *new_len);
+	return error;
+
+out_cur:
+	xfs_trans_brelse(tp, agbp);
+
+	return error;
+}
+
+/*
+ * Record a refcount intent for later processing.
+ */
+static int
+__xfs_refcount_add(
+	struct xfs_mount		*mp,
+	struct xfs_defer_ops		*dfops,
+	enum xfs_refcount_intent_type	type,
+	xfs_fsblock_t			startblock,
+	xfs_extlen_t			blockcount)
+{
+	struct xfs_refcount_intent	*ri;
+
+	trace_xfs_refcount_defer(mp, XFS_FSB_TO_AGNO(mp, startblock),
+			type, XFS_FSB_TO_AGBNO(mp, startblock),
+			blockcount);
+
+	ri = kmem_alloc(sizeof(struct xfs_refcount_intent),
+			KM_SLEEP | KM_NOFS);
+	INIT_LIST_HEAD(&ri->ri_list);
+	ri->ri_type = type;
+	ri->ri_startblock = startblock;
+	ri->ri_blockcount = blockcount;
+
+	xfs_defer_add(dfops, XFS_DEFER_OPS_TYPE_REFCOUNT, &ri->ri_list);
+	return 0;
+}
+
+/*
+ * Increase the reference count of the blocks backing a file's extent.
+ */
+int
+xfs_refcount_increase_extent(
+	struct xfs_mount		*mp,
+	struct xfs_defer_ops		*dfops,
+	struct xfs_bmbt_irec		*PREV)
+{
+	if (!xfs_sb_version_hasreflink(&mp->m_sb))
+		return 0;
+
+	return __xfs_refcount_add(mp, dfops, XFS_REFCOUNT_INCREASE,
+			PREV->br_startblock, PREV->br_blockcount);
+}
+
+/*
+ * Decrease the reference count of the blocks backing a file's extent.
+ */
+int
+xfs_refcount_decrease_extent(
+	struct xfs_mount		*mp,
+	struct xfs_defer_ops		*dfops,
+	struct xfs_bmbt_irec		*PREV)
+{
+	if (!xfs_sb_version_hasreflink(&mp->m_sb))
+		return 0;
+
+	return __xfs_refcount_add(mp, dfops, XFS_REFCOUNT_DECREASE,
+			PREV->br_startblock, PREV->br_blockcount);
+}

fs/xfs/libxfs/xfs_refcount.h

@@ -41,4 +41,17 @@ struct xfs_refcount_intent {
 	xfs_extlen_t				ri_blockcount;
 };
 
+extern int xfs_refcount_increase_extent(struct xfs_mount *mp,
+		struct xfs_defer_ops *dfops, struct xfs_bmbt_irec *irec);
+extern int xfs_refcount_decrease_extent(struct xfs_mount *mp,
+		struct xfs_defer_ops *dfops, struct xfs_bmbt_irec *irec);
+
+extern void xfs_refcount_finish_one_cleanup(struct xfs_trans *tp,
+		struct xfs_btree_cur *rcur, int error);
+extern int xfs_refcount_finish_one(struct xfs_trans *tp,
+		struct xfs_defer_ops *dfops, enum xfs_refcount_intent_type type,
+		xfs_fsblock_t startblock, xfs_extlen_t blockcount,
+		xfs_fsblock_t *new_fsb, xfs_extlen_t *new_len,
+		struct xfs_btree_cur **pcur);
+
 #endif	/* __XFS_REFCOUNT_H__ */

fs/xfs/xfs_error.h

@@ -93,7 +93,8 @@ extern void xfs_verifier_error(struct xfs_buf *bp);
 #define XFS_ERRTAG_FREE_EXTENT				22
 #define XFS_ERRTAG_RMAP_FINISH_ONE			23
 #define XFS_ERRTAG_REFCOUNT_CONTINUE_UPDATE		24
-#define XFS_ERRTAG_MAX					25
+#define XFS_ERRTAG_REFCOUNT_FINISH_ONE			25
+#define XFS_ERRTAG_MAX					26
 
 /*
  * Random factors for above tags, 1 means always, 2 means 1/2 time, etc.
@@ -123,6 +124,7 @@ extern void xfs_verifier_error(struct xfs_buf *bp);
 #define XFS_RANDOM_FREE_EXTENT				1
 #define XFS_RANDOM_RMAP_FINISH_ONE			1
 #define XFS_RANDOM_REFCOUNT_CONTINUE_UPDATE		1
+#define XFS_RANDOM_REFCOUNT_FINISH_ONE			1
 
 #ifdef DEBUG
 extern int xfs_error_test_active;

fs/xfs/xfs_refcount_item.c

@@ -396,9 +396,20 @@ xfs_cui_recover(
 {
 	int				i;
 	int				error = 0;
+	unsigned int			refc_type;
 	struct xfs_phys_extent		*refc;
 	xfs_fsblock_t			startblock_fsb;
 	bool				op_ok;
+	struct xfs_cud_log_item		*cudp;
+	struct xfs_trans		*tp;
+	struct xfs_btree_cur		*rcur = NULL;
+	enum xfs_refcount_intent_type	type;
+	xfs_fsblock_t			firstfsb;
+	xfs_fsblock_t			new_fsb;
+	xfs_extlen_t			new_len;
+	struct xfs_bmbt_irec		irec;
+	struct xfs_defer_ops		dfops;
+	bool				requeue_only = false;
 
 	ASSERT(!test_bit(XFS_CUI_RECOVERED, &cuip->cui_flags));
 
@@ -437,7 +448,80 @@ xfs_cui_recover(
 		}
 	}
 
+	/*
+	 * Under normal operation, refcount updates are deferred, so we
+	 * wouldn't be adding them directly to a transaction.  All
+	 * refcount updates manage reservation usage internally and
+	 * dynamically by deferring work that won't fit in the
+	 * transaction.  Normally, any work that needs to be deferred
+	 * gets attached to the same defer_ops that scheduled the
+	 * refcount update.  However, we're in log recovery here, so we
+	 * we create our own defer_ops and use that to finish up any
+	 * work that doesn't fit.
+	 */
+	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_itruncate, 0, 0, 0, &tp);
+	if (error)
+		return error;
+	cudp = xfs_trans_get_cud(tp, cuip);
+
+	xfs_defer_init(&dfops, &firstfsb);
+	for (i = 0; i < cuip->cui_format.cui_nextents; i++) {
+		refc = &cuip->cui_format.cui_extents[i];
+		refc_type = refc->pe_flags & XFS_REFCOUNT_EXTENT_TYPE_MASK;
+		switch (refc_type) {
+		case XFS_REFCOUNT_INCREASE:
+		case XFS_REFCOUNT_DECREASE:
+		case XFS_REFCOUNT_ALLOC_COW:
+		case XFS_REFCOUNT_FREE_COW:
+			type = refc_type;
+			break;
+		default:
+			error = -EFSCORRUPTED;
+			goto abort_error;
+		}
+		if (requeue_only) {
+			new_fsb = refc->pe_startblock;
+			new_len = refc->pe_len;
+		} else
+			error = xfs_trans_log_finish_refcount_update(tp, cudp,
+				&dfops, type, refc->pe_startblock, refc->pe_len,
+				&new_fsb, &new_len, &rcur);
+		if (error)
+			goto abort_error;
+
+		/* Requeue what we didn't finish. */
+		if (new_len > 0) {
+			irec.br_startblock = new_fsb;
+			irec.br_blockcount = new_len;
+			switch (type) {
+			case XFS_REFCOUNT_INCREASE:
+				error = xfs_refcount_increase_extent(
+						tp->t_mountp, &dfops, &irec);
+				break;
+			case XFS_REFCOUNT_DECREASE:
+				error = xfs_refcount_decrease_extent(
+						tp->t_mountp, &dfops, &irec);
+				break;
+			default:
+				ASSERT(0);
+			}
+			if (error)
+				goto abort_error;
+			requeue_only = true;
+		}
+	}
+
+	xfs_refcount_finish_one_cleanup(tp, rcur, error);
+	error = xfs_defer_finish(&tp, &dfops, NULL);
+	if (error)
+		goto abort_error;
 	set_bit(XFS_CUI_RECOVERED, &cuip->cui_flags);
-	xfs_cui_release(cuip);
+	error = xfs_trans_commit(tp);
+	return error;
+
+abort_error:
+	xfs_refcount_finish_one_cleanup(tp, rcur, error);
+	xfs_defer_cancel(&dfops);
+	xfs_trans_cancel(tp);
 	return error;
 }

fs/xfs/xfs_super.c

@@ -1903,6 +1903,7 @@ init_xfs_fs(void)
 
 	xfs_extent_free_init_defer_op();
 	xfs_rmap_update_init_defer_op();
+	xfs_refcount_update_init_defer_op();
 
 	xfs_dir_startup();
 

fs/xfs/xfs_trace.h

@@ -2931,6 +2931,9 @@ DEFINE_AG_ERROR_EVENT(xfs_refcount_find_right_extent_error);
 DEFINE_AG_EXTENT_EVENT(xfs_refcount_find_shared);
 DEFINE_AG_EXTENT_EVENT(xfs_refcount_find_shared_result);
 DEFINE_AG_ERROR_EVENT(xfs_refcount_find_shared_error);
+#define DEFINE_REFCOUNT_DEFERRED_EVENT DEFINE_PHYS_EXTENT_DEFERRED_EVENT
+DEFINE_REFCOUNT_DEFERRED_EVENT(xfs_refcount_defer);
+DEFINE_REFCOUNT_DEFERRED_EVENT(xfs_refcount_deferred);
 
 TRACE_EVENT(xfs_refcount_finish_one_leftover,
 	TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno,

fs/xfs/xfs_trans.h

@@ -37,6 +37,8 @@ struct xfs_rud_log_item;
 struct xfs_rui_log_item;
 struct xfs_btree_cur;
 struct xfs_cui_log_item;
+struct xfs_cud_log_item;
+struct xfs_defer_ops;
 
 typedef struct xfs_log_item {
 	struct list_head		li_ail;		/* AIL pointers */
@@ -252,11 +254,13 @@ int xfs_trans_log_finish_rmap_update(struct xfs_trans *tp,
 /* refcount updates */
 enum xfs_refcount_intent_type;
 
+void xfs_refcount_update_init_defer_op(void);
 struct xfs_cud_log_item *xfs_trans_get_cud(struct xfs_trans *tp,
 		struct xfs_cui_log_item *cuip);
 int xfs_trans_log_finish_refcount_update(struct xfs_trans *tp,
-		struct xfs_cud_log_item *cudp,
+		struct xfs_cud_log_item *cudp, struct xfs_defer_ops *dfops,
 		enum xfs_refcount_intent_type type, xfs_fsblock_t startblock,
-		xfs_extlen_t blockcount, struct xfs_btree_cur **pcur);
+		xfs_extlen_t blockcount, xfs_fsblock_t *new_fsb,
+		xfs_extlen_t *new_len, struct xfs_btree_cur **pcur);
 
 #endif	/* __XFS_TRANS_H__ */

fs/xfs/xfs_trans_refcount.c

@@ -56,15 +56,18 @@ int
 xfs_trans_log_finish_refcount_update(
 	struct xfs_trans		*tp,
 	struct xfs_cud_log_item		*cudp,
+	struct xfs_defer_ops		*dop,
 	enum xfs_refcount_intent_type	type,
 	xfs_fsblock_t			startblock,
 	xfs_extlen_t			blockcount,
+	xfs_fsblock_t			*new_fsb,
+	xfs_extlen_t			*new_len,
 	struct xfs_btree_cur		**pcur)
 {
 	int				error;
 
-	/* XXX: leave this empty for now */
-	error = -EFSCORRUPTED;
+	error = xfs_refcount_finish_one(tp, dop, type, startblock,
+			blockcount, new_fsb, new_len, pcur);
 
 	/*
 	 * Mark the transaction dirty, even on error. This ensures the
@@ -78,3 +81,184 @@ xfs_trans_log_finish_refcount_update(
 
 	return error;
 }
+
+/* Sort refcount intents by AG. */
+static int
+xfs_refcount_update_diff_items(
+	void				*priv,
+	struct list_head		*a,
+	struct list_head		*b)
+{
+	struct xfs_mount		*mp = priv;
+	struct xfs_refcount_intent	*ra;
+	struct xfs_refcount_intent	*rb;
+
+	ra = container_of(a, struct xfs_refcount_intent, ri_list);
+	rb = container_of(b, struct xfs_refcount_intent, ri_list);
+	return  XFS_FSB_TO_AGNO(mp, ra->ri_startblock) -
+		XFS_FSB_TO_AGNO(mp, rb->ri_startblock);
+}
+
+/* Get an CUI. */
+STATIC void *
+xfs_refcount_update_create_intent(
+	struct xfs_trans		*tp,
+	unsigned int			count)
+{
+	struct xfs_cui_log_item		*cuip;
+
+	ASSERT(tp != NULL);
+	ASSERT(count > 0);
+
+	cuip = xfs_cui_init(tp->t_mountp, count);
+	ASSERT(cuip != NULL);
+
+	/*
+	 * Get a log_item_desc to point at the new item.
+	 */
+	xfs_trans_add_item(tp, &cuip->cui_item);
+	return cuip;
+}
+
+/* Set the phys extent flags for this reverse mapping. */
+static void
+xfs_trans_set_refcount_flags(
+	struct xfs_phys_extent		*refc,
+	enum xfs_refcount_intent_type	type)
+{
+	refc->pe_flags = 0;
+	switch (type) {
+	case XFS_REFCOUNT_INCREASE:
+	case XFS_REFCOUNT_DECREASE:
+	case XFS_REFCOUNT_ALLOC_COW:
+	case XFS_REFCOUNT_FREE_COW:
+		refc->pe_flags |= type;
+		break;
+	default:
+		ASSERT(0);
+	}
+}
+
+/* Log refcount updates in the intent item. */
+STATIC void
+xfs_refcount_update_log_item(
+	struct xfs_trans		*tp,
+	void				*intent,
+	struct list_head		*item)
+{
+	struct xfs_cui_log_item		*cuip = intent;
+	struct xfs_refcount_intent	*refc;
+	uint				next_extent;
+	struct xfs_phys_extent		*ext;
+
+	refc = container_of(item, struct xfs_refcount_intent, ri_list);
+
+	tp->t_flags |= XFS_TRANS_DIRTY;
+	cuip->cui_item.li_desc->lid_flags |= XFS_LID_DIRTY;
+
+	/*
+	 * atomic_inc_return gives us the value after the increment;
+	 * we want to use it as an array index so we need to subtract 1 from
+	 * it.
+	 */
+	next_extent = atomic_inc_return(&cuip->cui_next_extent) - 1;
+	ASSERT(next_extent < cuip->cui_format.cui_nextents);
+	ext = &cuip->cui_format.cui_extents[next_extent];
+	ext->pe_startblock = refc->ri_startblock;
+	ext->pe_len = refc->ri_blockcount;
+	xfs_trans_set_refcount_flags(ext, refc->ri_type);
+}
+
+/* Get an CUD so we can process all the deferred refcount updates. */
+STATIC void *
+xfs_refcount_update_create_done(
+	struct xfs_trans		*tp,
+	void				*intent,
+	unsigned int			count)
+{
+	return xfs_trans_get_cud(tp, intent);
+}
+
+/* Process a deferred refcount update. */
+STATIC int
+xfs_refcount_update_finish_item(
+	struct xfs_trans		*tp,
+	struct xfs_defer_ops		*dop,
+	struct list_head		*item,
+	void				*done_item,
+	void				**state)
+{
+	struct xfs_refcount_intent	*refc;
+	xfs_fsblock_t			new_fsb;
+	xfs_extlen_t			new_aglen;
+	int				error;
+
+	refc = container_of(item, struct xfs_refcount_intent, ri_list);
+	error = xfs_trans_log_finish_refcount_update(tp, done_item, dop,
+			refc->ri_type,
+			refc->ri_startblock,
+			refc->ri_blockcount,
+			&new_fsb, &new_aglen,
+			(struct xfs_btree_cur **)state);
+	/* Did we run out of reservation?  Requeue what we didn't finish. */
+	if (!error && new_aglen > 0) {
+		ASSERT(refc->ri_type == XFS_REFCOUNT_INCREASE ||
+		       refc->ri_type == XFS_REFCOUNT_DECREASE);
+		refc->ri_startblock = new_fsb;
+		refc->ri_blockcount = new_aglen;
+		return -EAGAIN;
+	}
+	kmem_free(refc);
+	return error;
+}
+
+/* Clean up after processing deferred refcounts. */
+STATIC void
+xfs_refcount_update_finish_cleanup(
+	struct xfs_trans	*tp,
+	void			*state,
+	int			error)
+{
+	struct xfs_btree_cur	*rcur = state;
+
+	xfs_refcount_finish_one_cleanup(tp, rcur, error);
+}
+
+/* Abort all pending CUIs. */
+STATIC void
+xfs_refcount_update_abort_intent(
+	void				*intent)
+{
+	xfs_cui_release(intent);
+}
+
+/* Cancel a deferred refcount update. */
+STATIC void
+xfs_refcount_update_cancel_item(
+	struct list_head		*item)
+{
+	struct xfs_refcount_intent	*refc;
+
+	refc = container_of(item, struct xfs_refcount_intent, ri_list);
+	kmem_free(refc);
+}
+
+static const struct xfs_defer_op_type xfs_refcount_update_defer_type = {
+	.type		= XFS_DEFER_OPS_TYPE_REFCOUNT,
+	.max_items	= XFS_CUI_MAX_FAST_EXTENTS,
+	.diff_items	= xfs_refcount_update_diff_items,
+	.create_intent	= xfs_refcount_update_create_intent,
+	.abort_intent	= xfs_refcount_update_abort_intent,
+	.log_item	= xfs_refcount_update_log_item,
+	.create_done	= xfs_refcount_update_create_done,
+	.finish_item	= xfs_refcount_update_finish_item,
+	.finish_cleanup = xfs_refcount_update_finish_cleanup,
+	.cancel_item	= xfs_refcount_update_cancel_item,
+};
+
+/* Register the deferred op type. */
+void
+xfs_refcount_update_init_defer_op(void)
+{
+	xfs_defer_init_op_type(&xfs_refcount_update_defer_type);
+}