xfs: propagate bmap updates to rmapbt

When we map, unmap, or convert an extent in a file's data or attr
fork, schedule a respective update in the rmapbt.  Previous versions
of this patch required a 1:1 correspondence between bmap and rmap,
but this is no longer true as we now have ability to make interval
queries against the rmapbt.

We use the deferred operations code to handle redo operations
atomically and deadlock free.  This plumbs in all five rmap actions
(map, unmap, convert extent, alloc, free); we'll use the first three
now for file data, and reflink will want the last two.  We also add
an error injection site to test log recovery.

Finally, we need to fix the bmap shift extent code to adjust the
rmaps correctly.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

fs/xfs/libxfs/xfs_bmap.c

@@ -2179,6 +2179,11 @@ xfs_bmap_add_extent_delay_real(
 		ASSERT(0);
 	}
 
+	/* add reverse mapping */
+	error = xfs_rmap_map_extent(mp, bma->dfops, bma->ip, whichfork, new);
+	if (error)
+		goto done;
+
 	/* convert to a btree if necessary */
 	if (xfs_bmap_needs_btree(bma->ip, whichfork)) {
 		int	tmp_logflags;	/* partial log flag return val */
@@ -2715,6 +2720,11 @@ xfs_bmap_add_extent_unwritten_real(
 		ASSERT(0);
 	}
 
+	/* update reverse mappings */
+	error = xfs_rmap_convert_extent(mp, dfops, ip, XFS_DATA_FORK, new);
+	if (error)
+		goto done;
+
 	/* convert to a btree if necessary */
 	if (xfs_bmap_needs_btree(ip, XFS_DATA_FORK)) {
 		int	tmp_logflags;	/* partial log flag return val */
@@ -3107,6 +3117,11 @@ xfs_bmap_add_extent_hole_real(
 		break;
 	}
 
+	/* add reverse mapping */
+	error = xfs_rmap_map_extent(mp, bma->dfops, bma->ip, whichfork, new);
+	if (error)
+		goto done;
+
 	/* convert to a btree if necessary */
 	if (xfs_bmap_needs_btree(bma->ip, whichfork)) {
 		int	tmp_logflags;	/* partial log flag return val */
@@ -5034,6 +5049,14 @@ xfs_bmap_del_extent(
 		++*idx;
 		break;
 	}
+
+	/* remove reverse mapping */
+	if (!delay) {
+		error = xfs_rmap_unmap_extent(mp, dfops, ip, whichfork, del);
+		if (error)
+			goto done;
+	}
+
 	/*
 	 * If we need to, add to list of extents to delete.
 	 */
@@ -5573,7 +5596,8 @@ xfs_bmse_shift_one(
 	struct xfs_bmbt_rec_host	*gotp,
 	struct xfs_btree_cur		*cur,
 	int				*logflags,
-	enum shift_direction		direction)
+	enum shift_direction		direction,
+	struct xfs_defer_ops		*dfops)
 {
 	struct xfs_ifork		*ifp;
 	struct xfs_mount		*mp;
@@ -5621,9 +5645,13 @@ xfs_bmse_shift_one(
 		/* check whether to merge the extent or shift it down */
 		if (xfs_bmse_can_merge(&adj_irec, &got,
 				       offset_shift_fsb)) {
-			return xfs_bmse_merge(ip, whichfork, offset_shift_fsb,
+			error = xfs_bmse_merge(ip, whichfork, offset_shift_fsb,
 					       *current_ext, gotp, adj_irecp,
 					       cur, logflags);
+			if (error)
+				return error;
+			adj_irec = got;
+			goto update_rmap;
 		}
 	} else {
 		startoff = got.br_startoff + offset_shift_fsb;
@@ -5660,9 +5688,10 @@ xfs_bmse_shift_one(
 		(*current_ext)--;
 	xfs_bmbt_set_startoff(gotp, startoff);
 	*logflags |= XFS_ILOG_CORE;
+	adj_irec = got;
 	if (!cur) {
 		*logflags |= XFS_ILOG_DEXT;
-		return 0;
+		goto update_rmap;
 	}
 
 	error = xfs_bmbt_lookup_eq(cur, got.br_startoff, got.br_startblock,
@@ -5672,8 +5701,18 @@ xfs_bmse_shift_one(
 	XFS_WANT_CORRUPTED_RETURN(mp, i == 1);
 
 	got.br_startoff = startoff;
-	return xfs_bmbt_update(cur, got.br_startoff, got.br_startblock,
+	error = xfs_bmbt_update(cur, got.br_startoff, got.br_startblock,
 			got.br_blockcount, got.br_state);
+	if (error)
+		return error;
+
+update_rmap:
+	/* update reverse mapping */
+	error = xfs_rmap_unmap_extent(mp, dfops, ip, whichfork, &adj_irec);
+	if (error)
+		return error;
+	adj_irec.br_startoff = startoff;
+	return xfs_rmap_map_extent(mp, dfops, ip, whichfork, &adj_irec);
 }
 
 /*
@@ -5801,7 +5840,7 @@ xfs_bmap_shift_extents(
 	while (nexts++ < num_exts) {
 		error = xfs_bmse_shift_one(ip, whichfork, offset_shift_fsb,
 					   &current_ext, gotp, cur, &logflags,
-					   direction);
+					   direction, dfops);
 		if (error)
 			goto del_cursor;
 		/*

fs/xfs/libxfs/xfs_rmap.c

@@ -36,6 +36,8 @@
 #include "xfs_trace.h"
 #include "xfs_error.h"
 #include "xfs_extent_busy.h"
+#include "xfs_bmap.h"
+#include "xfs_inode.h"
 
 /*
  * Lookup the first record less than or equal to [bno, len, owner, offset]
@@ -1138,3 +1140,260 @@ xfs_rmap_query_range(
 	return xfs_btree_query_range(cur, &low_brec, &high_brec,
 			xfs_rmap_query_range_helper, &query);
 }
+
+/* Clean up after calling xfs_rmap_finish_one. */
+void
+xfs_rmap_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 rmap operations.  We pass back the
+ * btree cursor to maintain our lock on the rmapbt 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_rmap_finish_one(
+	struct xfs_trans		*tp,
+	enum xfs_rmap_intent_type	type,
+	__uint64_t			owner,
+	int				whichfork,
+	xfs_fileoff_t			startoff,
+	xfs_fsblock_t			startblock,
+	xfs_filblks_t			blockcount,
+	xfs_exntst_t			state,
+	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;
+	struct xfs_owner_info		oinfo;
+	xfs_agblock_t			bno;
+	bool				unwritten;
+
+	agno = XFS_FSB_TO_AGNO(mp, startblock);
+	ASSERT(agno != NULLAGNUMBER);
+	bno = XFS_FSB_TO_AGBNO(mp, startblock);
+
+	trace_xfs_rmap_deferred(mp, agno, type, bno, owner, whichfork,
+			startoff, blockcount, state);
+
+	if (XFS_TEST_ERROR(false, mp,
+			XFS_ERRTAG_RMAP_FINISH_ONE,
+			XFS_RANDOM_RMAP_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) {
+		xfs_rmap_finish_one_cleanup(tp, rcur, 0);
+		rcur = NULL;
+		*pcur = NULL;
+	}
+	if (rcur == NULL) {
+		/*
+		 * Refresh the freelist before we start changing the
+		 * rmapbt, because a shape change could cause us to
+		 * allocate blocks.
+		 */
+		error = xfs_free_extent_fix_freelist(tp, agno, &agbp);
+		if (error)
+			return error;
+		if (!agbp)
+			return -EFSCORRUPTED;
+
+		rcur = xfs_rmapbt_init_cursor(mp, tp, agbp, agno);
+		if (!rcur) {
+			error = -ENOMEM;
+			goto out_cur;
+		}
+	}
+	*pcur = rcur;
+
+	xfs_rmap_ino_owner(&oinfo, owner, whichfork, startoff);
+	unwritten = state == XFS_EXT_UNWRITTEN;
+	bno = XFS_FSB_TO_AGBNO(rcur->bc_mp, startblock);
+
+	switch (type) {
+	case XFS_RMAP_ALLOC:
+	case XFS_RMAP_MAP:
+		error = xfs_rmap_map(rcur, bno, blockcount, unwritten, &oinfo);
+		break;
+	case XFS_RMAP_FREE:
+	case XFS_RMAP_UNMAP:
+		error = xfs_rmap_unmap(rcur, bno, blockcount, unwritten,
+				&oinfo);
+		break;
+	case XFS_RMAP_CONVERT:
+		error = xfs_rmap_convert(rcur, bno, blockcount, !unwritten,
+				&oinfo);
+		break;
+	default:
+		ASSERT(0);
+		error = -EFSCORRUPTED;
+	}
+	return error;
+
+out_cur:
+	xfs_trans_brelse(tp, agbp);
+
+	return error;
+}
+
+/*
+ * Don't defer an rmap if we aren't an rmap filesystem.
+ */
+static bool
+xfs_rmap_update_is_needed(
+	struct xfs_mount	*mp)
+{
+	return xfs_sb_version_hasrmapbt(&mp->m_sb);
+}
+
+/*
+ * Record a rmap intent; the list is kept sorted first by AG and then by
+ * increasing age.
+ */
+static int
+__xfs_rmap_add(
+	struct xfs_mount		*mp,
+	struct xfs_defer_ops		*dfops,
+	enum xfs_rmap_intent_type	type,
+	__uint64_t			owner,
+	int				whichfork,
+	struct xfs_bmbt_irec		*bmap)
+{
+	struct xfs_rmap_intent	*ri;
+
+	trace_xfs_rmap_defer(mp, XFS_FSB_TO_AGNO(mp, bmap->br_startblock),
+			type,
+			XFS_FSB_TO_AGBNO(mp, bmap->br_startblock),
+			owner, whichfork,
+			bmap->br_startoff,
+			bmap->br_blockcount,
+			bmap->br_state);
+
+	ri = kmem_alloc(sizeof(struct xfs_rmap_intent), KM_SLEEP | KM_NOFS);
+	INIT_LIST_HEAD(&ri->ri_list);
+	ri->ri_type = type;
+	ri->ri_owner = owner;
+	ri->ri_whichfork = whichfork;
+	ri->ri_bmap = *bmap;
+
+	xfs_defer_add(dfops, XFS_DEFER_OPS_TYPE_RMAP, &ri->ri_list);
+	return 0;
+}
+
+/* Map an extent into a file. */
+int
+xfs_rmap_map_extent(
+	struct xfs_mount	*mp,
+	struct xfs_defer_ops	*dfops,
+	struct xfs_inode	*ip,
+	int			whichfork,
+	struct xfs_bmbt_irec	*PREV)
+{
+	if (!xfs_rmap_update_is_needed(mp))
+		return 0;
+
+	return __xfs_rmap_add(mp, dfops, XFS_RMAP_MAP, ip->i_ino,
+			whichfork, PREV);
+}
+
+/* Unmap an extent out of a file. */
+int
+xfs_rmap_unmap_extent(
+	struct xfs_mount	*mp,
+	struct xfs_defer_ops	*dfops,
+	struct xfs_inode	*ip,
+	int			whichfork,
+	struct xfs_bmbt_irec	*PREV)
+{
+	if (!xfs_rmap_update_is_needed(mp))
+		return 0;
+
+	return __xfs_rmap_add(mp, dfops, XFS_RMAP_UNMAP, ip->i_ino,
+			whichfork, PREV);
+}
+
+/* Convert a data fork extent from unwritten to real or vice versa. */
+int
+xfs_rmap_convert_extent(
+	struct xfs_mount	*mp,
+	struct xfs_defer_ops	*dfops,
+	struct xfs_inode	*ip,
+	int			whichfork,
+	struct xfs_bmbt_irec	*PREV)
+{
+	if (!xfs_rmap_update_is_needed(mp))
+		return 0;
+
+	return __xfs_rmap_add(mp, dfops, XFS_RMAP_CONVERT, ip->i_ino,
+			whichfork, PREV);
+}
+
+/* Schedule the creation of an rmap for non-file data. */
+int
+xfs_rmap_alloc_extent(
+	struct xfs_mount	*mp,
+	struct xfs_defer_ops	*dfops,
+	xfs_agnumber_t		agno,
+	xfs_agblock_t		bno,
+	xfs_extlen_t		len,
+	__uint64_t		owner)
+{
+	struct xfs_bmbt_irec	bmap;
+
+	if (!xfs_rmap_update_is_needed(mp))
+		return 0;
+
+	bmap.br_startblock = XFS_AGB_TO_FSB(mp, agno, bno);
+	bmap.br_blockcount = len;
+	bmap.br_startoff = 0;
+	bmap.br_state = XFS_EXT_NORM;
+
+	return __xfs_rmap_add(mp, dfops, XFS_RMAP_ALLOC, owner,
+			XFS_DATA_FORK, &bmap);
+}
+
+/* Schedule the deletion of an rmap for non-file data. */
+int
+xfs_rmap_free_extent(
+	struct xfs_mount	*mp,
+	struct xfs_defer_ops	*dfops,
+	xfs_agnumber_t		agno,
+	xfs_agblock_t		bno,
+	xfs_extlen_t		len,
+	__uint64_t		owner)
+{
+	struct xfs_bmbt_irec	bmap;
+
+	if (!xfs_rmap_update_is_needed(mp))
+		return 0;
+
+	bmap.br_startblock = XFS_AGB_TO_FSB(mp, agno, bno);
+	bmap.br_blockcount = len;
+	bmap.br_startoff = 0;
+	bmap.br_state = XFS_EXT_NORM;
+
+	return __xfs_rmap_add(mp, dfops, XFS_RMAP_FREE, owner,
+			XFS_DATA_FORK, &bmap);
+}

fs/xfs/libxfs/xfs_rmap.h

@@ -182,4 +182,28 @@ struct xfs_rmap_intent {
 	struct xfs_bmbt_irec			ri_bmap;
 };
 
+/* functions for updating the rmapbt based on bmbt map/unmap operations */
+int xfs_rmap_map_extent(struct xfs_mount *mp, struct xfs_defer_ops *dfops,
+		struct xfs_inode *ip, int whichfork,
+		struct xfs_bmbt_irec *imap);
+int xfs_rmap_unmap_extent(struct xfs_mount *mp, struct xfs_defer_ops *dfops,
+		struct xfs_inode *ip, int whichfork,
+		struct xfs_bmbt_irec *imap);
+int xfs_rmap_convert_extent(struct xfs_mount *mp, struct xfs_defer_ops *dfops,
+		struct xfs_inode *ip, int whichfork,
+		struct xfs_bmbt_irec *imap);
+int xfs_rmap_alloc_extent(struct xfs_mount *mp, struct xfs_defer_ops *dfops,
+		xfs_agnumber_t agno, xfs_agblock_t bno, xfs_extlen_t len,
+		__uint64_t owner);
+int xfs_rmap_free_extent(struct xfs_mount *mp, struct xfs_defer_ops *dfops,
+		xfs_agnumber_t agno, xfs_agblock_t bno, xfs_extlen_t len,
+		__uint64_t owner);
+
+void xfs_rmap_finish_one_cleanup(struct xfs_trans *tp,
+		struct xfs_btree_cur *rcur, int error);
+int xfs_rmap_finish_one(struct xfs_trans *tp, enum xfs_rmap_intent_type type,
+		__uint64_t owner, int whichfork, xfs_fileoff_t startoff,
+		xfs_fsblock_t startblock, xfs_filblks_t blockcount,
+		xfs_exntst_t state, struct xfs_btree_cur **pcur);
+
 #endif	/* __XFS_RMAP_H__ */

fs/xfs/xfs_bmap_util.c

@@ -41,6 +41,7 @@
 #include "xfs_trace.h"
 #include "xfs_icache.h"
 #include "xfs_log.h"
+#include "xfs_rmap_btree.h"
 
 /* Kernel only BMAP related definitions and functions */
 

fs/xfs/xfs_error.h

@@ -91,7 +91,8 @@ extern void xfs_verifier_error(struct xfs_buf *bp);
 #define XFS_ERRTAG_DIOWRITE_IOERR			20
 #define XFS_ERRTAG_BMAPIFORMAT				21
 #define XFS_ERRTAG_FREE_EXTENT				22
-#define XFS_ERRTAG_MAX					23
+#define XFS_ERRTAG_RMAP_FINISH_ONE			23
+#define XFS_ERRTAG_MAX					24
 
 /*
  * Random factors for above tags, 1 means always, 2 means 1/2 time, etc.
@@ -119,6 +120,7 @@ extern void xfs_verifier_error(struct xfs_buf *bp);
 #define XFS_RANDOM_DIOWRITE_IOERR			(XFS_RANDOM_DEFAULT/10)
 #define	XFS_RANDOM_BMAPIFORMAT				XFS_RANDOM_DEFAULT
 #define XFS_RANDOM_FREE_EXTENT				1
+#define XFS_RANDOM_RMAP_FINISH_ONE			1
 
 #ifdef DEBUG
 extern int xfs_error_test_active;

fs/xfs/xfs_rmap_item.c

@@ -24,11 +24,13 @@
 #include "xfs_trans_resv.h"
 #include "xfs_bit.h"
 #include "xfs_mount.h"
+#include "xfs_defer.h"
 #include "xfs_trans.h"
 #include "xfs_trans_priv.h"
 #include "xfs_buf_item.h"
 #include "xfs_rmap_item.h"
 #include "xfs_log.h"
+#include "xfs_rmap.h"
 
 
 kmem_zone_t	*xfs_rui_zone;
@@ -473,6 +475,12 @@ xfs_rui_recover(
 	struct xfs_map_extent		*rmap;
 	xfs_fsblock_t			startblock_fsb;
 	bool				op_ok;
+	struct xfs_rud_log_item		*rudp;
+	enum xfs_rmap_intent_type	type;
+	int				whichfork;
+	xfs_exntst_t			state;
+	struct xfs_trans		*tp;
+	struct xfs_btree_cur		*rcur = NULL;
 
 	ASSERT(!test_bit(XFS_RUI_RECOVERED, &ruip->rui_flags));
 
@@ -512,8 +520,53 @@ xfs_rui_recover(
 		}
 	}
 
-	/* XXX: do nothing for now */
+	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_itruncate, 0, 0, 0, &tp);
+	if (error)
+		return error;
+	rudp = xfs_trans_get_rud(tp, ruip, ruip->rui_format.rui_nextents);
+
+	for (i = 0; i < ruip->rui_format.rui_nextents; i++) {
+		rmap = &(ruip->rui_format.rui_extents[i]);
+		state = (rmap->me_flags & XFS_RMAP_EXTENT_UNWRITTEN) ?
+				XFS_EXT_UNWRITTEN : XFS_EXT_NORM;
+		whichfork = (rmap->me_flags & XFS_RMAP_EXTENT_ATTR_FORK) ?
+				XFS_ATTR_FORK : XFS_DATA_FORK;
+		switch (rmap->me_flags & XFS_RMAP_EXTENT_TYPE_MASK) {
+		case XFS_RMAP_EXTENT_MAP:
+			type = XFS_RMAP_MAP;
+			break;
+		case XFS_RMAP_EXTENT_UNMAP:
+			type = XFS_RMAP_UNMAP;
+			break;
+		case XFS_RMAP_EXTENT_CONVERT:
+			type = XFS_RMAP_CONVERT;
+			break;
+		case XFS_RMAP_EXTENT_ALLOC:
+			type = XFS_RMAP_ALLOC;
+			break;
+		case XFS_RMAP_EXTENT_FREE:
+			type = XFS_RMAP_FREE;
+			break;
+		default:
+			error = -EFSCORRUPTED;
+			goto abort_error;
+		}
+		error = xfs_trans_log_finish_rmap_update(tp, rudp, type,
+				rmap->me_owner, whichfork,
+				rmap->me_startoff, rmap->me_startblock,
+				rmap->me_len, state, &rcur);
+		if (error)
+			goto abort_error;
+
+	}
+
+	xfs_rmap_finish_one_cleanup(tp, rcur, error);
 	set_bit(XFS_RUI_RECOVERED, &ruip->rui_flags);
-	xfs_rui_release(ruip);
+	error = xfs_trans_commit(tp);
+	return error;
+
+abort_error:
+	xfs_rmap_finish_one_cleanup(tp, rcur, error);
+	xfs_trans_cancel(tp);
 	return error;
 }

fs/xfs/xfs_trans.h

@@ -35,6 +35,7 @@ struct xfs_dquot_acct;
 struct xfs_busy_extent;
 struct xfs_rud_log_item;
 struct xfs_rui_log_item;
+struct xfs_btree_cur;
 
 typedef struct xfs_log_item {
 	struct list_head		li_ail;		/* AIL pointers */
@@ -245,6 +246,6 @@ int xfs_trans_log_finish_rmap_update(struct xfs_trans *tp,
 		struct xfs_rud_log_item *rudp, enum xfs_rmap_intent_type type,
 		__uint64_t owner, int whichfork, xfs_fileoff_t startoff,
 		xfs_fsblock_t startblock, xfs_filblks_t blockcount,
-		xfs_exntst_t state);
+		xfs_exntst_t state, struct xfs_btree_cur **pcur);
 
 #endif	/* __XFS_TRANS_H__ */

fs/xfs/xfs_trans_rmap.c

@@ -171,14 +171,15 @@ xfs_trans_log_finish_rmap_update(
 	xfs_fileoff_t			startoff,
 	xfs_fsblock_t			startblock,
 	xfs_filblks_t			blockcount,
-	xfs_exntst_t			state)
+	xfs_exntst_t			state,
+	struct xfs_btree_cur		**pcur)
 {
 	uint				next_extent;
 	struct xfs_map_extent		*rmap;
 	int				error;
 
-	/* XXX: actually finish the rmap update here */
-	error = -EFSCORRUPTED;
+	error = xfs_rmap_finish_one(tp, type, owner, whichfork, startoff,
+			startblock, blockcount, state, pcur);
 
 	/*
 	 * Mark the transaction dirty, even on error. This ensures the
@@ -276,7 +277,8 @@ xfs_rmap_update_finish_item(
 			rmap->ri_bmap.br_startoff,
 			rmap->ri_bmap.br_startblock,
 			rmap->ri_bmap.br_blockcount,
-			rmap->ri_bmap.br_state);
+			rmap->ri_bmap.br_state,
+			(struct xfs_btree_cur **)state);
 	kmem_free(rmap);
 	return error;
 }
@@ -288,6 +290,9 @@ xfs_rmap_update_finish_cleanup(
 	void			*state,
 	int			error)
 {
+	struct xfs_btree_cur	*rcur = state;
+
+	xfs_rmap_finish_one_cleanup(tp, rcur, error);
 }
 
 /* Abort all pending RUIs. */