xfs: implement deferred bmbt map/unmap operations

Implement deferred versions of the inode block map/unmap functions.
These will be used in subsequent patches to make reflink operations
atomic.

Signed-off-by: Darrick J. Wong <[email protected]>
Reviewed-by: Christoph Hellwig <[email protected]>

fs/xfs/libxfs/xfs_bmap.c

@@ -6062,3 +6062,146 @@ xfs_bmap_split_extent(
 	xfs_trans_cancel(tp);
 	return error;
 }
+
+/* Deferred mapping is only for real extents in the data fork. */
+static bool
+xfs_bmap_is_update_needed(
+	struct xfs_bmbt_irec	*bmap)
+{
+	return  bmap->br_startblock != HOLESTARTBLOCK &&
+		bmap->br_startblock != DELAYSTARTBLOCK;
+}
+
+/* Record a bmap intent. */
+static int
+__xfs_bmap_add(
+	struct xfs_mount		*mp,
+	struct xfs_defer_ops		*dfops,
+	enum xfs_bmap_intent_type	type,
+	struct xfs_inode		*ip,
+	int				whichfork,
+	struct xfs_bmbt_irec		*bmap)
+{
+	int				error;
+	struct xfs_bmap_intent		*bi;
+
+	trace_xfs_bmap_defer(mp,
+			XFS_FSB_TO_AGNO(mp, bmap->br_startblock),
+			type,
+			XFS_FSB_TO_AGBNO(mp, bmap->br_startblock),
+			ip->i_ino, whichfork,
+			bmap->br_startoff,
+			bmap->br_blockcount,
+			bmap->br_state);
+
+	bi = kmem_alloc(sizeof(struct xfs_bmap_intent), KM_SLEEP | KM_NOFS);
+	INIT_LIST_HEAD(&bi->bi_list);
+	bi->bi_type = type;
+	bi->bi_owner = ip;
+	bi->bi_whichfork = whichfork;
+	bi->bi_bmap = *bmap;
+
+	error = xfs_defer_join(dfops, bi->bi_owner);
+	if (error) {
+		kmem_free(bi);
+		return error;
+	}
+
+	xfs_defer_add(dfops, XFS_DEFER_OPS_TYPE_BMAP, &bi->bi_list);
+	return 0;
+}
+
+/* Map an extent into a file. */
+int
+xfs_bmap_map_extent(
+	struct xfs_mount	*mp,
+	struct xfs_defer_ops	*dfops,
+	struct xfs_inode	*ip,
+	struct xfs_bmbt_irec	*PREV)
+{
+	if (!xfs_bmap_is_update_needed(PREV))
+		return 0;
+
+	return __xfs_bmap_add(mp, dfops, XFS_BMAP_MAP, ip,
+			XFS_DATA_FORK, PREV);
+}
+
+/* Unmap an extent out of a file. */
+int
+xfs_bmap_unmap_extent(
+	struct xfs_mount	*mp,
+	struct xfs_defer_ops	*dfops,
+	struct xfs_inode	*ip,
+	struct xfs_bmbt_irec	*PREV)
+{
+	if (!xfs_bmap_is_update_needed(PREV))
+		return 0;
+
+	return __xfs_bmap_add(mp, dfops, XFS_BMAP_UNMAP, ip,
+			XFS_DATA_FORK, PREV);
+}
+
+/*
+ * Process one of the deferred bmap operations.  We pass back the
+ * btree cursor to maintain our lock on the bmapbt between calls.
+ */
+int
+xfs_bmap_finish_one(
+	struct xfs_trans		*tp,
+	struct xfs_defer_ops		*dfops,
+	struct xfs_inode		*ip,
+	enum xfs_bmap_intent_type	type,
+	int				whichfork,
+	xfs_fileoff_t			startoff,
+	xfs_fsblock_t			startblock,
+	xfs_filblks_t			blockcount,
+	xfs_exntst_t			state)
+{
+	struct xfs_bmbt_irec		bmap;
+	int				nimaps = 1;
+	xfs_fsblock_t			firstfsb;
+	int				flags = XFS_BMAPI_REMAP;
+	int				done;
+	int				error = 0;
+
+	bmap.br_startblock = startblock;
+	bmap.br_startoff = startoff;
+	bmap.br_blockcount = blockcount;
+	bmap.br_state = state;
+
+	trace_xfs_bmap_deferred(tp->t_mountp,
+			XFS_FSB_TO_AGNO(tp->t_mountp, startblock), type,
+			XFS_FSB_TO_AGBNO(tp->t_mountp, startblock),
+			ip->i_ino, whichfork, startoff, blockcount, state);
+
+	if (whichfork != XFS_DATA_FORK && whichfork != XFS_ATTR_FORK)
+		return -EFSCORRUPTED;
+	if (whichfork == XFS_ATTR_FORK)
+		flags |= XFS_BMAPI_ATTRFORK;
+
+	if (XFS_TEST_ERROR(false, tp->t_mountp,
+			XFS_ERRTAG_BMAP_FINISH_ONE,
+			XFS_RANDOM_BMAP_FINISH_ONE))
+		return -EIO;
+
+	switch (type) {
+	case XFS_BMAP_MAP:
+		firstfsb = bmap.br_startblock;
+		error = xfs_bmapi_write(tp, ip, bmap.br_startoff,
+					bmap.br_blockcount, flags, &firstfsb,
+					bmap.br_blockcount, &bmap, &nimaps,
+					dfops);
+		break;
+	case XFS_BMAP_UNMAP:
+		error = xfs_bunmapi(tp, ip, bmap.br_startoff,
+				bmap.br_blockcount, flags, 1, &firstfsb,
+				dfops, &done);
+		ASSERT(done);
+		break;
+	default:
+		ASSERT(0);
+		error = -EFSCORRUPTED;
+	}
+
+	return error;
+}

fs/xfs/libxfs/xfs_bmap.h

@@ -231,4 +231,13 @@ struct xfs_bmap_intent {
 	struct xfs_bmbt_irec			bi_bmap;
 };
 
+int	xfs_bmap_finish_one(struct xfs_trans *tp, struct xfs_defer_ops *dfops,
+		struct xfs_inode *ip, enum xfs_bmap_intent_type type,
+		int whichfork, xfs_fileoff_t startoff, xfs_fsblock_t startblock,
+		xfs_filblks_t blockcount, xfs_exntst_t state);
+int	xfs_bmap_map_extent(struct xfs_mount *mp, struct xfs_defer_ops *dfops,
+		struct xfs_inode *ip, struct xfs_bmbt_irec *imap);
+int	xfs_bmap_unmap_extent(struct xfs_mount *mp, struct xfs_defer_ops *dfops,
+		struct xfs_inode *ip, struct xfs_bmbt_irec *imap);
+
 #endif	/* __XFS_BMAP_H__ */

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_BMAP,
 	XFS_DEFER_OPS_TYPE_REFCOUNT,
 	XFS_DEFER_OPS_TYPE_RMAP,
 	XFS_DEFER_OPS_TYPE_FREE,

fs/xfs/xfs_bmap_item.c

@@ -389,10 +389,19 @@ xfs_bui_recover(
 	struct xfs_bui_log_item		*buip)
 {
 	int				error = 0;
+	unsigned int			bui_type;
 	struct xfs_map_extent		*bmap;
 	xfs_fsblock_t			startblock_fsb;
 	xfs_fsblock_t			inode_fsb;
 	bool				op_ok;
+	struct xfs_bud_log_item		*budp;
+	enum xfs_bmap_intent_type	type;
+	int				whichfork;
+	xfs_exntst_t			state;
+	struct xfs_trans		*tp;
+	struct xfs_inode		*ip = NULL;
+	struct xfs_defer_ops		dfops;
+	xfs_fsblock_t			firstfsb;
 
 	ASSERT(!test_bit(XFS_BUI_RECOVERED, &buip->bui_flags));
 
@@ -437,7 +446,61 @@ xfs_bui_recover(
 		return -EIO;
 	}
 
+	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_itruncate, 0, 0, 0, &tp);
+	if (error)
+		return error;
+	budp = xfs_trans_get_bud(tp, buip);
+
+	/* Grab the inode. */
+	error = xfs_iget(mp, tp, bmap->me_owner, 0, XFS_ILOCK_EXCL, &ip);
+	if (error)
+		goto err_inode;
+
+	xfs_defer_init(&dfops, &firstfsb);
+
+	/* Process deferred bmap item. */
+	state = (bmap->me_flags & XFS_BMAP_EXTENT_UNWRITTEN) ?
+			XFS_EXT_UNWRITTEN : XFS_EXT_NORM;
+	whichfork = (bmap->me_flags & XFS_BMAP_EXTENT_ATTR_FORK) ?
+			XFS_ATTR_FORK : XFS_DATA_FORK;
+	bui_type = bmap->me_flags & XFS_BMAP_EXTENT_TYPE_MASK;
+	switch (bui_type) {
+	case XFS_BMAP_MAP:
+	case XFS_BMAP_UNMAP:
+		type = bui_type;
+		break;
+	default:
+		error = -EFSCORRUPTED;
+		goto err_dfops;
+	}
+	xfs_trans_ijoin(tp, ip, 0);
+
+	error = xfs_trans_log_finish_bmap_update(tp, budp, &dfops, type,
+			ip, whichfork, bmap->me_startoff,
+			bmap->me_startblock, bmap->me_len,
+			state);
+	if (error)
+		goto err_dfops;
+
+	/* Finish transaction, free inodes. */
+	error = xfs_defer_finish(&tp, &dfops, NULL);
+	if (error)
+		goto err_dfops;
+
 	set_bit(XFS_BUI_RECOVERED, &buip->bui_flags);
-	xfs_bui_release(buip);
+	error = xfs_trans_commit(tp);
+	xfs_iunlock(ip, XFS_ILOCK_EXCL);
+	IRELE(ip);
+
+	return error;
+
+err_dfops:
+	xfs_defer_cancel(&dfops);
+err_inode:
+	xfs_trans_cancel(tp);
+	if (ip) {
+		xfs_iunlock(ip, XFS_ILOCK_EXCL);
+		IRELE(ip);
+	}
 	return error;
 }

fs/xfs/xfs_error.h

@@ -94,7 +94,8 @@ extern void xfs_verifier_error(struct xfs_buf *bp);
 #define XFS_ERRTAG_RMAP_FINISH_ONE			23
 #define XFS_ERRTAG_REFCOUNT_CONTINUE_UPDATE		24
 #define XFS_ERRTAG_REFCOUNT_FINISH_ONE			25
-#define XFS_ERRTAG_MAX					26
+#define XFS_ERRTAG_BMAP_FINISH_ONE			26
+#define XFS_ERRTAG_MAX					27
 
 /*
  * Random factors for above tags, 1 means always, 2 means 1/2 time, etc.
@@ -125,6 +126,7 @@ extern void xfs_verifier_error(struct xfs_buf *bp);
 #define XFS_RANDOM_RMAP_FINISH_ONE			1
 #define XFS_RANDOM_REFCOUNT_CONTINUE_UPDATE		1
 #define XFS_RANDOM_REFCOUNT_FINISH_ONE			1
+#define XFS_RANDOM_BMAP_FINISH_ONE			1
 
 #ifdef DEBUG
 extern int xfs_error_test_active;

fs/xfs/xfs_super.c

@@ -1922,6 +1922,7 @@ init_xfs_fs(void)
 	xfs_extent_free_init_defer_op();
 	xfs_rmap_update_init_defer_op();
 	xfs_refcount_update_init_defer_op();
+	xfs_bmap_update_init_defer_op();
 
 	xfs_dir_startup();
 

fs/xfs/xfs_trace.h

@@ -2587,6 +2587,11 @@ DEFINE_RMAPBT_EVENT(xfs_rmap_lookup_le_range_result);
 DEFINE_RMAPBT_EVENT(xfs_rmap_find_right_neighbor_result);
 DEFINE_RMAPBT_EVENT(xfs_rmap_find_left_neighbor_result);
 
+/* deferred bmbt updates */
+#define DEFINE_BMAP_DEFERRED_EVENT	DEFINE_RMAP_DEFERRED_EVENT
+DEFINE_BMAP_DEFERRED_EVENT(xfs_bmap_defer);
+DEFINE_BMAP_DEFERRED_EVENT(xfs_bmap_deferred);
+
 /* per-AG reservation */
 DECLARE_EVENT_CLASS(xfs_ag_resv_class,
 	TP_PROTO(struct xfs_perag *pag, enum xfs_ag_resv_type resv,

fs/xfs/xfs_trans.h

@@ -40,6 +40,7 @@ struct xfs_cui_log_item;
 struct xfs_cud_log_item;
 struct xfs_defer_ops;
 struct xfs_bui_log_item;
+struct xfs_bud_log_item;
 
 typedef struct xfs_log_item {
 	struct list_head		li_ail;		/* AIL pointers */