[PATCH v8 00/71] xfsprogs: add reflink and dedupe support

[PATCH v8 00/71] xfsprogs: add reflink and dedupe support

[Darrick J. Wong]

Hi all,

This is the eighth revision of a patchset that adds to XFS kernel
support for mapping multiple file logical blocks to the same physical
block (reflink/deduplication), implements the beginnings of online
metadata scrubbing and preening, and implements reverse mapping for
the realtime device.  There shouldn't be any incompatible on-disk
format changes, pending a thorough review of the patches within.

(NOTE: In the git trees, this series is preceded by the pending rmap
fixes patches posted to linux-xfs a few days ago.)

At the beginning of this series are imports of all the kernel patches
against libxfs/ that are needed to provide refcount btrees and reflink
support.  Please note that these patches /do/ have diffs against files
in the libxfs/ directory (mostly for deferred ops) that do not start
with 'xfs_', so the libxfs-apply script cannot be used in isolation.

After that are the various small changes required for db, growfs, io,
logprint, and manpages to support the new reference counting data
structures.  The mkfs patch is, as usual, at the end.

Before that, however, are a bunch of patches to xfs_repair that
spot-check the refcount btree and use the rmap tracking code
introduced in xfsprogs 4.8 to regenerate the refcount btree.  It also
fixes discrepancies in the inode reflink flag.

If you're going to start using this mess, you probably ought to just
pull from my github trees for kernel[1], xfsprogs[2], xfstests[3],
xfs-docs[4], and man-pages[5].  The kernel patches in the git trees
should apply to 4.8-rc3; xfsprogs patches to for-next; and xfstest to
master.

The patches have been xfstested with x64, ppc64, and armhf; all tests
in the clone and rmap groups pass.  AFAICT they don't cause any new
failures for the 'auto' group.

This is an extraordinary way to eat your data.  Enjoy! 
Comments and questions are, as always, welcome.

--D

[1] https://github.com/djwong/linux/tree/djwong-devel
[2] https://github.com/djwong/xfsprogs/tree/djwong-devel
[3] https://github.com/djwong/xfstests/tree/djwong-devel
[4] https://github.com/djwong/xfs-documentation/tree/djwong-devel
[5] https://github.com/djwong/man-pages/tree/djwong-devel

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 01/71] xfs: remove xfs_btree_bigkey

[Darrick J. Wong]

Remove the xfs_btree_bigkey mess and simply make xfs_btree_key big enough
to hold both keys in-core.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_btree.c |   12 ++++++------
 libxfs/xfs_btree.h |   24 ++++++------------------
 2 files changed, 12 insertions(+), 24 deletions(-)


diff --git a/libxfs/xfs_btree.c b/libxfs/xfs_btree.c
index 27db736..99bf808 100644
--- a/libxfs/xfs_btree.c
+++ b/libxfs/xfs_btree.c
@@ -2066,7 +2066,7 @@ __xfs_btree_updkeys(
 	struct xfs_buf		*bp0,
 	bool			force_all)
 {
-	union xfs_btree_bigkey	key;	/* keys from current level */
+	union xfs_btree_key	key;	/* keys from current level */
 	union xfs_btree_key	*lkey;	/* keys from the next level up */
 	union xfs_btree_key	*hkey;
 	union xfs_btree_key	*nlkey;	/* keys from the next level up */
@@ -2082,7 +2082,7 @@ __xfs_btree_updkeys(
 
 	trace_xfs_btree_updkeys(cur, level, bp0);
 
-	lkey = (union xfs_btree_key *)&key;
+	lkey = &key;
 	hkey = xfs_btree_high_key_from_key(cur, lkey);
 	xfs_btree_get_keys(cur, block, lkey);
 	for (level++; level < cur->bc_nlevels; level++) {
@@ -3226,7 +3226,7 @@ xfs_btree_insrec(
 	struct xfs_buf		*bp;	/* buffer for block */
 	union xfs_btree_ptr	nptr;	/* new block ptr */
 	struct xfs_btree_cur	*ncur;	/* new btree cursor */
-	union xfs_btree_bigkey	nkey;	/* new block key */
+	union xfs_btree_key	nkey;	/* new block key */
 	union xfs_btree_key	*lkey;
 	int			optr;	/* old key/record index */
 	int			ptr;	/* key/record index */
@@ -3241,7 +3241,7 @@ xfs_btree_insrec(
 	XFS_BTREE_TRACE_ARGIPR(cur, level, *ptrp, &rec);
 
 	ncur = NULL;
-	lkey = (union xfs_btree_key *)&nkey;
+	lkey = &nkey;
 
 	/*
 	 * If we have an external root pointer, and we've made it to the
@@ -3444,14 +3444,14 @@ xfs_btree_insert(
 	union xfs_btree_ptr	nptr;	/* new block number (split result) */
 	struct xfs_btree_cur	*ncur;	/* new cursor (split result) */
 	struct xfs_btree_cur	*pcur;	/* previous level's cursor */
-	union xfs_btree_bigkey	bkey;	/* key of block to insert */
+	union xfs_btree_key	bkey;	/* key of block to insert */
 	union xfs_btree_key	*key;
 	union xfs_btree_rec	rec;	/* record to insert */
 
 	level = 0;
 	ncur = NULL;
 	pcur = cur;
-	key = (union xfs_btree_key *)&bkey;
+	key = &bkey;
 
 	xfs_btree_set_ptr_null(cur, &nptr);
 
diff --git a/libxfs/xfs_btree.h b/libxfs/xfs_btree.h
index 72f8b3c..870a5ab 100644
--- a/libxfs/xfs_btree.h
+++ b/libxfs/xfs_btree.h
@@ -37,30 +37,18 @@ union xfs_btree_ptr {
 	__be64			l;	/* long form ptr */
 };
 
-union xfs_btree_key {
-	struct xfs_bmbt_key		bmbt;
-	xfs_bmdr_key_t			bmbr;	/* bmbt root block */
-	xfs_alloc_key_t			alloc;
-	struct xfs_inobt_key		inobt;
-	struct xfs_rmap_key		rmap;
-};
-
 /*
- * In-core key that holds both low and high keys for overlapped btrees.
- * The two keys are packed next to each other on disk, so do the same
- * in memory.  Preserve the existing xfs_btree_key as a single key to
- * avoid the mental model breakage that would happen if we passed a
- * bigkey into a function that operates on a single key.
+ * The in-core btree key.  Overlapping btrees actually store two keys
+ * per pointer, so we reserve enough memory to hold both.  The __*bigkey
+ * items should never be accessed directly.
  */
-union xfs_btree_bigkey {
+union xfs_btree_key {
 	struct xfs_bmbt_key		bmbt;
 	xfs_bmdr_key_t			bmbr;	/* bmbt root block */
 	xfs_alloc_key_t			alloc;
 	struct xfs_inobt_key		inobt;
-	struct {
-		struct xfs_rmap_key	rmap;
-		struct xfs_rmap_key	rmap_hi;
-	};
+	struct xfs_rmap_key		rmap;
+	struct xfs_rmap_key		__rmap_bigkey[2];
 };
 
 union xfs_btree_rec {

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 02/71] xfs: create a standard btree size calculator code

[Darrick J. Wong]

Create a helper to generate AG btree height calculator functions.
This will be used (much) later when we get to the refcount btree.

v2: Use a helper function instead of a macro.
v3: We can (theoretically) store more than 2^32 records in a btree, so
    widen the fields to accept that.
v4: Don't modify xfs_bmap_worst_indlen; the purpose of /that/ function
    is to estimate the worst-case number of blocks needed for a bmbt
    expansion, not to calculate the space required to store nr records.
v5: More minor tweaks to the loop control; move down to reflink patches.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_btree.c |   24 ++++++++++++++++++++++++
 libxfs/xfs_btree.h |    2 ++
 2 files changed, 26 insertions(+)


diff --git a/libxfs/xfs_btree.c b/libxfs/xfs_btree.c
index 99bf808..b3e2ce9 100644
--- a/libxfs/xfs_btree.c
+++ b/libxfs/xfs_btree.c
@@ -4797,3 +4797,27 @@ xfs_btree_query_range(
 	return xfs_btree_overlapped_query_range(cur, &low_key, &high_key,
 			fn, priv);
 }
+
+/*
+ * Calculate the number of blocks needed to store a given number of records
+ * in a short-format (per-AG metadata) btree.
+ */
+xfs_extlen_t
+xfs_btree_calc_size(
+	struct xfs_mount	*mp,
+	uint			*limits,
+	unsigned long long	len)
+{
+	int			level;
+	int			maxrecs;
+	xfs_extlen_t		rval;
+
+	maxrecs = limits[0];
+	for (level = 0, rval = 0; len > 1; level++) {
+		len += maxrecs - 1;
+		do_div(len, maxrecs);
+		maxrecs = limits[1];
+		rval += len;
+	}
+	return rval;
+}
diff --git a/libxfs/xfs_btree.h b/libxfs/xfs_btree.h
index 870a5ab..607463a 100644
--- a/libxfs/xfs_btree.h
+++ b/libxfs/xfs_btree.h
@@ -501,6 +501,8 @@ bool xfs_btree_sblock_v5hdr_verify(struct xfs_buf *bp);
 bool xfs_btree_sblock_verify(struct xfs_buf *bp, unsigned int max_recs);
 uint xfs_btree_compute_maxlevels(struct xfs_mount *mp, uint *limits,
 				 unsigned long len);
+xfs_extlen_t xfs_btree_calc_size(struct xfs_mount *mp, uint *limits,
+		unsigned long long len);
 
 /* return codes */
 #define XFS_BTREE_QUERY_RANGE_CONTINUE	0	/* keep iterating */

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 03/71] xfs: count the blocks in a btree

[Darrick J. Wong]

Provide a helper method to count the number of blocks in a short form
btree.  The refcount and rmap btrees need to know the number of blocks
already in use to set up their per-AG block reservations during mount.

v2: Use btree_visit_blocks instead of open-coding our own traversal
routine.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_btree.c |   23 +++++++++++++++++++++++
 libxfs/xfs_btree.h |    2 ++
 2 files changed, 25 insertions(+)


diff --git a/libxfs/xfs_btree.c b/libxfs/xfs_btree.c
index b3e2ce9..43c809f 100644
--- a/libxfs/xfs_btree.c
+++ b/libxfs/xfs_btree.c
@@ -4821,3 +4821,26 @@ xfs_btree_calc_size(
 	}
 	return rval;
 }
+
+int
+xfs_btree_count_blocks_helper(
+	struct xfs_btree_cur	*cur,
+	int			level,
+	void			*data)
+{
+	xfs_extlen_t		*blocks = data;
+	(*blocks)++;
+
+	return 0;
+}
+
+/* Count the blocks in a btree and return the result in *blocks. */
+int
+xfs_btree_count_blocks(
+	struct xfs_btree_cur	*cur,
+	xfs_extlen_t		*blocks)
+{
+	*blocks = 0;
+	return xfs_btree_visit_blocks(cur, xfs_btree_count_blocks_helper,
+			blocks);
+}
diff --git a/libxfs/xfs_btree.h b/libxfs/xfs_btree.h
index 607463a..be4a0c1 100644
--- a/libxfs/xfs_btree.h
+++ b/libxfs/xfs_btree.h
@@ -519,4 +519,6 @@ typedef int (*xfs_btree_visit_blocks_fn)(struct xfs_btree_cur *cur, int level,
 int xfs_btree_visit_blocks(struct xfs_btree_cur *cur,
 		xfs_btree_visit_blocks_fn fn, void *data);
 
+int xfs_btree_count_blocks(struct xfs_btree_cur *cur, xfs_extlen_t *blocks);
+
 #endif	/* __XFS_BTREE_H__ */

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 04/71] xfs: defer should allow ->finish_item to request a new transaction

[Darrick J. Wong]

When xfs_defer_finish calls ->finish_item, it's possible that
(refcount) won't be able to finish all the work in a single
transaction.  When this happens, the ->finish_item handler should
shorten the log done item's list count, update the work item to
reflect where work should continue, and return -EAGAIN so that
defer_finish knows to retain the pending item on the pending list,
roll the transaction, and restart processing where we left off.

Plumb in the code and document how this mechanism is supposed to work.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_defer.c |   78 +++++++++++++++++++++++++++++++++++++++++++++++-----
 1 file changed, 71 insertions(+), 7 deletions(-)


diff --git a/libxfs/xfs_defer.c b/libxfs/xfs_defer.c
index 06d56ba..27fef4b 100644
--- a/libxfs/xfs_defer.c
+++ b/libxfs/xfs_defer.c
@@ -81,6 +81,10 @@
  *   - For each work item attached to the log intent item,
  *     * Perform the described action.
  *     * Attach the work item to the log done item.
+ *     * If the result of doing the work was -EAGAIN, ->finish work
+ *       wants a new transaction.  See the "Requesting a Fresh
+ *       Transaction while Finishing Deferred Work" section below for
+ *       details.
  *
  * The key here is that we must log an intent item for all pending
  * work items every time we roll the transaction, and that we must log
@@ -88,6 +92,34 @@
  * we can perform complex remapping operations, chaining intent items
  * as needed.
  *
+ * Requesting a Fresh Transaction while Finishing Deferred Work
+ *
+ * If ->finish_item decides that it needs a fresh transaction to
+ * finish the work, it must ask its caller (xfs_defer_finish) for a
+ * continuation.  The most likely cause of this circumstance are the
+ * refcount adjust functions deciding that they've logged enough items
+ * to be at risk of exceeding the transaction reservation.
+ *
+ * To get a fresh transaction, we want to log the existing log done
+ * item to prevent the log intent item from replaying, immediately log
+ * a new log intent item with the unfinished work items, roll the
+ * transaction, and re-call ->finish_item wherever it left off.  The
+ * log done item and the new log intent item must be in the same
+ * transaction or atomicity cannot be guaranteed; defer_finish ensures
+ * that this happens.
+ *
+ * This requires some coordination between ->finish_item and
+ * defer_finish.  Upon deciding to request a new transaction,
+ * ->finish_item should update the current work item to reflect the
+ * unfinished work.  Next, it should reset the log done item's list
+ * count to the number of items finished, and return -EAGAIN.
+ * defer_finish sees the -EAGAIN, logs the new log intent item
+ * with the remaining work items, and leaves the xfs_defer_pending
+ * item at the head of the dop_work queue.  Then it rolls the
+ * transaction and picks up processing where it left off.  It is
+ * required that ->finish_item must be careful to leave enough
+ * transaction reservation to fit the new log intent item.
+ *
  * This is an example of remapping the extent (E, E+B) into file X at
  * offset A and dealing with the extent (C, C+B) already being mapped
  * there:
@@ -104,21 +136,26 @@
  * | Intent to add rmap (X, E, A, B)                 |
  * +-------------------------------------------------+
  * | Reduce refcount for extent (C, B)               | t2
- * | Done reducing refcount for extent (C, B)        |
+ * | Done reducing refcount for extent (C, 9)        |
+ * | Intent to reduce refcount for extent (C+9, B-9) |
+ * | (ran out of space after 9 refcount updates)     |
+ * +-------------------------------------------------+
+ * | Reduce refcount for extent (C+9, B+9)           | t3
+ * | Done reducing refcount for extent (C+9, B-9)    |
  * | Increase refcount for extent (E, B)             |
  * | Done increasing refcount for extent (E, B)      |
  * | Intent to free extent (C, B)                    |
  * | Intent to free extent (F, 1) (refcountbt block) |
  * | Intent to remove rmap (F, 1, REFC)              |
  * +-------------------------------------------------+
- * | Remove rmap (X, C, A, B)                        | t3
+ * | Remove rmap (X, C, A, B)                        | t4
  * | Done removing rmap (X, C, A, B)                 |
  * | Add rmap (X, E, A, B)                           |
  * | Done adding rmap (X, E, A, B)                   |
  * | Remove rmap (F, 1, REFC)                        |
  * | Done removing rmap (F, 1, REFC)                 |
  * +-------------------------------------------------+
- * | Free extent (C, B)                              | t4
+ * | Free extent (C, B)                              | t5
  * | Done freeing extent (C, B)                      |
  * | Free extent (D, 1)                              |
  * | Done freeing extent (D, 1)                      |
@@ -141,6 +178,9 @@
  * - Intent to free extent (C, B)
  * - Intent to free extent (F, 1) (refcountbt block)
  * - Intent to remove rmap (F, 1, REFC)
+ *
+ * Note that the continuation requested between t2 and t3 is likely to
+ * reoccur.
  */
 
 static const struct xfs_defer_op_type *defer_op_types[XFS_DEFER_OPS_TYPE_MAX];
@@ -332,7 +372,16 @@ xfs_defer_finish(
 			dfp->dfp_count--;
 			error = dfp->dfp_type->finish_item(*tp, dop, li,
 					done_item, &state);
-			if (error) {
+			if (error == -EAGAIN) {
+				/*
+				 * Caller wants a fresh transaction;
+				 * put the work item back on the list
+				 * and jump out.
+				 */
+				list_add(li, &dfp->dfp_work);
+				dfp->dfp_count++;
+				break;
+			} else if (error) {
 				/*
 				 * Clean up after ourselves and jump out.
 				 * xfs_defer_cancel will take care of freeing
@@ -344,9 +393,24 @@ xfs_defer_finish(
 				goto out;
 			}
 		}
-		/* Done with the dfp, free it. */
-		list_del(&dfp->dfp_list);
-		kmem_free(dfp);
+		if (error == -EAGAIN) {
+			/*
+			 * Caller wants a fresh transaction, so log a
+			 * new log intent item to replace the old one
+			 * and roll the transaction.  See "Requesting
+			 * a Fresh Transaction while Finishing
+			 * Deferred Work" above.
+			 */
+			dfp->dfp_intent = dfp->dfp_type->create_intent(*tp,
+					dfp->dfp_count);
+			list_for_each(li, &dfp->dfp_work)
+				dfp->dfp_type->log_item(*tp, dfp->dfp_intent,
+						li);
+		} else {
+			/* Done with the dfp, free it. */
+			list_del(&dfp->dfp_list);
+			kmem_free(dfp);
+		}
 
 		if (cleanup_fn)
 			cleanup_fn(*tp, state, error);

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 05/71] xfs: set up per-AG free space reservations

[Darrick J. Wong]

One unfortunate quirk of the reference count btree -- it can expand in
size when blocks are written to *other* allocation groups if, say, one
large extent becomes a lot of tiny extents.  Since we don't want to
start throwing errors in the middle of CoWing, we need to reserve some
blocks to handle future expansion.

Use the count of how many reserved blocks we need to have on hand to
create a virtual reservation in the AG.  Through selective clamping of
the maximum length of allocation requests and of the length of the
longest free extent, we can make it look like there's less free space
in the AG unless the reservation owner is asking for blocks.

In other words, play some accounting tricks in-core to make sure that
we always have blocks available.  On the plus side, there's nothing to
clean up if we crash, which is contrast to the strategy that the rough
draft used (actually removing extents from the freespace btrees).

v2: There's really only two kinds of per-AG reservation pools -- one
to feed the AGFL (rmapbt), and one to feed everything else
(refcountbt).  Bearing that in mind, we can embed the reservation
controls in xfs_perag and greatly simplify the block accounting.
Furthermore, fix some longstanding accounting bugs that were a direct
result of the goofy "allocate a block and later fix up the accounting"
strategy by integrating the reservation accounting code more tightly
with the allocator.  This eliminates the ENOSPC complaints resulting
from refcount btree splits during truncate operations.

v3: Since AGFL blocks are considered to be "free", we mustn't change
fdblocks in response to any AGFL grow/shrink operation.  Therefore, we
must give back to fdblocks at umount whatever we borrowed at mount.
We have to let ag_reserved go down to zero because it's used to
determine if we're critically low on reservation.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 include/xfs_mount.h       |   36 +++++
 include/xfs_trace.h       |   14 ++
 libxfs/Makefile           |    2 
 libxfs/defer_item.c       |    3 
 libxfs/xfs_ag_resv.c      |  325 +++++++++++++++++++++++++++++++++++++++++++++
 libxfs/xfs_ag_resv.h      |   35 +++++
 libxfs/xfs_alloc.c        |  109 ++++++++++-----
 libxfs/xfs_alloc.h        |    8 +
 libxfs/xfs_bmap.c         |    6 +
 libxfs/xfs_ialloc_btree.c |    2 
 10 files changed, 499 insertions(+), 41 deletions(-)
 create mode 100644 libxfs/xfs_ag_resv.c
 create mode 100644 libxfs/xfs_ag_resv.h


diff --git a/include/xfs_mount.h b/include/xfs_mount.h
index 5cd9464..62b9ef2 100644
--- a/include/xfs_mount.h
+++ b/include/xfs_mount.h
@@ -112,6 +112,22 @@ typedef struct xfs_mount {
 	struct xlog		*m_log;
 } xfs_mount_t;
 
+/* per-AG block reservation data structures*/
+enum xfs_ag_resv_type {
+	XFS_AG_RESV_NONE = 0,
+	XFS_AG_RESV_METADATA,
+	XFS_AG_RESV_AGFL,
+};
+
+struct xfs_ag_resv {
+	/* number of blocks originally reserved here */
+	xfs_extlen_t			ar_orig_reserved;
+	/* number of blocks reserved here */
+	xfs_extlen_t			ar_reserved;
+	/* number of blocks originally asked for */
+	xfs_extlen_t			ar_asked;
+};
+
 /*
  * Per-ag incore structure, copies of information in agf and agi,
  * to improve the performance of allocation group selection.
@@ -142,8 +158,28 @@ typedef struct xfs_perag {
 	xfs_agino_t	pagl_leftrec;
 	xfs_agino_t	pagl_rightrec;
 	int		pagb_count;	/* pagb slots in use */
+
+	/* Blocks reserved for all kinds of metadata. */
+	struct xfs_ag_resv	pag_meta_resv;
+	/* Blocks reserved for just AGFL-based metadata. */
+	struct xfs_ag_resv	pag_agfl_resv;
 } xfs_perag_t;
 
+static inline struct xfs_ag_resv *
+xfs_perag_resv(
+	struct xfs_perag	*pag,
+	enum xfs_ag_resv_type	type)
+{
+	switch (type) {
+	case XFS_AG_RESV_METADATA:
+		return &pag->pag_meta_resv;
+	case XFS_AG_RESV_AGFL:
+		return &pag->pag_agfl_resv;
+	default:
+		return NULL;
+	}
+}
+
 #define LIBXFS_MOUNT_DEBUGGER		0x0001
 #define LIBXFS_MOUNT_32BITINODES	0x0002
 #define LIBXFS_MOUNT_32BITINOOPT	0x0004
diff --git a/include/xfs_trace.h b/include/xfs_trace.h
index a636a8c..a32eec3 100644
--- a/include/xfs_trace.h
+++ b/include/xfs_trace.h
@@ -215,4 +215,18 @@
 #define trace_xfs_rmapbt_free_block(...)	((void) 0)
 #define trace_xfs_rmapbt_alloc_block(...)	((void) 0)
 
+#define trace_xfs_ag_resv_critical(...)		((void) 0)
+#define trace_xfs_ag_resv_needed(...)		((void) 0)
+#define trace_xfs_ag_resv_free(...)		((void) 0)
+#define trace_xfs_ag_resv_free_error(...)	((void) 0)
+#define trace_xfs_ag_resv_init(...)		((void) 0)
+#define trace_xfs_ag_resv_init_error(...)	((void) 0)
+#define trace_xfs_ag_resv_alloc_extent(...)	((void) 0)
+#define trace_xfs_ag_resv_free_extent(...)	((void) 0)
+
+/* set c = c to avoid unused var warnings */
+#define trace_xfs_perag_get(a,b,c,d)	((c) = (c))
+#define trace_xfs_perag_get_tag(a,b,c,d) ((c) = (c))
+#define trace_xfs_perag_put(a,b,c,d)	((c) = (c))
+
 #endif /* __TRACE_H__ */
diff --git a/libxfs/Makefile b/libxfs/Makefile
index 62608bd..1b3bcca 100644
--- a/libxfs/Makefile
+++ b/libxfs/Makefile
@@ -18,6 +18,7 @@ PKGHFILES = xfs_fs.h \
 	xfs_log_format.h
 
 HFILES = \
+	xfs_ag_resv.h \
 	xfs_alloc.h \
 	xfs_alloc_btree.h \
 	xfs_attr_leaf.h \
@@ -60,6 +61,7 @@ CFILES = cache.c \
 	rdwr.c \
 	trans.c \
 	util.c \
+	xfs_ag_resv.c \
 	xfs_alloc.c \
 	xfs_alloc_btree.c \
 	xfs_attr.c \
diff --git a/libxfs/defer_item.c b/libxfs/defer_item.c
index de4f769..f60a11b 100644
--- a/libxfs/defer_item.c
+++ b/libxfs/defer_item.c
@@ -95,7 +95,8 @@ xfs_extent_free_finish_item(
 
 	free = container_of(item, struct xfs_extent_free_item, xefi_list);
 	error = xfs_free_extent(tp, free->xefi_startblock,
-			free->xefi_blockcount, &free->xefi_oinfo);
+			free->xefi_blockcount, &free->xefi_oinfo,
+			XFS_AG_RESV_NONE);
 	kmem_free(free);
 	return error;
 }
diff --git a/libxfs/xfs_ag_resv.c b/libxfs/xfs_ag_resv.c
new file mode 100644
index 0000000..af69235
--- /dev/null
+++ b/libxfs/xfs_ag_resv.c
@@ -0,0 +1,325 @@
+/*
+ * Copyright (C) 2016 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#include "libxfs_priv.h"
+#include "xfs_fs.h"
+#include "xfs_shared.h"
+#include "xfs_format.h"
+#include "xfs_log_format.h"
+#include "xfs_trans_resv.h"
+#include "xfs_sb.h"
+#include "xfs_mount.h"
+#include "xfs_defer.h"
+#include "xfs_alloc.h"
+#include "xfs_trace.h"
+#include "xfs_cksum.h"
+#include "xfs_trans.h"
+#include "xfs_bit.h"
+#include "xfs_bmap.h"
+#include "xfs_bmap_btree.h"
+#include "xfs_ag_resv.h"
+#include "xfs_trans_space.h"
+#include "xfs_rmap_btree.h"
+#include "xfs_btree.h"
+
+/*
+ * Per-AG Block Reservations
+ *
+ * For some kinds of allocation group metadata structures, it is advantageous
+ * to reserve a small number of blocks in each AG so that future expansions of
+ * that data structure do not encounter ENOSPC because errors during a btree
+ * split cause the filesystem to go offline.
+ *
+ * Prior to the introduction of reflink, this wasn't an issue because the free
+ * space btrees maintain a reserve of space (the AGFL) to handle any expansion
+ * that may be necessary; and allocations of other metadata (inodes, BMBT,
+ * dir/attr) aren't restricted to a single AG.  However, with reflink it is
+ * possible to allocate all the space in an AG, have subsequent reflink/CoW
+ * activity expand the refcount btree, and discover that there's no space left
+ * to handle that expansion.  Since we can calculate the maximum size of the
+ * refcount btree, we can reserve space for it and avoid ENOSPC.
+ *
+ * Handling per-AG reservations consists of three changes to the allocator's
+ * behavior:  First, because these reservations are always needed, we decrease
+ * the ag_max_usable counter to reflect the size of the AG after the reserved
+ * blocks are taken.  Second, the reservations must be reflected in the
+ * fdblocks count to maintain proper accounting.  Third, each AG must maintain
+ * its own reserved block counter so that we can calculate the amount of space
+ * that must remain free to maintain the reservations.  Fourth, the "remaining
+ * reserved blocks" count must be used when calculating the length of the
+ * longest free extent in an AG and to clamp maxlen in the per-AG allocation
+ * functions.  In other words, we maintain a virtual allocation via in-core
+ * accounting tricks so that we don't have to clean up after a crash. :)
+ *
+ * Reserved blocks can be managed by passing one of the enum xfs_ag_resv_type
+ * values via struct xfs_alloc_arg or directly to the xfs_free_extent
+ * function.  It might seem a little funny to maintain a reservoir of blocks
+ * to feed another reservoir, but the AGFL only holds enough blocks to get
+ * through the next transaction.  The per-AG reservation is to ensure (we
+ * hope) that each AG never runs out of blocks.  Each data structure wanting
+ * to use the reservation system should update ask/used in xfs_ag_resv_init.
+ */
+
+/*
+ * Are we critically low on blocks?  For now we'll define that as the number
+ * of blocks we can get our hands on being less than 10% of what we reserved
+ * or less than some arbitrary number (eight).
+ */
+bool
+xfs_ag_resv_critical(
+	struct xfs_perag		*pag,
+	enum xfs_ag_resv_type		type)
+{
+	xfs_extlen_t			avail;
+	xfs_extlen_t			orig;
+
+	switch (type) {
+	case XFS_AG_RESV_METADATA:
+		avail = pag->pagf_freeblks - pag->pag_agfl_resv.ar_reserved;
+		orig = pag->pag_meta_resv.ar_asked;
+		break;
+	case XFS_AG_RESV_AGFL:
+		avail = pag->pagf_freeblks + pag->pagf_flcount -
+			pag->pag_meta_resv.ar_reserved;
+		orig = pag->pag_agfl_resv.ar_asked;
+		break;
+	default:
+		ASSERT(0);
+		return false;
+	}
+
+	trace_xfs_ag_resv_critical(pag, type, avail);
+
+	return avail < orig / 10 || avail < XFS_BTREE_MAXLEVELS;
+}
+
+/*
+ * How many blocks are reserved but not used, and therefore must not be
+ * allocated away?
+ */
+xfs_extlen_t
+xfs_ag_resv_needed(
+	struct xfs_perag		*pag,
+	enum xfs_ag_resv_type		type)
+{
+	xfs_extlen_t			len;
+
+	len = pag->pag_meta_resv.ar_reserved + pag->pag_agfl_resv.ar_reserved;
+	switch (type) {
+	case XFS_AG_RESV_METADATA:
+	case XFS_AG_RESV_AGFL:
+		len -= xfs_perag_resv(pag, type)->ar_reserved;
+		break;
+	case XFS_AG_RESV_NONE:
+		/* empty */
+		break;
+	default:
+		ASSERT(0);
+	}
+
+	trace_xfs_ag_resv_needed(pag, type, len);
+
+	return len;
+}
+
+/* Clean out a reservation */
+static int
+__xfs_ag_resv_free(
+	struct xfs_perag		*pag,
+	enum xfs_ag_resv_type		type)
+{
+	struct xfs_ag_resv		*resv;
+	xfs_extlen_t			oldresv;
+	int				error;
+
+	trace_xfs_ag_resv_free(pag, type, 0);
+
+	resv = xfs_perag_resv(pag, type);
+	pag->pag_mount->m_ag_max_usable += resv->ar_asked;
+	/*
+	 * AGFL blocks are always considered "free", so whatever
+	 * was reserved at mount time must be given back at umount.
+	 */
+	oldresv = (type == XFS_AG_RESV_AGFL ? resv->ar_orig_reserved :
+					      resv->ar_reserved);
+	error = xfs_mod_fdblocks(pag->pag_mount, oldresv, true);
+	resv->ar_reserved = 0;
+	resv->ar_asked = 0;
+
+	if (error)
+		trace_xfs_ag_resv_free_error(pag->pag_mount, pag->pag_agno,
+				error, _RET_IP_);
+	return error;
+}
+
+/* Free a per-AG reservation. */
+int
+xfs_ag_resv_free(
+	struct xfs_perag		*pag)
+{
+	int				error = 0;
+	int				err2;
+
+	err2 = __xfs_ag_resv_free(pag, XFS_AG_RESV_AGFL);
+	if (err2 && !error)
+		error = err2;
+	err2 = __xfs_ag_resv_free(pag, XFS_AG_RESV_METADATA);
+	if (err2 && !error)
+		error = err2;
+	return error;
+}
+
+static int
+__xfs_ag_resv_init(
+	struct xfs_perag		*pag,
+	enum xfs_ag_resv_type		type,
+	xfs_extlen_t			ask,
+	xfs_extlen_t			used)
+{
+	struct xfs_mount		*mp = pag->pag_mount;
+	struct xfs_ag_resv		*resv;
+	int				error;
+
+	resv = xfs_perag_resv(pag, type);
+	if (used > ask)
+		ask = used;
+	resv->ar_asked = ask;
+	resv->ar_reserved = resv->ar_orig_reserved = ask - used;
+	mp->m_ag_max_usable -= ask;
+
+	trace_xfs_ag_resv_init(pag, type, ask);
+
+	error = xfs_mod_fdblocks(mp, -(int64_t)resv->ar_reserved, true);
+	if (error)
+		trace_xfs_ag_resv_init_error(pag->pag_mount, pag->pag_agno,
+				error, _RET_IP_);
+
+	return error;
+}
+
+/* Create a per-AG block reservation. */
+int
+xfs_ag_resv_init(
+	struct xfs_perag		*pag)
+{
+	xfs_extlen_t			ask;
+	xfs_extlen_t			used;
+	int				error = 0;
+	int				err2;
+
+	if (pag->pag_meta_resv.ar_asked)
+		goto init_agfl;
+
+	/* Create the metadata reservation. */
+	ask = used = 0;
+
+	err2 = __xfs_ag_resv_init(pag, XFS_AG_RESV_METADATA, ask, used);
+	if (err2 && !error)
+		error = err2;
+
+init_agfl:
+	if (pag->pag_agfl_resv.ar_asked)
+		return error;
+
+	/* Create the AGFL metadata reservation */
+	ask = used = 0;
+
+	err2 = __xfs_ag_resv_init(pag, XFS_AG_RESV_AGFL, ask, used);
+	if (err2 && !error)
+		error = err2;
+
+	return error;
+}
+
+/* Allocate a block from the reservation. */
+void
+xfs_ag_resv_alloc_extent(
+	struct xfs_perag		*pag,
+	enum xfs_ag_resv_type		type,
+	struct xfs_alloc_arg		*args)
+{
+	struct xfs_ag_resv		*resv;
+	xfs_extlen_t			len;
+	uint				field;
+
+	trace_xfs_ag_resv_alloc_extent(pag, type, args->len);
+
+	switch (type) {
+	case XFS_AG_RESV_METADATA:
+	case XFS_AG_RESV_AGFL:
+		resv = xfs_perag_resv(pag, type);
+		break;
+	default:
+		ASSERT(0);
+		/* fall through */
+	case XFS_AG_RESV_NONE:
+		field = args->wasdel ? XFS_TRANS_SB_RES_FDBLOCKS :
+				       XFS_TRANS_SB_FDBLOCKS;
+		xfs_trans_mod_sb(args->tp, field, -(int64_t)args->len);
+		return;
+	}
+
+	len = min_t(xfs_extlen_t, args->len, resv->ar_reserved);
+	resv->ar_reserved -= len;
+	if (type == XFS_AG_RESV_AGFL)
+		return;
+	/* Allocations of reserved blocks only need on-disk sb updates... */
+	xfs_trans_mod_sb(args->tp, XFS_TRANS_SB_RES_FDBLOCKS, -(int64_t)len);
+	/* ...but non-reserved blocks need in-core and on-disk updates. */
+	if (args->len > len)
+		xfs_trans_mod_sb(args->tp, XFS_TRANS_SB_FDBLOCKS,
+				-((int64_t)args->len - len));
+}
+
+/* Free a block to the reservation. */
+void
+xfs_ag_resv_free_extent(
+	struct xfs_perag		*pag,
+	enum xfs_ag_resv_type		type,
+	struct xfs_trans		*tp,
+	xfs_extlen_t			len)
+{
+	xfs_extlen_t			leftover;
+	struct xfs_ag_resv		*resv;
+
+	trace_xfs_ag_resv_free_extent(pag, type, len);
+
+	switch (type) {
+	case XFS_AG_RESV_METADATA:
+	case XFS_AG_RESV_AGFL:
+		resv = xfs_perag_resv(pag, type);
+		break;
+	default:
+		ASSERT(0);
+		/* fall through */
+	case XFS_AG_RESV_NONE:
+		xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, (int64_t)len);
+		return;
+	}
+
+	leftover = min_t(xfs_extlen_t, len, resv->ar_asked - resv->ar_reserved);
+	resv->ar_reserved += leftover;
+	if (type == XFS_AG_RESV_AGFL)
+		return;
+	/* Freeing into the reserved pool only requires on-disk update... */
+	xfs_trans_mod_sb(tp, XFS_TRANS_SB_RES_FDBLOCKS, len);
+	/* ...but freeing beyond that requires in-core and on-disk update. */
+	if (len > leftover)
+		xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, len - leftover);
+}
diff --git a/libxfs/xfs_ag_resv.h b/libxfs/xfs_ag_resv.h
new file mode 100644
index 0000000..8d6c687
--- /dev/null
+++ b/libxfs/xfs_ag_resv.h
@@ -0,0 +1,35 @@
+/*
+ * Copyright (C) 2016 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#ifndef __XFS_AG_RESV_H__
+#define	__XFS_AG_RESV_H__
+
+int xfs_ag_resv_free(struct xfs_perag *pag);
+int xfs_ag_resv_init(struct xfs_perag *pag);
+
+bool xfs_ag_resv_critical(struct xfs_perag *pag, enum xfs_ag_resv_type type);
+xfs_extlen_t xfs_ag_resv_needed(struct xfs_perag *pag,
+		enum xfs_ag_resv_type type);
+
+void xfs_ag_resv_alloc_extent(struct xfs_perag *pag, enum xfs_ag_resv_type type,
+		struct xfs_alloc_arg *args);
+void xfs_ag_resv_free_extent(struct xfs_perag *pag, enum xfs_ag_resv_type type,
+		struct xfs_trans *tp, xfs_extlen_t len);
+
+#endif	/* __XFS_AG_RESV_H__ */
diff --git a/libxfs/xfs_alloc.c b/libxfs/xfs_alloc.c
index 8353c77..5f111dd 100644
--- a/libxfs/xfs_alloc.c
+++ b/libxfs/xfs_alloc.c
@@ -33,6 +33,7 @@
 #include "xfs_cksum.h"
 #include "xfs_trace.h"
 #include "xfs_trans.h"
+#include "xfs_ag_resv.h"
 
 struct workqueue_struct *xfs_alloc_wq;
 
@@ -70,14 +71,8 @@ xfs_prealloc_blocks(
  * extents need to be actually allocated. To get around this, we explicitly set
  * aside a few blocks which will not be reserved in delayed allocation.
  *
- * When rmap is disabled, we need to reserve 4 fsbs _per AG_ for the freelist
- * and 4 more to handle a potential split of the file's bmap btree.
- *
- * When rmap is enabled, we must also be able to handle two rmap btree inserts
- * to record both the file data extent and a new bmbt block.  The bmbt block
- * might not be in the same AG as the file data extent.  In the worst case
- * the bmap btree splits multiple levels and all the new blocks come from
- * different AGs, so set aside enough to handle rmap btree splits in all AGs.
+ * We need to reserve 4 fsbs _per AG_ for the freelist and 4 more to handle a
+ * potential split of the file's bmap btree.
  */
 unsigned int
 xfs_alloc_set_aside(
@@ -86,8 +81,6 @@ xfs_alloc_set_aside(
 	unsigned int		blocks;
 
 	blocks = 4 + (mp->m_sb.sb_agcount * XFS_ALLOC_AGFL_RESERVE);
-	if (xfs_sb_version_hasrmapbt(&mp->m_sb))
-		blocks += mp->m_sb.sb_agcount * mp->m_rmap_maxlevels;
 	return blocks;
 }
 
@@ -676,12 +669,29 @@ xfs_alloc_ag_vextent(
 	xfs_alloc_arg_t	*args)	/* argument structure for allocation */
 {
 	int		error=0;
+	xfs_extlen_t	reservation;
+	xfs_extlen_t	oldmax;
 
 	ASSERT(args->minlen > 0);
 	ASSERT(args->maxlen > 0);
 	ASSERT(args->minlen <= args->maxlen);
 	ASSERT(args->mod < args->prod);
 	ASSERT(args->alignment > 0);
+
+	/*
+	 * Clamp maxlen to the amount of free space minus any reservations
+	 * that have been made.
+	 */
+	oldmax = args->maxlen;
+	reservation = xfs_ag_resv_needed(args->pag, args->resv);
+	if (args->maxlen > args->pag->pagf_freeblks - reservation)
+		args->maxlen = args->pag->pagf_freeblks - reservation;
+	if (args->maxlen == 0) {
+		args->agbno = NULLAGBLOCK;
+		args->maxlen = oldmax;
+		return 0;
+	}
+
 	/*
 	 * Branch to correct routine based on the type.
 	 */
@@ -701,12 +711,14 @@ xfs_alloc_ag_vextent(
 		/* NOTREACHED */
 	}
 
+	args->maxlen = oldmax;
+
 	if (error || args->agbno == NULLAGBLOCK)
 		return error;
 
 	ASSERT(args->len >= args->minlen);
 	ASSERT(args->len <= args->maxlen);
-	ASSERT(!args->wasfromfl || !args->isfl);
+	ASSERT(!args->wasfromfl || args->resv != XFS_AG_RESV_AGFL);
 	ASSERT(args->agbno % args->alignment == 0);
 
 	/* if not file data, insert new block into the reverse map btree */
@@ -728,12 +740,7 @@ xfs_alloc_ag_vextent(
 					      args->agbno, args->len));
 	}
 
-	if (!args->isfl) {
-		xfs_trans_mod_sb(args->tp, args->wasdel ?
-				 XFS_TRANS_SB_RES_FDBLOCKS :
-				 XFS_TRANS_SB_FDBLOCKS,
-				 -((long)(args->len)));
-	}
+	xfs_ag_resv_alloc_extent(args->pag, args->resv, args);
 
 	XFS_STATS_INC(args->mp, xs_allocx);
 	XFS_STATS_ADD(args->mp, xs_allocb, args->len);
@@ -1597,7 +1604,8 @@ xfs_alloc_ag_vextent_small(
 	 * to respect minleft even when pulling from the
 	 * freelist.
 	 */
-	else if (args->minlen == 1 && args->alignment == 1 && !args->isfl &&
+	else if (args->minlen == 1 && args->alignment == 1 &&
+		 args->resv != XFS_AG_RESV_AGFL &&
 		 (be32_to_cpu(XFS_BUF_TO_AGF(args->agbp)->agf_flcount)
 		  > args->minleft)) {
 		error = xfs_alloc_get_freelist(args->tp, args->agbp, &fbno, 0);
@@ -1626,7 +1634,8 @@ xfs_alloc_ag_vextent_small(
 			/*
 			 * If we're feeding an AGFL block to something that
 			 * doesn't live in the free space, we need to clear
-			 * out the OWN_AG rmap.
+			 * out the OWN_AG rmap and add the block back to
+			 * the AGFL per-AG reservation.
 			 */
 			xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_AG);
 			error = xfs_rmap_free(args->tp, args->agbp, args->agno,
@@ -1634,6 +1643,8 @@ xfs_alloc_ag_vextent_small(
 			if (error)
 				goto error0;
 			pag = xfs_perag_get(args->mp, args->agno);
+			xfs_ag_resv_free_extent(pag, XFS_AG_RESV_AGFL,
+					args->tp, 1);
 			xfs_perag_put(pag);
 
 			*stat = 0;
@@ -1682,7 +1693,7 @@ xfs_free_ag_extent(
 	xfs_agblock_t		bno,
 	xfs_extlen_t		len,
 	struct xfs_owner_info	*oinfo,
-	int			isfl)
+	enum xfs_ag_resv_type	type)
 {
 	xfs_btree_cur_t	*bno_cur;	/* cursor for by-block btree */
 	xfs_btree_cur_t	*cnt_cur;	/* cursor for by-size btree */
@@ -1910,21 +1921,22 @@ xfs_free_ag_extent(
 	 */
 	pag = xfs_perag_get(mp, agno);
 	error = xfs_alloc_update_counters(tp, pag, agbp, len);
+	xfs_ag_resv_free_extent(pag, type, tp, len);
 	xfs_perag_put(pag);
 	if (error)
 		goto error0;
 
-	if (!isfl)
-		xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, (long)len);
 	XFS_STATS_INC(mp, xs_freex);
 	XFS_STATS_ADD(mp, xs_freeb, len);
 
-	trace_xfs_free_extent(mp, agno, bno, len, isfl, haveleft, haveright);
+	trace_xfs_free_extent(mp, agno, bno, len, type == XFS_AG_RESV_AGFL,
+			haveleft, haveright);
 
 	return 0;
 
  error0:
-	trace_xfs_free_extent(mp, agno, bno, len, isfl, -1, -1);
+	trace_xfs_free_extent(mp, agno, bno, len, type == XFS_AG_RESV_AGFL,
+			-1, -1);
 	if (bno_cur)
 		xfs_btree_del_cursor(bno_cur, XFS_BTREE_ERROR);
 	if (cnt_cur)
@@ -1949,21 +1961,43 @@ xfs_alloc_compute_maxlevels(
 }
 
 /*
- * Find the length of the longest extent in an AG.
+ * Find the length of the longest extent in an AG.  The 'need' parameter
+ * specifies how much space we're going to need for the AGFL and the
+ * 'reserved' parameter tells us how many blocks in this AG are reserved for
+ * other callers.
  */
 xfs_extlen_t
 xfs_alloc_longest_free_extent(
 	struct xfs_mount	*mp,
 	struct xfs_perag	*pag,
-	xfs_extlen_t		need)
+	xfs_extlen_t		need,
+	xfs_extlen_t		reserved)
 {
 	xfs_extlen_t		delta = 0;
 
+	/*
+	 * If the AGFL needs a recharge, we'll have to subtract that from the
+	 * longest extent.
+	 */
 	if (need > pag->pagf_flcount)
 		delta = need - pag->pagf_flcount;
 
+	/*
+	 * If we cannot maintain others' reservations with space from the
+	 * not-longest freesp extents, we'll have to subtract /that/ from
+	 * the longest extent too.
+	 */
+	if (pag->pagf_freeblks - pag->pagf_longest < reserved)
+		delta += reserved - (pag->pagf_freeblks - pag->pagf_longest);
+
+	/*
+	 * If the longest extent is long enough to satisfy all the
+	 * reservations and AGFL rules in place, we can return this extent.
+	 */
 	if (pag->pagf_longest > delta)
 		return pag->pagf_longest - delta;
+
+	/* Otherwise, let the caller try for 1 block if there's space. */
 	return pag->pagf_flcount > 0 || pag->pagf_longest > 0;
 }
 
@@ -2003,20 +2037,24 @@ xfs_alloc_space_available(
 {
 	struct xfs_perag	*pag = args->pag;
 	xfs_extlen_t		longest;
+	xfs_extlen_t		reservation; /* blocks that are still reserved */
 	int			available;
 
 	if (flags & XFS_ALLOC_FLAG_FREEING)
 		return true;
 
+	reservation = xfs_ag_resv_needed(pag, args->resv);
+
 	/* do we have enough contiguous free space for the allocation? */
-	longest = xfs_alloc_longest_free_extent(args->mp, pag, min_free);
+	longest = xfs_alloc_longest_free_extent(args->mp, pag, min_free,
+			reservation);
 	if ((args->minlen + args->alignment + args->minalignslop - 1) > longest)
 		return false;
 
-	/* do have enough free space remaining for the allocation? */
+	/* do we have enough free space remaining for the allocation? */
 	available = (int)(pag->pagf_freeblks + pag->pagf_flcount -
-			  min_free - args->total);
-	if (available < (int)args->minleft)
+			  reservation - min_free - args->total);
+	if (available < (int)args->minleft || available <= 0)
 		return false;
 
 	return true;
@@ -2123,7 +2161,7 @@ xfs_alloc_fix_freelist(
 		if (error)
 			goto out_agbp_relse;
 		error = xfs_free_ag_extent(tp, agbp, args->agno, bno, 1,
-					   &targs.oinfo, 1);
+					   &targs.oinfo, XFS_AG_RESV_AGFL);
 		if (error)
 			goto out_agbp_relse;
 		bp = xfs_btree_get_bufs(mp, tp, args->agno, bno, 0);
@@ -2134,7 +2172,7 @@ xfs_alloc_fix_freelist(
 	targs.mp = mp;
 	targs.agbp = agbp;
 	targs.agno = args->agno;
-	targs.alignment = targs.minlen = targs.prod = targs.isfl = 1;
+	targs.alignment = targs.minlen = targs.prod = 1;
 	targs.type = XFS_ALLOCTYPE_THIS_AG;
 	targs.pag = pag;
 	error = xfs_alloc_read_agfl(mp, tp, targs.agno, &agflbp);
@@ -2145,6 +2183,7 @@ xfs_alloc_fix_freelist(
 	while (pag->pagf_flcount < need) {
 		targs.agbno = 0;
 		targs.maxlen = need - pag->pagf_flcount;
+		targs.resv = XFS_AG_RESV_AGFL;
 
 		/* Allocate as many blocks as possible at once. */
 		error = xfs_alloc_ag_vextent(&targs);
@@ -2826,7 +2865,8 @@ xfs_free_extent(
 	struct xfs_trans	*tp,	/* transaction pointer */
 	xfs_fsblock_t		bno,	/* starting block number of extent */
 	xfs_extlen_t		len,	/* length of extent */
-	struct xfs_owner_info	*oinfo)	/* extent owner */
+	struct xfs_owner_info	*oinfo,	/* extent owner */
+	enum xfs_ag_resv_type	type)	/* block reservation type */
 {
 	struct xfs_mount	*mp = tp->t_mountp;
 	struct xfs_buf		*agbp;
@@ -2835,6 +2875,7 @@ xfs_free_extent(
 	int			error;
 
 	ASSERT(len != 0);
+	ASSERT(type != XFS_AG_RESV_AGFL);
 
 	if (XFS_TEST_ERROR(false, mp,
 			XFS_ERRTAG_FREE_EXTENT,
@@ -2852,7 +2893,7 @@ xfs_free_extent(
 		agbno + len <= be32_to_cpu(XFS_BUF_TO_AGF(agbp)->agf_length),
 				err);
 
-	error = xfs_free_ag_extent(tp, agbp, agno, agbno, len, oinfo, 0);
+	error = xfs_free_ag_extent(tp, agbp, agno, agbno, len, oinfo, type);
 	if (error)
 		goto err;
 
diff --git a/libxfs/xfs_alloc.h b/libxfs/xfs_alloc.h
index 6fe2d6b..f7c5201 100644
--- a/libxfs/xfs_alloc.h
+++ b/libxfs/xfs_alloc.h
@@ -87,10 +87,10 @@ typedef struct xfs_alloc_arg {
 	xfs_alloctype_t	otype;		/* original allocation type */
 	char		wasdel;		/* set if allocation was prev delayed */
 	char		wasfromfl;	/* set if allocation is from freelist */
-	char		isfl;		/* set if is freelist blocks - !acctg */
 	char		userdata;	/* mask defining userdata treatment */
 	xfs_fsblock_t	firstblock;	/* io first block allocated */
 	struct xfs_owner_info	oinfo;	/* owner of blocks being allocated */
+	enum xfs_ag_resv_type	resv;	/* block reservation to use */
 } xfs_alloc_arg_t;
 
 /*
@@ -106,7 +106,8 @@ unsigned int xfs_alloc_set_aside(struct xfs_mount *mp);
 unsigned int xfs_alloc_ag_max_usable(struct xfs_mount *mp);
 
 xfs_extlen_t xfs_alloc_longest_free_extent(struct xfs_mount *mp,
-		struct xfs_perag *pag, xfs_extlen_t need);
+		struct xfs_perag *pag, xfs_extlen_t need,
+		xfs_extlen_t reserved);
 unsigned int xfs_alloc_min_freelist(struct xfs_mount *mp,
 		struct xfs_perag *pag);
 
@@ -184,7 +185,8 @@ xfs_free_extent(
 	struct xfs_trans	*tp,	/* transaction pointer */
 	xfs_fsblock_t		bno,	/* starting block number of extent */
 	xfs_extlen_t		len,	/* length of extent */
-	struct xfs_owner_info	*oinfo);/* extent owner */
+	struct xfs_owner_info	*oinfo,	/* extent owner */
+	enum xfs_ag_resv_type	type);	/* block reservation type */
 
 int				/* error */
 xfs_alloc_lookup_ge(
diff --git a/libxfs/xfs_bmap.c b/libxfs/xfs_bmap.c
index 7da6b60..a50a8af 100644
--- a/libxfs/xfs_bmap.c
+++ b/libxfs/xfs_bmap.c
@@ -39,6 +39,7 @@
 #include "xfs_attr_leaf.h"
 #include "xfs_quota_defs.h"
 #include "xfs_rmap.h"
+#include "xfs_ag_resv.h"
 
 
 kmem_zone_t		*xfs_bmap_free_item_zone;
@@ -3493,7 +3494,8 @@ xfs_bmap_longest_free_extent(
 	}
 
 	longest = xfs_alloc_longest_free_extent(mp, pag,
-					xfs_alloc_min_freelist(mp, pag));
+				xfs_alloc_min_freelist(mp, pag),
+				xfs_ag_resv_needed(pag, XFS_AG_RESV_NONE));
 	if (*blen < longest)
 		*blen = longest;
 
@@ -3773,7 +3775,7 @@ xfs_bmap_btalloc(
 	}
 	args.minleft = ap->minleft;
 	args.wasdel = ap->wasdel;
-	args.isfl = 0;
+	args.resv = XFS_AG_RESV_NONE;
 	args.userdata = ap->userdata;
 	if (ap->userdata & XFS_ALLOC_USERDATA_ZERO)
 		args.ip = ap->ip;
diff --git a/libxfs/xfs_ialloc_btree.c b/libxfs/xfs_ialloc_btree.c
index 642e3bb..7bf6040 100644
--- a/libxfs/xfs_ialloc_btree.c
+++ b/libxfs/xfs_ialloc_btree.c
@@ -131,7 +131,7 @@ xfs_inobt_free_block(
 	xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_INOBT);
 	return xfs_free_extent(cur->bc_tp,
 			XFS_DADDR_TO_FSB(cur->bc_mp, XFS_BUF_ADDR(bp)), 1,
-			&oinfo);
+			&oinfo, XFS_AG_RESV_NONE);
 }
 
 STATIC int

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 06/71] xfs: introduce refcount btree definitions

[Darrick J. Wong]

Add new per-AG refcount btree definitions to the per-AG structures.

v2: Move the reflink inode flag out of the way of the DAX flag, and
add the new cowextsize flag.

v3: Don't allow pNFS to export reflinked files; this will be removed
some day when the Linux pNFS server supports it.

[hch: don't allow pNFS export of reflinked files]
[darrick: fix the feature test in hch's patch]

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/xfs_inode.h     |    5 +++++
 include/xfs_mount.h     |    3 +++
 libxfs/xfs_alloc.c      |    5 +++++
 libxfs/xfs_btree.c      |    5 +++--
 libxfs/xfs_btree.h      |    3 +++
 libxfs/xfs_format.h     |   31 +++++++++++++++++++++++++++----
 libxfs/xfs_rmap_btree.c |    7 +++++--
 libxfs/xfs_types.h      |    2 +-
 8 files changed, 52 insertions(+), 9 deletions(-)


diff --git a/include/xfs_inode.h b/include/xfs_inode.h
index 8141d97..3876fa6 100644
--- a/include/xfs_inode.h
+++ b/include/xfs_inode.h
@@ -122,6 +122,11 @@ xfs_set_projid(struct xfs_icdinode *id, prid_t projid)
 	id->di_projid_lo = (__uint16_t) (projid & 0xffff);
 }
 
+static inline bool xfs_is_reflink_inode(struct xfs_inode *ip)
+{
+	return ip->i_d.di_flags2 & XFS_DIFLAG2_REFLINK;
+}
+
 typedef struct cred {
 	uid_t	cr_uid;
 	gid_t	cr_gid;
diff --git a/include/xfs_mount.h b/include/xfs_mount.h
index 62b9ef2..8301b78 100644
--- a/include/xfs_mount.h
+++ b/include/xfs_mount.h
@@ -163,6 +163,9 @@ typedef struct xfs_perag {
 	struct xfs_ag_resv	pag_meta_resv;
 	/* Blocks reserved for just AGFL-based metadata. */
 	struct xfs_ag_resv	pag_agfl_resv;
+
+	/* reference count */
+	__uint8_t	pagf_refcount_level;
 } xfs_perag_t;
 
 static inline struct xfs_ag_resv *
diff --git a/libxfs/xfs_alloc.c b/libxfs/xfs_alloc.c
index 5f111dd..a9a0820 100644
--- a/libxfs/xfs_alloc.c
+++ b/libxfs/xfs_alloc.c
@@ -2453,6 +2453,10 @@ xfs_agf_verify(
 	    be32_to_cpu(agf->agf_btreeblks) > be32_to_cpu(agf->agf_length))
 		return false;
 
+	if (xfs_sb_version_hasreflink(&mp->m_sb) &&
+	    be32_to_cpu(agf->agf_refcount_level) > XFS_BTREE_MAXLEVELS)
+		return false;
+
 	return true;;
 
 }
@@ -2573,6 +2577,7 @@ xfs_alloc_read_agf(
 			be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNTi]);
 		pag->pagf_levels[XFS_BTNUM_RMAPi] =
 			be32_to_cpu(agf->agf_levels[XFS_BTNUM_RMAPi]);
+		pag->pagf_refcount_level = be32_to_cpu(agf->agf_refcount_level);
 		spin_lock_init(&pag->pagb_lock);
 		pag->pagb_count = 0;
 #ifdef __KERNEL__
diff --git a/libxfs/xfs_btree.c b/libxfs/xfs_btree.c
index 43c809f..3ad060f 100644
--- a/libxfs/xfs_btree.c
+++ b/libxfs/xfs_btree.c
@@ -41,9 +41,10 @@ kmem_zone_t	*xfs_btree_cur_zone;
  */
 static const __uint32_t xfs_magics[2][XFS_BTNUM_MAX] = {
 	{ XFS_ABTB_MAGIC, XFS_ABTC_MAGIC, 0, XFS_BMAP_MAGIC, XFS_IBT_MAGIC,
-	  XFS_FIBT_MAGIC },
+	  XFS_FIBT_MAGIC, 0 },
 	{ XFS_ABTB_CRC_MAGIC, XFS_ABTC_CRC_MAGIC, XFS_RMAP_CRC_MAGIC,
-	  XFS_BMAP_CRC_MAGIC, XFS_IBT_CRC_MAGIC, XFS_FIBT_CRC_MAGIC }
+	  XFS_BMAP_CRC_MAGIC, XFS_IBT_CRC_MAGIC, XFS_FIBT_CRC_MAGIC,
+	  XFS_REFC_CRC_MAGIC }
 };
 #define xfs_btree_magic(cur) \
 	xfs_magics[!!((cur)->bc_flags & XFS_BTREE_CRC_BLOCKS)][cur->bc_btnum]
diff --git a/libxfs/xfs_btree.h b/libxfs/xfs_btree.h
index be4a0c1..b5d820e 100644
--- a/libxfs/xfs_btree.h
+++ b/libxfs/xfs_btree.h
@@ -72,6 +72,7 @@ union xfs_btree_rec {
 #define	XFS_BTNUM_INO	((xfs_btnum_t)XFS_BTNUM_INOi)
 #define	XFS_BTNUM_FINO	((xfs_btnum_t)XFS_BTNUM_FINOi)
 #define	XFS_BTNUM_RMAP	((xfs_btnum_t)XFS_BTNUM_RMAPi)
+#define	XFS_BTNUM_REFC	((xfs_btnum_t)XFS_BTNUM_REFCi)
 
 /*
  * For logging record fields.
@@ -105,6 +106,7 @@ do {    \
 	case XFS_BTNUM_INO: __XFS_BTREE_STATS_INC(__mp, ibt, stat); break; \
 	case XFS_BTNUM_FINO: __XFS_BTREE_STATS_INC(__mp, fibt, stat); break; \
 	case XFS_BTNUM_RMAP: __XFS_BTREE_STATS_INC(__mp, rmap, stat); break; \
+	case XFS_BTNUM_REFC: break; \
 	case XFS_BTNUM_MAX: ASSERT(0); __mp = __mp /* fucking gcc */ ; break; \
 	}       \
 } while (0)
@@ -127,6 +129,7 @@ do {    \
 		__XFS_BTREE_STATS_ADD(__mp, fibt, stat, val); break; \
 	case XFS_BTNUM_RMAP:	\
 		__XFS_BTREE_STATS_ADD(__mp, rmap, stat, val); break; \
+	case XFS_BTNUM_REFC: break; \
 	case XFS_BTNUM_MAX: ASSERT(0); __mp = __mp /* fucking gcc */ ; break; \
 	}       \
 } while (0)
diff --git a/libxfs/xfs_format.h b/libxfs/xfs_format.h
index ffa1931..3d1bad7 100644
--- a/libxfs/xfs_format.h
+++ b/libxfs/xfs_format.h
@@ -456,6 +456,7 @@ xfs_sb_has_compat_feature(
 
 #define XFS_SB_FEAT_RO_COMPAT_FINOBT   (1 << 0)		/* free inode btree */
 #define XFS_SB_FEAT_RO_COMPAT_RMAPBT   (1 << 1)		/* reverse map btree */
+#define XFS_SB_FEAT_RO_COMPAT_REFLINK  (1 << 2)		/* reflinked files */
 #define XFS_SB_FEAT_RO_COMPAT_ALL \
 		(XFS_SB_FEAT_RO_COMPAT_FINOBT | \
 		 XFS_SB_FEAT_RO_COMPAT_RMAPBT)
@@ -546,6 +547,12 @@ static inline bool xfs_sb_version_hasrmapbt(struct xfs_sb *sbp)
 		(sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_RMAPBT);
 }
 
+static inline bool xfs_sb_version_hasreflink(struct xfs_sb *sbp)
+{
+	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) &&
+		(sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_REFLINK);
+}
+
 /*
  * end of superblock version macros
  */
@@ -641,14 +648,17 @@ typedef struct xfs_agf {
 	uuid_t		agf_uuid;	/* uuid of filesystem */
 
 	__be32		agf_rmap_blocks;	/* rmapbt blocks used */
-	__be32		agf_padding;		/* padding */
+	__be32		agf_refcount_blocks;	/* refcountbt blocks used */
+
+	__be32		agf_refcount_root;	/* refcount tree root block */
+	__be32		agf_refcount_level;	/* refcount btree levels */
 
 	/*
 	 * reserve some contiguous space for future logged fields before we add
 	 * the unlogged fields. This makes the range logging via flags and
 	 * structure offsets much simpler.
 	 */
-	__be64		agf_spare64[15];
+	__be64		agf_spare64[14];
 
 	/* unlogged fields, written during buffer writeback. */
 	__be64		agf_lsn;	/* last write sequence */
@@ -1040,9 +1050,14 @@ static inline void xfs_dinode_put_rdev(struct xfs_dinode *dip, xfs_dev_t rdev)
  * 16 bits of the XFS_XFLAG_s range.
  */
 #define XFS_DIFLAG2_DAX_BIT	0	/* use DAX for this inode */
+#define XFS_DIFLAG2_REFLINK_BIT	1	/* file's blocks may be shared */
+#define XFS_DIFLAG2_COWEXTSIZE_BIT   2  /* copy on write extent size hint */
 #define XFS_DIFLAG2_DAX		(1 << XFS_DIFLAG2_DAX_BIT)
+#define XFS_DIFLAG2_REFLINK     (1 << XFS_DIFLAG2_REFLINK_BIT)
+#define XFS_DIFLAG2_COWEXTSIZE  (1 << XFS_DIFLAG2_COWEXTSIZE_BIT)
 
-#define XFS_DIFLAG2_ANY		(XFS_DIFLAG2_DAX)
+#define XFS_DIFLAG2_ANY \
+	(XFS_DIFLAG2_DAX | XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE)
 
 /*
  * Inode number format:
@@ -1352,7 +1367,8 @@ struct xfs_owner_info {
 #define XFS_RMAP_OWN_AG		(-5ULL)	/* AG freespace btree blocks */
 #define XFS_RMAP_OWN_INOBT	(-6ULL)	/* Inode btree blocks */
 #define XFS_RMAP_OWN_INODES	(-7ULL)	/* Inode chunk */
-#define XFS_RMAP_OWN_MIN	(-8ULL) /* guard */
+#define XFS_RMAP_OWN_REFC	(-8ULL) /* refcount tree */
+#define XFS_RMAP_OWN_MIN	(-9ULL) /* guard */
 
 #define XFS_RMAP_NON_INODE_OWNER(owner)	(!!((owner) & (1ULL << 63)))
 
@@ -1433,6 +1449,13 @@ typedef __be32 xfs_rmap_ptr_t;
 	 XFS_IBT_BLOCK(mp) + 1)
 
 /*
+ * Reference Count Btree format definitions
+ *
+ */
+#define	XFS_REFC_CRC_MAGIC	0x52334643	/* 'R3FC' */
+
+
+/*
  * BMAP Btree format definitions
  *
  * This includes both the root block definition that sits inside an inode fork
diff --git a/libxfs/xfs_rmap_btree.c b/libxfs/xfs_rmap_btree.c
index d1b3c08..29a4dd6 100644
--- a/libxfs/xfs_rmap_btree.c
+++ b/libxfs/xfs_rmap_btree.c
@@ -510,6 +510,9 @@ void
 xfs_rmapbt_compute_maxlevels(
 	struct xfs_mount		*mp)
 {
-	mp->m_rmap_maxlevels = xfs_btree_compute_maxlevels(mp,
-			mp->m_rmap_mnr, mp->m_sb.sb_agblocks);
+	if (xfs_sb_version_hasreflink(&mp->m_sb))
+		mp->m_rmap_maxlevels = XFS_BTREE_MAXLEVELS;
+	else
+		mp->m_rmap_maxlevels = xfs_btree_compute_maxlevels(mp,
+				mp->m_rmap_mnr, mp->m_sb.sb_agblocks);
 }
diff --git a/libxfs/xfs_types.h b/libxfs/xfs_types.h
index da87796..690d616 100644
--- a/libxfs/xfs_types.h
+++ b/libxfs/xfs_types.h
@@ -112,7 +112,7 @@ typedef enum {
 
 typedef enum {
 	XFS_BTNUM_BNOi, XFS_BTNUM_CNTi, XFS_BTNUM_RMAPi, XFS_BTNUM_BMAPi,
-	XFS_BTNUM_INOi, XFS_BTNUM_FINOi, XFS_BTNUM_MAX
+	XFS_BTNUM_INOi, XFS_BTNUM_FINOi, XFS_BTNUM_REFCi, XFS_BTNUM_MAX
 } xfs_btnum_t;
 
 struct xfs_name {

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 07/71] xfs: add refcount btree stats infrastructure

[Darrick J. Wong]

The refcount btree presents the same stats as the other btrees, so
add all the code for that now.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_btree.h |    5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)


diff --git a/libxfs/xfs_btree.h b/libxfs/xfs_btree.h
index b5d820e..81ee006 100644
--- a/libxfs/xfs_btree.h
+++ b/libxfs/xfs_btree.h
@@ -106,7 +106,7 @@ do {    \
 	case XFS_BTNUM_INO: __XFS_BTREE_STATS_INC(__mp, ibt, stat); break; \
 	case XFS_BTNUM_FINO: __XFS_BTREE_STATS_INC(__mp, fibt, stat); break; \
 	case XFS_BTNUM_RMAP: __XFS_BTREE_STATS_INC(__mp, rmap, stat); break; \
-	case XFS_BTNUM_REFC: break; \
+	case XFS_BTNUM_REFC: __XFS_BTREE_STATS_INC(__mp, refcbt, stat); break; \
 	case XFS_BTNUM_MAX: ASSERT(0); __mp = __mp /* fucking gcc */ ; break; \
 	}       \
 } while (0)
@@ -129,7 +129,8 @@ do {    \
 		__XFS_BTREE_STATS_ADD(__mp, fibt, stat, val); break; \
 	case XFS_BTNUM_RMAP:	\
 		__XFS_BTREE_STATS_ADD(__mp, rmap, stat, val); break; \
-	case XFS_BTNUM_REFC: break; \
+	case XFS_BTNUM_REFC:	\
+		__XFS_BTREE_STATS_ADD(__mp, refcbt, stat, val); break; \
 	case XFS_BTNUM_MAX: ASSERT(0); __mp = __mp /* fucking gcc */ ; break; \
 	}       \
 } while (0)

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 08/71] xfs: refcount btree add more reserved blocks

[Darrick J. Wong]

Since XFS reserves a small amount of space in each AG as the minimum
free space needed for an operation, save some more space in case we
touch the refcount btree.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_alloc.c  |   13 +++++++++++++
 libxfs/xfs_format.h |    2 ++
 2 files changed, 15 insertions(+)


diff --git a/libxfs/xfs_alloc.c b/libxfs/xfs_alloc.c
index a9a0820..8920c71 100644
--- a/libxfs/xfs_alloc.c
+++ b/libxfs/xfs_alloc.c
@@ -48,10 +48,23 @@ STATIC int xfs_alloc_ag_vextent_size(xfs_alloc_arg_t *);
 STATIC int xfs_alloc_ag_vextent_small(xfs_alloc_arg_t *,
 		xfs_btree_cur_t *, xfs_agblock_t *, xfs_extlen_t *, int *);
 
+unsigned int
+xfs_refc_block(
+	struct xfs_mount	*mp)
+{
+	if (xfs_sb_version_hasrmapbt(&mp->m_sb))
+		return XFS_RMAP_BLOCK(mp) + 1;
+	if (xfs_sb_version_hasfinobt(&mp->m_sb))
+		return XFS_FIBT_BLOCK(mp) + 1;
+	return XFS_IBT_BLOCK(mp) + 1;
+}
+
 xfs_extlen_t
 xfs_prealloc_blocks(
 	struct xfs_mount	*mp)
 {
+	if (xfs_sb_version_hasreflink(&mp->m_sb))
+		return xfs_refc_block(mp) + 1;
 	if (xfs_sb_version_hasrmapbt(&mp->m_sb))
 		return XFS_RMAP_BLOCK(mp) + 1;
 	if (xfs_sb_version_hasfinobt(&mp->m_sb))
diff --git a/libxfs/xfs_format.h b/libxfs/xfs_format.h
index 3d1bad7..b028466 100644
--- a/libxfs/xfs_format.h
+++ b/libxfs/xfs_format.h
@@ -1454,6 +1454,8 @@ typedef __be32 xfs_rmap_ptr_t;
  */
 #define	XFS_REFC_CRC_MAGIC	0x52334643	/* 'R3FC' */
 
+unsigned int xfs_refc_block(struct xfs_mount *mp);
+
 
 /*
  * BMAP Btree format definitions

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 09/71] xfs: define the on-disk refcount btree format

[Darrick J. Wong]

Start constructing the refcount btree implementation by establishing
the on-disk format and everything needed to read, write, and
manipulate the refcount btree blocks.

v2: Calculate a separate maxlevels for the refcount btree.

v3: Enable the tracking of per-cursor stats for refcount btrees.
The refcount update code will use this to guess if it's time to
split a refcountbt update across two transactions to avoid
exhausing the transaction reservation.

xfs_refcountbt_init_cursor can be called under the ilock, so
use KM_NOFS to prevent fs activity with a lock held.  This
should shut up some of the lockdep warnings.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
[hch: allocate the cursor with KM_NOFS to quiet lockdep]
Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/darwin.h            |    1 
 include/freebsd.h           |    1 
 include/gnukfreebsd.h       |    1 
 include/irix.h              |    1 
 include/libxfs.h            |    1 
 include/linux.h             |    1 
 include/xfs_mount.h         |    3 +
 libxfs/Makefile             |    2 
 libxfs/init.c               |    2 
 libxfs/xfs_btree.c          |    3 +
 libxfs/xfs_btree.h          |   12 +++
 libxfs/xfs_format.h         |   32 ++++++++
 libxfs/xfs_refcount_btree.c |  177 +++++++++++++++++++++++++++++++++++++++++++
 libxfs/xfs_refcount_btree.h |   67 ++++++++++++++++
 libxfs/xfs_sb.c             |    9 ++
 libxfs/xfs_shared.h         |    2 
 libxfs/xfs_trans_resv.c     |    2 
 libxfs/xfs_trans_resv.h     |    1 
 18 files changed, 317 insertions(+), 1 deletion(-)
 create mode 100644 libxfs/xfs_refcount_btree.c
 create mode 100644 libxfs/xfs_refcount_btree.h


diff --git a/include/darwin.h b/include/darwin.h
index 45e0c03..140386a 100644
--- a/include/darwin.h
+++ b/include/darwin.h
@@ -140,6 +140,7 @@ typedef off_t		xfs_off_t;
 typedef u_int64_t	xfs_ino_t;
 typedef u_int32_t	xfs_dev_t;
 typedef int64_t		xfs_daddr_t;
+typedef u_int32_t	xfs_nlink_t;
 
 #define stat64		stat
 #define fstat64		fstat
diff --git a/include/freebsd.h b/include/freebsd.h
index 6e77427..668dcea 100644
--- a/include/freebsd.h
+++ b/include/freebsd.h
@@ -51,6 +51,7 @@ typedef off_t		off64_t;
 typedef __uint64_t	xfs_ino_t;
 typedef __uint32_t	xfs_dev_t;
 typedef __int64_t	xfs_daddr_t;
+typedef __uint32_t	xfs_nlink_t;
 
 typedef unsigned char		__u8;
 typedef signed char		__s8;
diff --git a/include/gnukfreebsd.h b/include/gnukfreebsd.h
index d55acfb..dfd31ae 100644
--- a/include/gnukfreebsd.h
+++ b/include/gnukfreebsd.h
@@ -40,6 +40,7 @@ typedef off_t		xfs_off_t;
 typedef __uint64_t	xfs_ino_t;
 typedef __uint32_t	xfs_dev_t;
 typedef __int64_t	xfs_daddr_t;
+typedef __uint32_t	xfs_nlink_t;
 
 typedef unsigned char		__u8;
 typedef signed char		__s8;
diff --git a/include/irix.h b/include/irix.h
index b92e01b..8b9d588 100644
--- a/include/irix.h
+++ b/include/irix.h
@@ -47,6 +47,7 @@ typedef off64_t		xfs_off_t;
 typedef __int64_t	xfs_ino_t;
 typedef __int32_t	xfs_dev_t;
 typedef __int64_t	xfs_daddr_t;
+typedef __int32_t	xfs_nlink_t;
 
 typedef unsigned char		__u8;
 typedef signed char		__s8;
diff --git a/include/libxfs.h b/include/libxfs.h
index cf59d6c..ec8f6ab 100644
--- a/include/libxfs.h
+++ b/include/libxfs.h
@@ -79,6 +79,7 @@ extern uint32_t crc32c_le(uint32_t crc, unsigned char const *p, size_t len);
 #include "xfs_trans.h"
 #include "xfs_rmap_btree.h"
 #include "xfs_rmap.h"
+#include "xfs_refcount_btree.h"
 
 #ifndef ARRAY_SIZE
 #define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0]))
diff --git a/include/linux.h b/include/linux.h
index 5614719..1663930 100644
--- a/include/linux.h
+++ b/include/linux.h
@@ -141,6 +141,7 @@ typedef off64_t		xfs_off_t;
 typedef __uint64_t	xfs_ino_t;
 typedef __uint32_t	xfs_dev_t;
 typedef __int64_t	xfs_daddr_t;
+typedef __uint32_t	xfs_nlink_t;
 
 /**
  * Abstraction of mountpoints.
diff --git a/include/xfs_mount.h b/include/xfs_mount.h
index 8301b78..97b4ad7 100644
--- a/include/xfs_mount.h
+++ b/include/xfs_mount.h
@@ -66,10 +66,13 @@ typedef struct xfs_mount {
 	uint			m_inobt_mnr[2];	/* XFS_INOBT_BLOCK_MINRECS */
 	uint			m_rmap_mxr[2];	/* max rmap btree records */
 	uint			m_rmap_mnr[2];	/* min rmap btree records */
+	uint			m_refc_mxr[2];	/* max refc btree records */
+	uint			m_refc_mnr[2];	/* min refc btree records */
 	uint			m_ag_maxlevels;	/* XFS_AG_MAXLEVELS */
 	uint			m_bm_maxlevels[2]; /* XFS_BM_MAXLEVELS */
 	uint			m_in_maxlevels;	/* XFS_IN_MAXLEVELS */
 	uint			m_rmap_maxlevels; /* max rmap btree levels */
+	uint			m_refc_maxlevels; /* max refcount btree level */
 	xfs_extlen_t		m_ag_prealloc_blocks; /* reserved ag blocks */
 	uint			m_alloc_set_aside; /* space we can't use */
 	uint			m_ag_max_usable; /* max space per AG */
diff --git a/libxfs/Makefile b/libxfs/Makefile
index 1b3bcca..ccfd8e9 100644
--- a/libxfs/Makefile
+++ b/libxfs/Makefile
@@ -36,6 +36,7 @@ HFILES = \
 	xfs_inode_buf.h \
 	xfs_inode_fork.h \
 	xfs_quota_defs.h \
+	xfs_refcount_btree.h \
 	xfs_rmap.h \
 	xfs_rmap_btree.h \
 	xfs_sb.h \
@@ -86,6 +87,7 @@ CFILES = cache.c \
 	xfs_inode_fork.c \
 	xfs_ialloc_btree.c \
 	xfs_log_rlimit.c \
+	xfs_refcount_btree.c \
 	xfs_rmap.c \
 	xfs_rmap_btree.c \
 	xfs_rtbitmap.c \
diff --git a/libxfs/init.c b/libxfs/init.c
index c13b123..706925d 100644
--- a/libxfs/init.c
+++ b/libxfs/init.c
@@ -32,6 +32,7 @@
 #include "xfs_inode.h"
 #include "xfs_trans.h"
 #include "xfs_rmap_btree.h"
+#include "xfs_refcount_btree.h"
 
 #include "libxfs.h"		/* for now */
 
@@ -687,6 +688,7 @@ libxfs_mount(
 	xfs_bmap_compute_maxlevels(mp, XFS_ATTR_FORK);
 	xfs_ialloc_compute_maxlevels(mp);
 	xfs_rmapbt_compute_maxlevels(mp);
+	xfs_refcountbt_compute_maxlevels(mp);
 
 	if (sbp->sb_imax_pct) {
 		/* Make sure the maximum inode count is a multiple of the
diff --git a/libxfs/xfs_btree.c b/libxfs/xfs_btree.c
index 3ad060f..0468a9e 100644
--- a/libxfs/xfs_btree.c
+++ b/libxfs/xfs_btree.c
@@ -1213,6 +1213,9 @@ xfs_btree_set_refs(
 	case XFS_BTNUM_RMAP:
 		xfs_buf_set_ref(bp, XFS_RMAP_BTREE_REF);
 		break;
+	case XFS_BTNUM_REFC:
+		xfs_buf_set_ref(bp, XFS_REFC_BTREE_REF);
+		break;
 	default:
 		ASSERT(0);
 	}
diff --git a/libxfs/xfs_btree.h b/libxfs/xfs_btree.h
index 81ee006..eb20376 100644
--- a/libxfs/xfs_btree.h
+++ b/libxfs/xfs_btree.h
@@ -49,6 +49,7 @@ union xfs_btree_key {
 	struct xfs_inobt_key		inobt;
 	struct xfs_rmap_key		rmap;
 	struct xfs_rmap_key		__rmap_bigkey[2];
+	struct xfs_refcount_key		refc;
 };
 
 union xfs_btree_rec {
@@ -57,6 +58,7 @@ union xfs_btree_rec {
 	struct xfs_alloc_rec		alloc;
 	struct xfs_inobt_rec		inobt;
 	struct xfs_rmap_rec		rmap;
+	struct xfs_refcount_rec		refc;
 };
 
 /*
@@ -221,6 +223,15 @@ union xfs_btree_irec {
 	struct xfs_bmbt_irec		b;
 	struct xfs_inobt_rec_incore	i;
 	struct xfs_rmap_irec		r;
+	struct xfs_refcount_irec	rc;
+};
+
+/* Per-AG btree private information. */
+union xfs_btree_cur_private {
+	struct {
+		unsigned long	nr_ops;		/* # record updates */
+		int		shape_changes;	/* # of extent splits */
+	} refc;
 };
 
 /*
@@ -247,6 +258,7 @@ typedef struct xfs_btree_cur
 			struct xfs_buf	*agbp;	/* agf/agi buffer pointer */
 			struct xfs_defer_ops *dfops;	/* deferred updates */
 			xfs_agnumber_t	agno;	/* ag number */
+			union xfs_btree_cur_private	priv;
 		} a;
 		struct {			/* needed for BMAP */
 			struct xfs_inode *ip;	/* pointer to our inode */
diff --git a/libxfs/xfs_format.h b/libxfs/xfs_format.h
index b028466..6c81ce7 100644
--- a/libxfs/xfs_format.h
+++ b/libxfs/xfs_format.h
@@ -1456,6 +1456,38 @@ typedef __be32 xfs_rmap_ptr_t;
 
 unsigned int xfs_refc_block(struct xfs_mount *mp);
 
+/*
+ * Data record/key structure
+ *
+ * Each record associates a range of physical blocks (starting at
+ * rc_startblock and ending rc_blockcount blocks later) with a
+ * reference count (rc_refcount).  A record is only stored in the
+ * btree if the refcount is > 2.  An entry in the free block btree
+ * means that the refcount is 0, and no entries anywhere means that
+ * the refcount is 1, as was true in XFS before reflinking.
+ */
+struct xfs_refcount_rec {
+	__be32		rc_startblock;	/* starting block number */
+	__be32		rc_blockcount;	/* count of blocks */
+	__be32		rc_refcount;	/* number of inodes linked here */
+};
+
+struct xfs_refcount_key {
+	__be32		rc_startblock;	/* starting block number */
+};
+
+struct xfs_refcount_irec {
+	xfs_agblock_t	rc_startblock;	/* starting block number */
+	xfs_extlen_t	rc_blockcount;	/* count of free blocks */
+	xfs_nlink_t	rc_refcount;	/* number of inodes linked here */
+};
+
+#define MAXREFCOUNT	((xfs_nlink_t)~0U)
+#define MAXREFCEXTLEN	((xfs_extlen_t)~0U)
+
+/* btree pointer type */
+typedef __be32 xfs_refcount_ptr_t;
+
 
 /*
  * BMAP Btree format definitions
diff --git a/libxfs/xfs_refcount_btree.c b/libxfs/xfs_refcount_btree.c
new file mode 100644
index 0000000..a7b99e4
--- /dev/null
+++ b/libxfs/xfs_refcount_btree.c
@@ -0,0 +1,177 @@
+/*
+ * Copyright (C) 2016 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#include "libxfs_priv.h"
+#include "xfs_fs.h"
+#include "xfs_shared.h"
+#include "xfs_format.h"
+#include "xfs_log_format.h"
+#include "xfs_trans_resv.h"
+#include "xfs_sb.h"
+#include "xfs_mount.h"
+#include "xfs_btree.h"
+#include "xfs_bmap.h"
+#include "xfs_refcount_btree.h"
+#include "xfs_alloc.h"
+#include "xfs_trace.h"
+#include "xfs_cksum.h"
+#include "xfs_trans.h"
+#include "xfs_bit.h"
+
+static struct xfs_btree_cur *
+xfs_refcountbt_dup_cursor(
+	struct xfs_btree_cur	*cur)
+{
+	return xfs_refcountbt_init_cursor(cur->bc_mp, cur->bc_tp,
+			cur->bc_private.a.agbp, cur->bc_private.a.agno,
+			cur->bc_private.a.dfops);
+}
+
+STATIC bool
+xfs_refcountbt_verify(
+	struct xfs_buf		*bp)
+{
+	struct xfs_mount	*mp = bp->b_target->bt_mount;
+	struct xfs_btree_block	*block = XFS_BUF_TO_BLOCK(bp);
+	struct xfs_perag	*pag = bp->b_pag;
+	unsigned int		level;
+
+	if (block->bb_magic != cpu_to_be32(XFS_REFC_CRC_MAGIC))
+		return false;
+
+	if (!xfs_sb_version_hasreflink(&mp->m_sb))
+		return false;
+	if (!xfs_btree_sblock_v5hdr_verify(bp))
+		return false;
+
+	level = be16_to_cpu(block->bb_level);
+	if (pag && pag->pagf_init) {
+		if (level >= pag->pagf_refcount_level)
+			return false;
+	} else if (level >= mp->m_refc_maxlevels)
+		return false;
+
+	return xfs_btree_sblock_verify(bp, mp->m_refc_mxr[level != 0]);
+}
+
+STATIC void
+xfs_refcountbt_read_verify(
+	struct xfs_buf	*bp)
+{
+	if (!xfs_btree_sblock_verify_crc(bp))
+		xfs_buf_ioerror(bp, -EFSBADCRC);
+	else if (!xfs_refcountbt_verify(bp))
+		xfs_buf_ioerror(bp, -EFSCORRUPTED);
+
+	if (bp->b_error) {
+		trace_xfs_btree_corrupt(bp, _RET_IP_);
+		xfs_verifier_error(bp);
+	}
+}
+
+STATIC void
+xfs_refcountbt_write_verify(
+	struct xfs_buf	*bp)
+{
+	if (!xfs_refcountbt_verify(bp)) {
+		trace_xfs_btree_corrupt(bp, _RET_IP_);
+		xfs_buf_ioerror(bp, -EFSCORRUPTED);
+		xfs_verifier_error(bp);
+		return;
+	}
+	xfs_btree_sblock_calc_crc(bp);
+
+}
+
+const struct xfs_buf_ops xfs_refcountbt_buf_ops = {
+	.name			= "xfs_refcountbt",
+	.verify_read		= xfs_refcountbt_read_verify,
+	.verify_write		= xfs_refcountbt_write_verify,
+};
+
+static const struct xfs_btree_ops xfs_refcountbt_ops = {
+	.rec_len		= sizeof(struct xfs_refcount_rec),
+	.key_len		= sizeof(struct xfs_refcount_key),
+
+	.dup_cursor		= xfs_refcountbt_dup_cursor,
+	.buf_ops		= &xfs_refcountbt_buf_ops,
+};
+
+/*
+ * Allocate a new refcount btree cursor.
+ */
+struct xfs_btree_cur *
+xfs_refcountbt_init_cursor(
+	struct xfs_mount	*mp,
+	struct xfs_trans	*tp,
+	struct xfs_buf		*agbp,
+	xfs_agnumber_t		agno,
+	struct xfs_defer_ops	*dfops)
+{
+	struct xfs_agf		*agf = XFS_BUF_TO_AGF(agbp);
+	struct xfs_btree_cur	*cur;
+
+	ASSERT(agno != NULLAGNUMBER);
+	ASSERT(agno < mp->m_sb.sb_agcount);
+	cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_NOFS);
+
+	cur->bc_tp = tp;
+	cur->bc_mp = mp;
+	cur->bc_btnum = XFS_BTNUM_REFC;
+	cur->bc_blocklog = mp->m_sb.sb_blocklog;
+	cur->bc_ops = &xfs_refcountbt_ops;
+
+	cur->bc_nlevels = be32_to_cpu(agf->agf_refcount_level);
+
+	cur->bc_private.a.agbp = agbp;
+	cur->bc_private.a.agno = agno;
+	cur->bc_private.a.dfops = dfops;
+	cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;
+
+	cur->bc_private.a.priv.refc.nr_ops = 0;
+	cur->bc_private.a.priv.refc.shape_changes = 0;
+
+	return cur;
+}
+
+/*
+ * Calculate the number of records in a refcount btree block.
+ */
+int
+xfs_refcountbt_maxrecs(
+	struct xfs_mount	*mp,
+	int			blocklen,
+	bool			leaf)
+{
+	blocklen -= XFS_REFCOUNT_BLOCK_LEN;
+
+	if (leaf)
+		return blocklen / sizeof(struct xfs_refcount_rec);
+	return blocklen / (sizeof(struct xfs_refcount_key) +
+			   sizeof(xfs_refcount_ptr_t));
+}
+
+/* Compute the maximum height of a refcount btree. */
+void
+xfs_refcountbt_compute_maxlevels(
+	struct xfs_mount		*mp)
+{
+	mp->m_refc_maxlevels = xfs_btree_compute_maxlevels(mp,
+			mp->m_refc_mnr, mp->m_sb.sb_agblocks);
+}
diff --git a/libxfs/xfs_refcount_btree.h b/libxfs/xfs_refcount_btree.h
new file mode 100644
index 0000000..9e9ad7c
--- /dev/null
+++ b/libxfs/xfs_refcount_btree.h
@@ -0,0 +1,67 @@
+/*
+ * Copyright (C) 2016 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#ifndef __XFS_REFCOUNT_BTREE_H__
+#define	__XFS_REFCOUNT_BTREE_H__
+
+/*
+ * Reference Count Btree on-disk structures
+ */
+
+struct xfs_buf;
+struct xfs_btree_cur;
+struct xfs_mount;
+
+/*
+ * Btree block header size
+ */
+#define XFS_REFCOUNT_BLOCK_LEN	XFS_BTREE_SBLOCK_CRC_LEN
+
+/*
+ * Record, key, and pointer address macros for btree blocks.
+ *
+ * (note that some of these may appear unused, but they are used in userspace)
+ */
+#define XFS_REFCOUNT_REC_ADDR(block, index) \
+	((struct xfs_refcount_rec *) \
+		((char *)(block) + \
+		 XFS_REFCOUNT_BLOCK_LEN + \
+		 (((index) - 1) * sizeof(struct xfs_refcount_rec))))
+
+#define XFS_REFCOUNT_KEY_ADDR(block, index) \
+	((struct xfs_refcount_key *) \
+		((char *)(block) + \
+		 XFS_REFCOUNT_BLOCK_LEN + \
+		 ((index) - 1) * sizeof(struct xfs_refcount_key)))
+
+#define XFS_REFCOUNT_PTR_ADDR(block, index, maxrecs) \
+	((xfs_refcount_ptr_t *) \
+		((char *)(block) + \
+		 XFS_REFCOUNT_BLOCK_LEN + \
+		 (maxrecs) * sizeof(struct xfs_refcount_key) + \
+		 ((index) - 1) * sizeof(xfs_refcount_ptr_t)))
+
+extern struct xfs_btree_cur *xfs_refcountbt_init_cursor(struct xfs_mount *mp,
+		struct xfs_trans *tp, struct xfs_buf *agbp, xfs_agnumber_t agno,
+		struct xfs_defer_ops *dfops);
+extern int xfs_refcountbt_maxrecs(struct xfs_mount *mp, int blocklen,
+		bool leaf);
+extern void xfs_refcountbt_compute_maxlevels(struct xfs_mount *mp);
+
+#endif	/* __XFS_REFCOUNT_BTREE_H__ */
diff --git a/libxfs/xfs_sb.c b/libxfs/xfs_sb.c
index 0a1462f..de25489 100644
--- a/libxfs/xfs_sb.c
+++ b/libxfs/xfs_sb.c
@@ -35,6 +35,8 @@
 #include "xfs_alloc_btree.h"
 #include "xfs_ialloc_btree.h"
 #include "xfs_rmap_btree.h"
+#include "xfs_bmap.h"
+#include "xfs_refcount_btree.h"
 
 /*
  * Physical superblock buffer manipulations. Shared with libxfs in userspace.
@@ -740,6 +742,13 @@ xfs_sb_mount_common(
 	mp->m_rmap_mnr[0] = mp->m_rmap_mxr[0] / 2;
 	mp->m_rmap_mnr[1] = mp->m_rmap_mxr[1] / 2;
 
+	mp->m_refc_mxr[0] = xfs_refcountbt_maxrecs(mp, sbp->sb_blocksize,
+			true);
+	mp->m_refc_mxr[1] = xfs_refcountbt_maxrecs(mp, sbp->sb_blocksize,
+			false);
+	mp->m_refc_mnr[0] = mp->m_refc_mxr[0] / 2;
+	mp->m_refc_mnr[1] = mp->m_refc_mxr[1] / 2;
+
 	mp->m_bsize = XFS_FSB_TO_BB(mp, 1);
 	mp->m_ialloc_inos = (int)MAX((__uint16_t)XFS_INODES_PER_CHUNK,
 					sbp->sb_inopblock);
diff --git a/libxfs/xfs_shared.h b/libxfs/xfs_shared.h
index 0c5b30b..c6f4eb4 100644
--- a/libxfs/xfs_shared.h
+++ b/libxfs/xfs_shared.h
@@ -39,6 +39,7 @@ extern const struct xfs_buf_ops xfs_agf_buf_ops;
 extern const struct xfs_buf_ops xfs_agfl_buf_ops;
 extern const struct xfs_buf_ops xfs_allocbt_buf_ops;
 extern const struct xfs_buf_ops xfs_rmapbt_buf_ops;
+extern const struct xfs_buf_ops xfs_refcountbt_buf_ops;
 extern const struct xfs_buf_ops xfs_attr3_leaf_buf_ops;
 extern const struct xfs_buf_ops xfs_attr3_rmt_buf_ops;
 extern const struct xfs_buf_ops xfs_bmbt_buf_ops;
@@ -122,6 +123,7 @@ int	xfs_log_calc_minimum_size(struct xfs_mount *);
 #define	XFS_INO_REF		2
 #define	XFS_ATTR_BTREE_REF	1
 #define	XFS_DQUOT_REF		1
+#define	XFS_REFC_BTREE_REF	1
 
 /*
  * Flags for xfs_trans_ichgtime().
diff --git a/libxfs/xfs_trans_resv.c b/libxfs/xfs_trans_resv.c
index 2ed80a5..10234bb 100644
--- a/libxfs/xfs_trans_resv.c
+++ b/libxfs/xfs_trans_resv.c
@@ -72,7 +72,7 @@ xfs_calc_buf_res(
  *
  * Keep in mind that max depth is calculated separately for each type of tree.
  */
-static uint
+uint
 xfs_allocfree_log_count(
 	struct xfs_mount *mp,
 	uint		num_ops)
diff --git a/libxfs/xfs_trans_resv.h b/libxfs/xfs_trans_resv.h
index 0eb46ed..36a1511 100644
--- a/libxfs/xfs_trans_resv.h
+++ b/libxfs/xfs_trans_resv.h
@@ -102,5 +102,6 @@ struct xfs_trans_resv {
 #define	XFS_ATTRRM_LOG_COUNT		3
 
 void xfs_trans_resv_calc(struct xfs_mount *mp, struct xfs_trans_resv *resp);
+uint xfs_allocfree_log_count(struct xfs_mount *mp, uint num_ops);
 
 #endif	/* __XFS_TRANS_RESV_H__ */

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 10/71] xfs: account for the refcount btree in the alloc/free log reservation

[Darrick J. Wong]

Every time we allocate or free an extent, we might need to split the
refcount btree.  Reserve some blocks in the transaction to handle
this possibility.

(Reproduced by generic/167 over NFS atop XFS)

Signed-off-by: Christoph Hellwig <hch@lst.de>
[darrick.wong@oracle.com: add commit message]
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_trans_resv.c |    5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)


diff --git a/libxfs/xfs_trans_resv.c b/libxfs/xfs_trans_resv.c
index 10234bb..5b6bbcd 100644
--- a/libxfs/xfs_trans_resv.c
+++ b/libxfs/xfs_trans_resv.c
@@ -66,7 +66,8 @@ xfs_calc_buf_res(
  * Per-extent log reservation for the btree changes involved in freeing or
  * allocating an extent.  In classic XFS there were two trees that will be
  * modified (bnobt + cntbt).  With rmap enabled, there are three trees
- * (rmapbt).  The number of blocks reserved is based on the formula:
+ * (rmapbt).  With reflink, there are four trees (refcountbt).  The number of
+ * blocks reserved is based on the formula:
  *
  * num trees * ((2 blocks/level * max depth) - 1)
  *
@@ -82,6 +83,8 @@ xfs_allocfree_log_count(
 	blocks = num_ops * 2 * (2 * mp->m_ag_maxlevels - 1);
 	if (xfs_sb_version_hasrmapbt(&mp->m_sb))
 		blocks += num_ops * (2 * mp->m_rmap_maxlevels - 1);
+	if (xfs_sb_version_hasreflink(&mp->m_sb))
+		blocks += num_ops * (2 * mp->m_refc_maxlevels - 1);
 
 	return blocks;
 }

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 11/71] xfs: add refcount btree operations

[Darrick J. Wong]

Implement the generic btree operations required to manipulate refcount
btree blocks.  The implementation is similar to the bmapbt, though it
will only allocate and free blocks from the AG.

v2: Remove init_rec_from_key since we no longer need it, and add
tracepoints when refcount btree operations fail.

Since the refcount root and level fields are separate from the
existing roots and levels array, they need a separate logging flag.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
[hch: fix logging of AGF refcount btree fields]
Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/libxfs.h            |    1 
 include/xfs_trace.h         |   12 +++
 libxfs/Makefile             |    2 
 libxfs/xfs_alloc.c          |    3 +
 libxfs/xfs_format.h         |   10 ++
 libxfs/xfs_refcount.c       |  176 +++++++++++++++++++++++++++++++++++++
 libxfs/xfs_refcount.h       |   30 ++++++
 libxfs/xfs_refcount_btree.c |  205 +++++++++++++++++++++++++++++++++++++++++++
 8 files changed, 437 insertions(+), 2 deletions(-)
 create mode 100644 libxfs/xfs_refcount.c
 create mode 100644 libxfs/xfs_refcount.h


diff --git a/include/libxfs.h b/include/libxfs.h
index ec8f6ab..e5e1523 100644
--- a/include/libxfs.h
+++ b/include/libxfs.h
@@ -80,6 +80,7 @@ extern uint32_t crc32c_le(uint32_t crc, unsigned char const *p, size_t len);
 #include "xfs_rmap_btree.h"
 #include "xfs_rmap.h"
 #include "xfs_refcount_btree.h"
+#include "xfs_refcount.h"
 
 #ifndef ARRAY_SIZE
 #define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0]))
diff --git a/include/xfs_trace.h b/include/xfs_trace.h
index a32eec3..937d5f5 100644
--- a/include/xfs_trace.h
+++ b/include/xfs_trace.h
@@ -224,6 +224,18 @@
 #define trace_xfs_ag_resv_alloc_extent(...)	((void) 0)
 #define trace_xfs_ag_resv_free_extent(...)	((void) 0)
 
+#define trace_xfs_refcount_lookup(...)	((void) 0)
+#define trace_xfs_refcount_get(...)		((void) 0)
+#define trace_xfs_refcount_update(...)	((void) 0)
+#define trace_xfs_refcount_update_error(...)	((void) 0)
+#define trace_xfs_refcount_insert(...)	((void) 0)
+#define trace_xfs_refcount_insert_error(...)	((void) 0)
+#define trace_xfs_refcount_delete(...)	((void) 0)
+#define trace_xfs_refcount_delete_error(...)	((void) 0)
+#define trace_xfs_refcountbt_free_block(...)	((void) 0)
+#define trace_xfs_refcountbt_alloc_block(...)	((void) 0)
+#define trace_xfs_refcount_rec_order_error(...)	((void) 0)
+
 /* set c = c to avoid unused var warnings */
 #define trace_xfs_perag_get(a,b,c,d)	((c) = (c))
 #define trace_xfs_perag_get_tag(a,b,c,d) ((c) = (c))
diff --git a/libxfs/Makefile b/libxfs/Makefile
index ccfd8e9..6499731 100644
--- a/libxfs/Makefile
+++ b/libxfs/Makefile
@@ -36,6 +36,7 @@ HFILES = \
 	xfs_inode_buf.h \
 	xfs_inode_fork.h \
 	xfs_quota_defs.h \
+	xfs_refcount.h \
 	xfs_refcount_btree.h \
 	xfs_rmap.h \
 	xfs_rmap_btree.h \
@@ -87,6 +88,7 @@ CFILES = cache.c \
 	xfs_inode_fork.c \
 	xfs_ialloc_btree.c \
 	xfs_log_rlimit.c \
+	xfs_refcount.c \
 	xfs_refcount_btree.c \
 	xfs_rmap.c \
 	xfs_rmap_btree.c \
diff --git a/libxfs/xfs_alloc.c b/libxfs/xfs_alloc.c
index 8920c71..e754647 100644
--- a/libxfs/xfs_alloc.c
+++ b/libxfs/xfs_alloc.c
@@ -2329,6 +2329,9 @@ xfs_alloc_log_agf(
 		offsetof(xfs_agf_t, agf_btreeblks),
 		offsetof(xfs_agf_t, agf_uuid),
 		offsetof(xfs_agf_t, agf_rmap_blocks),
+		offsetof(xfs_agf_t, agf_refcount_blocks),
+		offsetof(xfs_agf_t, agf_refcount_root),
+		offsetof(xfs_agf_t, agf_refcount_level),
 		/* needed so that we don't log the whole rest of the structure: */
 		offsetof(xfs_agf_t, agf_spare64),
 		sizeof(xfs_agf_t)
diff --git a/libxfs/xfs_format.h b/libxfs/xfs_format.h
index 6c81ce7..bbb9334 100644
--- a/libxfs/xfs_format.h
+++ b/libxfs/xfs_format.h
@@ -684,8 +684,11 @@ typedef struct xfs_agf {
 #define	XFS_AGF_BTREEBLKS	0x00000800
 #define	XFS_AGF_UUID		0x00001000
 #define	XFS_AGF_RMAP_BLOCKS	0x00002000
-#define	XFS_AGF_SPARE64		0x00004000
-#define	XFS_AGF_NUM_BITS	15
+#define	XFS_AGF_REFCOUNT_BLOCKS	0x00004000
+#define	XFS_AGF_REFCOUNT_ROOT	0x00008000
+#define	XFS_AGF_REFCOUNT_LEVEL	0x00010000
+#define	XFS_AGF_SPARE64		0x00020000
+#define	XFS_AGF_NUM_BITS	18
 #define	XFS_AGF_ALL_BITS	((1 << XFS_AGF_NUM_BITS) - 1)
 
 #define XFS_AGF_FLAGS \
@@ -703,6 +706,9 @@ typedef struct xfs_agf {
 	{ XFS_AGF_BTREEBLKS,	"BTREEBLKS" }, \
 	{ XFS_AGF_UUID,		"UUID" }, \
 	{ XFS_AGF_RMAP_BLOCKS,	"RMAP_BLOCKS" }, \
+	{ XFS_AGF_REFCOUNT_BLOCKS,	"REFCOUNT_BLOCKS" }, \
+	{ XFS_AGF_REFCOUNT_ROOT,	"REFCOUNT_ROOT" }, \
+	{ XFS_AGF_REFCOUNT_LEVEL,	"REFCOUNT_LEVEL" }, \
 	{ XFS_AGF_SPARE64,	"SPARE64" }
 
 /* disk block (xfs_daddr_t) in the AG */
diff --git a/libxfs/xfs_refcount.c b/libxfs/xfs_refcount.c
new file mode 100644
index 0000000..8b522e2
--- /dev/null
+++ b/libxfs/xfs_refcount.c
@@ -0,0 +1,176 @@
+/*
+ * Copyright (C) 2016 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#include "libxfs_priv.h"
+#include "xfs_fs.h"
+#include "xfs_shared.h"
+#include "xfs_format.h"
+#include "xfs_log_format.h"
+#include "xfs_trans_resv.h"
+#include "xfs_sb.h"
+#include "xfs_mount.h"
+#include "xfs_defer.h"
+#include "xfs_btree.h"
+#include "xfs_bmap.h"
+#include "xfs_refcount_btree.h"
+#include "xfs_alloc.h"
+#include "xfs_trace.h"
+#include "xfs_cksum.h"
+#include "xfs_trans.h"
+#include "xfs_bit.h"
+#include "xfs_refcount.h"
+
+/*
+ * Look up the first record less than or equal to [bno, len] in the btree
+ * given by cur.
+ */
+int
+xfs_refcount_lookup_le(
+	struct xfs_btree_cur	*cur,
+	xfs_agblock_t		bno,
+	int			*stat)
+{
+	trace_xfs_refcount_lookup(cur->bc_mp, cur->bc_private.a.agno, bno,
+			XFS_LOOKUP_LE);
+	cur->bc_rec.rc.rc_startblock = bno;
+	cur->bc_rec.rc.rc_blockcount = 0;
+	return xfs_btree_lookup(cur, XFS_LOOKUP_LE, stat);
+}
+
+/*
+ * Look up the first record greater than or equal to [bno, len] in the btree
+ * given by cur.
+ */
+int
+xfs_refcount_lookup_ge(
+	struct xfs_btree_cur	*cur,
+	xfs_agblock_t		bno,
+	int			*stat)
+{
+	trace_xfs_refcount_lookup(cur->bc_mp, cur->bc_private.a.agno, bno,
+			XFS_LOOKUP_GE);
+	cur->bc_rec.rc.rc_startblock = bno;
+	cur->bc_rec.rc.rc_blockcount = 0;
+	return xfs_btree_lookup(cur, XFS_LOOKUP_GE, stat);
+}
+
+/*
+ * Get the data from the pointed-to record.
+ */
+int
+xfs_refcount_get_rec(
+	struct xfs_btree_cur		*cur,
+	struct xfs_refcount_irec	*irec,
+	int				*stat)
+{
+	union xfs_btree_rec	*rec;
+	int			error;
+
+	error = xfs_btree_get_rec(cur, &rec, stat);
+	if (!error && *stat == 1) {
+		irec->rc_startblock = be32_to_cpu(rec->refc.rc_startblock);
+		irec->rc_blockcount = be32_to_cpu(rec->refc.rc_blockcount);
+		irec->rc_refcount = be32_to_cpu(rec->refc.rc_refcount);
+		trace_xfs_refcount_get(cur->bc_mp, cur->bc_private.a.agno,
+				irec);
+	}
+	return error;
+}
+
+/*
+ * Update the record referred to by cur to the value given
+ * by [bno, len, refcount].
+ * This either works (return 0) or gets an EFSCORRUPTED error.
+ */
+STATIC int
+xfs_refcount_update(
+	struct xfs_btree_cur		*cur,
+	struct xfs_refcount_irec	*irec)
+{
+	union xfs_btree_rec	rec;
+	int			error;
+
+	trace_xfs_refcount_update(cur->bc_mp, cur->bc_private.a.agno, irec);
+	rec.refc.rc_startblock = cpu_to_be32(irec->rc_startblock);
+	rec.refc.rc_blockcount = cpu_to_be32(irec->rc_blockcount);
+	rec.refc.rc_refcount = cpu_to_be32(irec->rc_refcount);
+	error = xfs_btree_update(cur, &rec);
+	if (error)
+		trace_xfs_refcount_update_error(cur->bc_mp,
+				cur->bc_private.a.agno, error, _RET_IP_);
+	return error;
+}
+
+/*
+ * Insert the record referred to by cur to the value given
+ * by [bno, len, refcount].
+ * This either works (return 0) or gets an EFSCORRUPTED error.
+ */
+STATIC int
+xfs_refcount_insert(
+	struct xfs_btree_cur		*cur,
+	struct xfs_refcount_irec	*irec,
+	int				*i)
+{
+	int				error;
+
+	trace_xfs_refcount_insert(cur->bc_mp, cur->bc_private.a.agno, irec);
+	cur->bc_rec.rc.rc_startblock = irec->rc_startblock;
+	cur->bc_rec.rc.rc_blockcount = irec->rc_blockcount;
+	cur->bc_rec.rc.rc_refcount = irec->rc_refcount;
+	error = xfs_btree_insert(cur, i);
+	XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, *i == 1, out_error);
+out_error:
+	if (error)
+		trace_xfs_refcount_insert_error(cur->bc_mp,
+				cur->bc_private.a.agno, error, _RET_IP_);
+	return error;
+}
+
+/*
+ * Remove the record referred to by cur, then set the pointer to the spot
+ * where the record could be re-inserted, in case we want to increment or
+ * decrement the cursor.
+ * This either works (return 0) or gets an EFSCORRUPTED error.
+ */
+STATIC int
+xfs_refcount_delete(
+	struct xfs_btree_cur	*cur,
+	int			*i)
+{
+	struct xfs_refcount_irec	irec;
+	int			found_rec;
+	int			error;
+
+	error = xfs_refcount_get_rec(cur, &irec, &found_rec);
+	if (error)
+		goto out_error;
+	XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, found_rec == 1, out_error);
+	trace_xfs_refcount_delete(cur->bc_mp, cur->bc_private.a.agno, &irec);
+	error = xfs_btree_delete(cur, i);
+	XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, *i == 1, out_error);
+	if (error)
+		goto out_error;
+	error = xfs_refcount_lookup_ge(cur, irec.rc_startblock, &found_rec);
+out_error:
+	if (error)
+		trace_xfs_refcount_delete_error(cur->bc_mp,
+				cur->bc_private.a.agno, error, _RET_IP_);
+	return error;
+}
diff --git a/libxfs/xfs_refcount.h b/libxfs/xfs_refcount.h
new file mode 100644
index 0000000..4dc335a
--- /dev/null
+++ b/libxfs/xfs_refcount.h
@@ -0,0 +1,30 @@
+/*
+ * Copyright (C) 2016 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#ifndef __XFS_REFCOUNT_H__
+#define __XFS_REFCOUNT_H__
+
+extern int xfs_refcount_lookup_le(struct xfs_btree_cur *cur,
+		xfs_agblock_t bno, int *stat);
+extern int xfs_refcount_lookup_ge(struct xfs_btree_cur *cur,
+		xfs_agblock_t bno, int *stat);
+extern int xfs_refcount_get_rec(struct xfs_btree_cur *cur,
+		struct xfs_refcount_irec *irec, int *stat);
+
+#endif	/* __XFS_REFCOUNT_H__ */
diff --git a/libxfs/xfs_refcount_btree.c b/libxfs/xfs_refcount_btree.c
index a7b99e4..04f8db0 100644
--- a/libxfs/xfs_refcount_btree.c
+++ b/libxfs/xfs_refcount_btree.c
@@ -33,6 +33,7 @@
 #include "xfs_cksum.h"
 #include "xfs_trans.h"
 #include "xfs_bit.h"
+#include "xfs_rmap.h"
 
 static struct xfs_btree_cur *
 xfs_refcountbt_dup_cursor(
@@ -43,6 +44,160 @@ xfs_refcountbt_dup_cursor(
 			cur->bc_private.a.dfops);
 }
 
+STATIC void
+xfs_refcountbt_set_root(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_ptr	*ptr,
+	int			inc)
+{
+	struct xfs_buf		*agbp = cur->bc_private.a.agbp;
+	struct xfs_agf		*agf = XFS_BUF_TO_AGF(agbp);
+	xfs_agnumber_t		seqno = be32_to_cpu(agf->agf_seqno);
+	struct xfs_perag	*pag = xfs_perag_get(cur->bc_mp, seqno);
+
+	ASSERT(ptr->s != 0);
+
+	agf->agf_refcount_root = ptr->s;
+	be32_add_cpu(&agf->agf_refcount_level, inc);
+	pag->pagf_refcount_level += inc;
+	xfs_perag_put(pag);
+
+	xfs_alloc_log_agf(cur->bc_tp, agbp,
+			XFS_AGF_REFCOUNT_ROOT | XFS_AGF_REFCOUNT_LEVEL);
+}
+
+STATIC int
+xfs_refcountbt_alloc_block(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_ptr	*start,
+	union xfs_btree_ptr	*new,
+	int			*stat)
+{
+	struct xfs_buf		*agbp = cur->bc_private.a.agbp;
+	struct xfs_agf		*agf = XFS_BUF_TO_AGF(agbp);
+	struct xfs_alloc_arg	args;		/* block allocation args */
+	int			error;		/* error return value */
+
+	memset(&args, 0, sizeof(args));
+	args.tp = cur->bc_tp;
+	args.mp = cur->bc_mp;
+	args.type = XFS_ALLOCTYPE_NEAR_BNO;
+	args.fsbno = XFS_AGB_TO_FSB(cur->bc_mp, cur->bc_private.a.agno,
+			xfs_refc_block(args.mp));
+	args.firstblock = args.fsbno;
+	xfs_rmap_ag_owner(&args.oinfo, XFS_RMAP_OWN_REFC);
+	args.minlen = args.maxlen = args.prod = 1;
+
+	error = xfs_alloc_vextent(&args);
+	if (error)
+		goto out_error;
+	trace_xfs_refcountbt_alloc_block(cur->bc_mp, cur->bc_private.a.agno,
+			args.agbno, 1);
+	if (args.fsbno == NULLFSBLOCK) {
+		XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+		*stat = 0;
+		return 0;
+	}
+	ASSERT(args.agno == cur->bc_private.a.agno);
+	ASSERT(args.len == 1);
+
+	new->s = cpu_to_be32(args.agbno);
+	be32_add_cpu(&agf->agf_refcount_blocks, 1);
+	xfs_alloc_log_agf(cur->bc_tp, agbp, XFS_AGF_REFCOUNT_BLOCKS);
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+	*stat = 1;
+	return 0;
+
+out_error:
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
+	return error;
+}
+
+STATIC int
+xfs_refcountbt_free_block(
+	struct xfs_btree_cur	*cur,
+	struct xfs_buf		*bp)
+{
+	struct xfs_mount	*mp = cur->bc_mp;
+	struct xfs_buf		*agbp = cur->bc_private.a.agbp;
+	struct xfs_agf		*agf = XFS_BUF_TO_AGF(agbp);
+	xfs_fsblock_t		fsbno = XFS_DADDR_TO_FSB(mp, XFS_BUF_ADDR(bp));
+	struct xfs_owner_info	oinfo;
+
+	trace_xfs_refcountbt_free_block(cur->bc_mp, cur->bc_private.a.agno,
+			XFS_FSB_TO_AGBNO(cur->bc_mp, fsbno), 1);
+	xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_REFC);
+	be32_add_cpu(&agf->agf_refcount_blocks, -1);
+	xfs_alloc_log_agf(cur->bc_tp, agbp, XFS_AGF_REFCOUNT_BLOCKS);
+	xfs_bmap_add_free(mp, cur->bc_private.a.dfops, fsbno, 1,
+			&oinfo);
+
+	return 0;
+}
+
+STATIC int
+xfs_refcountbt_get_minrecs(
+	struct xfs_btree_cur	*cur,
+	int			level)
+{
+	return cur->bc_mp->m_refc_mnr[level != 0];
+}
+
+STATIC int
+xfs_refcountbt_get_maxrecs(
+	struct xfs_btree_cur	*cur,
+	int			level)
+{
+	return cur->bc_mp->m_refc_mxr[level != 0];
+}
+
+STATIC void
+xfs_refcountbt_init_key_from_rec(
+	union xfs_btree_key	*key,
+	union xfs_btree_rec	*rec)
+{
+	ASSERT(rec->refc.rc_startblock != 0);
+
+	key->refc.rc_startblock = rec->refc.rc_startblock;
+}
+
+STATIC void
+xfs_refcountbt_init_rec_from_cur(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_rec	*rec)
+{
+	ASSERT(cur->bc_rec.rc.rc_startblock != 0);
+
+	rec->refc.rc_startblock = cpu_to_be32(cur->bc_rec.rc.rc_startblock);
+	rec->refc.rc_blockcount = cpu_to_be32(cur->bc_rec.rc.rc_blockcount);
+	rec->refc.rc_refcount = cpu_to_be32(cur->bc_rec.rc.rc_refcount);
+}
+
+STATIC void
+xfs_refcountbt_init_ptr_from_cur(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_ptr	*ptr)
+{
+	struct xfs_agf		*agf = XFS_BUF_TO_AGF(cur->bc_private.a.agbp);
+
+	ASSERT(cur->bc_private.a.agno == be32_to_cpu(agf->agf_seqno));
+	ASSERT(agf->agf_refcount_root != 0);
+
+	ptr->s = agf->agf_refcount_root;
+}
+
+STATIC __int64_t
+xfs_refcountbt_key_diff(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_key	*key)
+{
+	struct xfs_refcount_irec	*rec = &cur->bc_rec.rc;
+	struct xfs_refcount_key		*kp = &key->refc;
+
+	return (__int64_t)be32_to_cpu(kp->rc_startblock) - rec->rc_startblock;
+}
+
 STATIC bool
 xfs_refcountbt_verify(
 	struct xfs_buf		*bp)
@@ -105,12 +260,62 @@ const struct xfs_buf_ops xfs_refcountbt_buf_ops = {
 	.verify_write		= xfs_refcountbt_write_verify,
 };
 
+#if defined(DEBUG) || defined(XFS_WARN)
+STATIC int
+xfs_refcountbt_keys_inorder(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_key	*k1,
+	union xfs_btree_key	*k2)
+{
+	return be32_to_cpu(k1->refc.rc_startblock) <
+	       be32_to_cpu(k2->refc.rc_startblock);
+}
+
+STATIC int
+xfs_refcountbt_recs_inorder(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_rec	*r1,
+	union xfs_btree_rec	*r2)
+{
+	struct xfs_refcount_irec	a, b;
+
+	int ret = be32_to_cpu(r1->refc.rc_startblock) +
+		be32_to_cpu(r1->refc.rc_blockcount) <=
+		be32_to_cpu(r2->refc.rc_startblock);
+	if (!ret) {
+		a.rc_startblock = be32_to_cpu(r1->refc.rc_startblock);
+		a.rc_blockcount = be32_to_cpu(r1->refc.rc_blockcount);
+		a.rc_refcount = be32_to_cpu(r1->refc.rc_refcount);
+		b.rc_startblock = be32_to_cpu(r2->refc.rc_startblock);
+		b.rc_blockcount = be32_to_cpu(r2->refc.rc_blockcount);
+		b.rc_refcount = be32_to_cpu(r2->refc.rc_refcount);
+		trace_xfs_refcount_rec_order_error(cur->bc_mp,
+				cur->bc_private.a.agno, &a, &b);
+	}
+
+	return ret;
+}
+#endif	/* DEBUG */
+
 static const struct xfs_btree_ops xfs_refcountbt_ops = {
 	.rec_len		= sizeof(struct xfs_refcount_rec),
 	.key_len		= sizeof(struct xfs_refcount_key),
 
 	.dup_cursor		= xfs_refcountbt_dup_cursor,
+	.set_root		= xfs_refcountbt_set_root,
+	.alloc_block		= xfs_refcountbt_alloc_block,
+	.free_block		= xfs_refcountbt_free_block,
+	.get_minrecs		= xfs_refcountbt_get_minrecs,
+	.get_maxrecs		= xfs_refcountbt_get_maxrecs,
+	.init_key_from_rec	= xfs_refcountbt_init_key_from_rec,
+	.init_rec_from_cur	= xfs_refcountbt_init_rec_from_cur,
+	.init_ptr_from_cur	= xfs_refcountbt_init_ptr_from_cur,
+	.key_diff		= xfs_refcountbt_key_diff,
 	.buf_ops		= &xfs_refcountbt_buf_ops,
+#if defined(DEBUG) || defined(XFS_WARN)
+	.keys_inorder		= xfs_refcountbt_keys_inorder,
+	.recs_inorder		= xfs_refcountbt_recs_inorder,
+#endif
 };
 
 /*

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 12/71] xfs: create refcount update intent log items

[Darrick J. Wong]

Create refcount update intent/done log items to record redo
information in the log.  Because we need to roll transactions between
updating the bmbt mapping and updating the reverse mapping, we also
have to track the status of the metadata updates that will be recorded
in the post-roll transactions, just in case we crash before committing
the final transaction.  This mechanism enables log recovery to finish
what was already started.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_log_format.h |   53 +++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 51 insertions(+), 2 deletions(-)


diff --git a/libxfs/xfs_log_format.h b/libxfs/xfs_log_format.h
index 96c7010..ebf5dc0 100644
--- a/libxfs/xfs_log_format.h
+++ b/libxfs/xfs_log_format.h
@@ -112,7 +112,9 @@ static inline uint xlog_get_cycle(char *ptr)
 #define XLOG_REG_TYPE_ICREATE		20
 #define XLOG_REG_TYPE_RUI_FORMAT	21
 #define XLOG_REG_TYPE_RUD_FORMAT	22
-#define XLOG_REG_TYPE_MAX		22
+#define XLOG_REG_TYPE_CUI_FORMAT	23
+#define XLOG_REG_TYPE_CUD_FORMAT	24
+#define XLOG_REG_TYPE_MAX		24
 
 /*
  * Flags to log operation header
@@ -231,6 +233,8 @@ typedef struct xfs_trans_header {
 #define	XFS_LI_ICREATE		0x123f
 #define	XFS_LI_RUI		0x1240	/* rmap update intent */
 #define	XFS_LI_RUD		0x1241
+#define	XFS_LI_CUI		0x1242	/* refcount update intent */
+#define	XFS_LI_CUD		0x1243
 
 #define XFS_LI_TYPE_DESC \
 	{ XFS_LI_EFI,		"XFS_LI_EFI" }, \
@@ -242,7 +246,9 @@ typedef struct xfs_trans_header {
 	{ XFS_LI_QUOTAOFF,	"XFS_LI_QUOTAOFF" }, \
 	{ XFS_LI_ICREATE,	"XFS_LI_ICREATE" }, \
 	{ XFS_LI_RUI,		"XFS_LI_RUI" }, \
-	{ XFS_LI_RUD,		"XFS_LI_RUD" }
+	{ XFS_LI_RUD,		"XFS_LI_RUD" }, \
+	{ XFS_LI_CUI,		"XFS_LI_CUI" }, \
+	{ XFS_LI_CUD,		"XFS_LI_CUD" }
 
 /*
  * Inode Log Item Format definitions.
@@ -663,6 +669,49 @@ struct xfs_rud_log_format {
 };
 
 /*
+ * CUI/CUD (refcount update) log format definitions
+ */
+struct xfs_phys_extent {
+	__uint64_t		pe_startblock;
+	__uint32_t		pe_len;
+	__uint32_t		pe_flags;
+};
+
+/* refcount pe_flags: upper bits are flags, lower byte is type code */
+#define XFS_REFCOUNT_EXTENT_INCREASE	1
+#define XFS_REFCOUNT_EXTENT_DECREASE	2
+#define XFS_REFCOUNT_EXTENT_ALLOC_COW	3
+#define XFS_REFCOUNT_EXTENT_FREE_COW	4
+#define XFS_REFCOUNT_EXTENT_TYPE_MASK	0xFF
+
+#define XFS_REFCOUNT_EXTENT_FLAGS	(XFS_REFCOUNT_EXTENT_TYPE_MASK)
+
+/*
+ * This is the structure used to lay out a cui log item in the
+ * log.  The cui_extents field is a variable size array whose
+ * size is given by cui_nextents.
+ */
+struct xfs_cui_log_format {
+	__uint16_t		cui_type;	/* cui log item type */
+	__uint16_t		cui_size;	/* size of this item */
+	__uint32_t		cui_nextents;	/* # extents to free */
+	__uint64_t		cui_id;		/* cui identifier */
+	struct xfs_phys_extent	cui_extents[1];	/* array of extents */
+};
+
+/*
+ * This is the structure used to lay out a cud log item in the
+ * log.  The cud_extents array is a variable size array whose
+ * size is given by cud_nextents;
+ */
+struct xfs_cud_log_format {
+	__uint16_t		cud_type;	/* cud log item type */
+	__uint16_t		cud_size;	/* size of this item */
+	__uint32_t		__pad;
+	__uint64_t		cud_cui_id;	/* id of corresponding cui */
+};
+
+/*
  * Dquot Log format definitions.
  *
  * The first two fields must be the type and size fitting into

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 13/71] xfs: log refcount intent items

[Darrick J. Wong]

Provide a mechanism for higher levels to create CUI/CUD items, submit
them to the log, and a stub function to deal with recovered CUI items.
These parts will be connected to the refcountbt in a later patch.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_refcount.h |   14 ++++++++++++++
 1 file changed, 14 insertions(+)


diff --git a/libxfs/xfs_refcount.h b/libxfs/xfs_refcount.h
index 4dc335a..2ef2b28 100644
--- a/libxfs/xfs_refcount.h
+++ b/libxfs/xfs_refcount.h
@@ -27,4 +27,18 @@ extern int xfs_refcount_lookup_ge(struct xfs_btree_cur *cur,
 extern int xfs_refcount_get_rec(struct xfs_btree_cur *cur,
 		struct xfs_refcount_irec *irec, int *stat);
 
+enum xfs_refcount_intent_type {
+	XFS_REFCOUNT_INCREASE,
+	XFS_REFCOUNT_DECREASE,
+	XFS_REFCOUNT_ALLOC_COW,
+	XFS_REFCOUNT_FREE_COW,
+};
+
+struct xfs_refcount_intent {
+	struct list_head			ri_list;
+	enum xfs_refcount_intent_type		ri_type;
+	xfs_fsblock_t				ri_startblock;
+	xfs_extlen_t				ri_blockcount;
+};
+
 #endif	/* __XFS_REFCOUNT_H__ */

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 14/71] xfs: adjust refcount of an extent of blocks in refcount btree

[Darrick J. Wong]

Provide functions to adjust the reference counts for an extent of
physical blocks stored in the refcount btree.

v2: Refactor the left/right split code into a single function.  Track
the number of btree shape changes and record updates during a refcount
update so that we can decide if we need to get a fresh transaction to
continue.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 include/xfs_trace.h   |   20 +
 libxfs/xfs_refcount.c |  783 +++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 803 insertions(+)


diff --git a/include/xfs_trace.h b/include/xfs_trace.h
index 937d5f5..5605c3c 100644
--- a/include/xfs_trace.h
+++ b/include/xfs_trace.h
@@ -235,6 +235,26 @@
 #define trace_xfs_refcountbt_free_block(...)	((void) 0)
 #define trace_xfs_refcountbt_alloc_block(...)	((void) 0)
 #define trace_xfs_refcount_rec_order_error(...)	((void) 0)
+#define trace_xfs_refcount_split_extent(...)	((void) 0)
+#define trace_xfs_refcount_split_extent_error(...)	((void) 0)
+#define trace_xfs_refcount_merge_center_extents_error(...)	((void) 0)
+#define trace_xfs_refcount_merge_left_extent_error(...)	((void) 0)
+#define trace_xfs_refcount_merge_right_extent_error(...)	((void) 0)
+#define trace_xfs_refcount_find_left_extent(...)	((void) 0)
+#define trace_xfs_refcount_find_left_extent_error(...)	((void) 0)
+#define trace_xfs_refcount_find_right_extent(...)	((void) 0)
+#define trace_xfs_refcount_find_right_extent_error(...)	((void) 0)
+#define trace_xfs_refcount_merge_center_extents(...)	((void) 0)
+#define trace_xfs_refcount_merge_left_extent(...)	((void) 0)
+#define trace_xfs_refcount_merge_right_extent(...)	((void) 0)
+#define trace_xfs_refcount_modify_extent(...)	((void) 0)
+#define trace_xfs_refcount_modify_extent_error(...)	((void) 0)
+#define trace_xfs_refcount_adjust_error(...)	((void) 0)
+#define trace_xfs_refcount_increase(...)	((void) 0)
+#define trace_xfs_refcount_decrease(...)	((void) 0)
+#define trace_xfs_refcount_deferred(...)	((void) 0)
+#define trace_xfs_refcount_defer(...)		((void) 0)
+#define trace_xfs_refcount_finish_one_leftover(...)	((void) 0)
 
 /* set c = c to avoid unused var warnings */
 #define trace_xfs_perag_get(a,b,c,d)	((c) = (c))
diff --git a/libxfs/xfs_refcount.c b/libxfs/xfs_refcount.c
index 8b522e2..08866f8 100644
--- a/libxfs/xfs_refcount.c
+++ b/libxfs/xfs_refcount.c
@@ -36,6 +36,12 @@
 #include "xfs_bit.h"
 #include "xfs_refcount.h"
 
+/* Allowable refcount adjustment amounts. */
+enum xfs_refc_adjust_op {
+	XFS_REFCOUNT_ADJUST_INCREASE	= 1,
+	XFS_REFCOUNT_ADJUST_DECREASE	= -1,
+};
+
 /*
  * Look up the first record less than or equal to [bno, len] in the btree
  * given by cur.
@@ -174,3 +180,780 @@ out_error:
 				cur->bc_private.a.agno, error, _RET_IP_);
 	return error;
 }
+
+/*
+ * Adjusting the Reference Count
+ *
+ * As stated elsewhere, the reference count btree (refcbt) stores
+ * >1 reference counts for extents of physical blocks.  In this
+ * operation, we're either raising or lowering the reference count of
+ * some subrange stored in the tree:
+ *
+ *      <------ adjustment range ------>
+ * ----+   +---+-----+ +--+--------+---------
+ *  2  |   | 3 |  4  | |17|   55   |   10
+ * ----+   +---+-----+ +--+--------+---------
+ * X axis is physical blocks number;
+ * reference counts are the numbers inside the rectangles
+ *
+ * The first thing we need to do is to ensure that there are no
+ * refcount extents crossing either boundary of the range to be
+ * adjusted.  For any extent that does cross a boundary, split it into
+ * two extents so that we can increment the refcount of one of the
+ * pieces later:
+ *
+ *      <------ adjustment range ------>
+ * ----+   +---+-----+ +--+--------+----+----
+ *  2  |   | 3 |  2  | |17|   55   | 10 | 10
+ * ----+   +---+-----+ +--+--------+----+----
+ *
+ * For this next step, let's assume that all the physical blocks in
+ * the adjustment range are mapped to a file and are therefore in use
+ * at least once.  Therefore, we can infer that any gap in the
+ * refcount tree within the adjustment range represents a physical
+ * extent with refcount == 1:
+ *
+ *      <------ adjustment range ------>
+ * ----+---+---+-----+-+--+--------+----+----
+ *  2  |"1"| 3 |  2  |1|17|   55   | 10 | 10
+ * ----+---+---+-----+-+--+--------+----+----
+ *      ^
+ *
+ * For each extent that falls within the interval range, figure out
+ * which extent is to the left or the right of that extent.  Now we
+ * have a left, current, and right extent.  If the new reference count
+ * of the center extent enables us to merge left, center, and right
+ * into one record covering all three, do so.  If the center extent is
+ * at the left end of the range, abuts the left extent, and its new
+ * reference count matches the left extent's record, then merge them.
+ * If the center extent is at the right end of the range, abuts the
+ * right extent, and the reference counts match, merge those.  In the
+ * example, we can left merge (assuming an increment operation):
+ *
+ *      <------ adjustment range ------>
+ * --------+---+-----+-+--+--------+----+----
+ *    2    | 3 |  2  |1|17|   55   | 10 | 10
+ * --------+---+-----+-+--+--------+----+----
+ *          ^
+ *
+ * For all other extents within the range, adjust the reference count
+ * or delete it if the refcount falls below 2.  If we were
+ * incrementing, the end result looks like this:
+ *
+ *      <------ adjustment range ------>
+ * --------+---+-----+-+--+--------+----+----
+ *    2    | 4 |  3  |2|18|   56   | 11 | 10
+ * --------+---+-----+-+--+--------+----+----
+ *
+ * The result of a decrement operation looks as such:
+ *
+ *      <------ adjustment range ------>
+ * ----+   +---+       +--+--------+----+----
+ *  2  |   | 2 |       |16|   54   |  9 | 10
+ * ----+   +---+       +--+--------+----+----
+ *      DDDD    111111DD
+ *
+ * The blocks marked "D" are freed; the blocks marked "1" are only
+ * referenced once and therefore the record is removed from the
+ * refcount btree.
+ */
+
+#define RCNEXT(rc)	((rc).rc_startblock + (rc).rc_blockcount)
+/*
+ * Split a refcount extent that crosses agbno.
+ */
+STATIC int
+xfs_refcount_split_extent(
+	struct xfs_btree_cur		*cur,
+	xfs_agblock_t			agbno,
+	bool				*shape_changed)
+{
+	struct xfs_refcount_irec	rcext, tmp;
+	int				found_rec;
+	int				error;
+
+	*shape_changed = false;
+	error = xfs_refcount_lookup_le(cur, agbno, &found_rec);
+	if (error)
+		goto out_error;
+	if (!found_rec)
+		return 0;
+
+	error = xfs_refcount_get_rec(cur, &rcext, &found_rec);
+	if (error)
+		goto out_error;
+	XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, found_rec == 1, out_error);
+	if (rcext.rc_startblock == agbno || RCNEXT(rcext) <= agbno)
+		return 0;
+
+	*shape_changed = true;
+	trace_xfs_refcount_split_extent(cur->bc_mp, cur->bc_private.a.agno,
+			&rcext, agbno);
+
+	/* Establish the right extent. */
+	tmp = rcext;
+	tmp.rc_startblock = agbno;
+	tmp.rc_blockcount -= (agbno - rcext.rc_startblock);
+	error = xfs_refcount_update(cur, &tmp);
+	if (error)
+		goto out_error;
+
+	/* Insert the left extent. */
+	tmp = rcext;
+	tmp.rc_blockcount = agbno - rcext.rc_startblock;
+	error = xfs_refcount_insert(cur, &tmp, &found_rec);
+	if (error)
+		goto out_error;
+	XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, found_rec == 1, out_error);
+	return error;
+
+out_error:
+	trace_xfs_refcount_split_extent_error(cur->bc_mp,
+			cur->bc_private.a.agno, error, _RET_IP_);
+	return error;
+}
+
+/*
+ * Merge the left, center, and right extents.
+ */
+STATIC int
+xfs_refcount_merge_center_extent(
+	struct xfs_btree_cur		*cur,
+	struct xfs_refcount_irec	*left,
+	struct xfs_refcount_irec	*center,
+	unsigned long long		extlen,
+	xfs_agblock_t			*agbno,
+	xfs_extlen_t			*aglen)
+{
+	int				error;
+	int				found_rec;
+
+	error = xfs_refcount_lookup_ge(cur, center->rc_startblock,
+			&found_rec);
+	if (error)
+		goto out_error;
+	XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, found_rec == 1, out_error);
+
+	error = xfs_refcount_delete(cur, &found_rec);
+	if (error)
+		goto out_error;
+	XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, found_rec == 1, out_error);
+
+	if (center->rc_refcount > 1) {
+		error = xfs_refcount_delete(cur, &found_rec);
+		if (error)
+			goto out_error;
+		XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, found_rec == 1,
+				out_error);
+	}
+
+	error = xfs_refcount_lookup_le(cur, left->rc_startblock,
+			&found_rec);
+	if (error)
+		goto out_error;
+	XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, found_rec == 1, out_error);
+
+	left->rc_blockcount = extlen;
+	error = xfs_refcount_update(cur, left);
+	if (error)
+		goto out_error;
+
+	*aglen = 0;
+	return error;
+
+out_error:
+	trace_xfs_refcount_merge_center_extents_error(cur->bc_mp,
+			cur->bc_private.a.agno, error, _RET_IP_);
+	return error;
+}
+
+/*
+ * Merge with the left extent.
+ */
+STATIC int
+xfs_refcount_merge_left_extent(
+	struct xfs_btree_cur		*cur,
+	struct xfs_refcount_irec	*left,
+	struct xfs_refcount_irec	*cleft,
+	xfs_agblock_t			*agbno,
+	xfs_extlen_t			*aglen)
+{
+	int				error;
+	int				found_rec;
+
+	if (cleft->rc_refcount > 1) {
+		error = xfs_refcount_lookup_le(cur, cleft->rc_startblock,
+				&found_rec);
+		if (error)
+			goto out_error;
+		XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, found_rec == 1,
+				out_error);
+
+		error = xfs_refcount_delete(cur, &found_rec);
+		if (error)
+			goto out_error;
+		XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, found_rec == 1,
+				out_error);
+	}
+
+	error = xfs_refcount_lookup_le(cur, left->rc_startblock,
+			&found_rec);
+	if (error)
+		goto out_error;
+	XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, found_rec == 1, out_error);
+
+	left->rc_blockcount += cleft->rc_blockcount;
+	error = xfs_refcount_update(cur, left);
+	if (error)
+		goto out_error;
+
+	*agbno += cleft->rc_blockcount;
+	*aglen -= cleft->rc_blockcount;
+	return error;
+
+out_error:
+	trace_xfs_refcount_merge_left_extent_error(cur->bc_mp,
+			cur->bc_private.a.agno, error, _RET_IP_);
+	return error;
+}
+
+/*
+ * Merge with the right extent.
+ */
+STATIC int
+xfs_refcount_merge_right_extent(
+	struct xfs_btree_cur		*cur,
+	struct xfs_refcount_irec	*right,
+	struct xfs_refcount_irec	*cright,
+	xfs_agblock_t			*agbno,
+	xfs_extlen_t			*aglen)
+{
+	int				error;
+	int				found_rec;
+
+	if (cright->rc_refcount > 1) {
+		error = xfs_refcount_lookup_le(cur, cright->rc_startblock,
+			&found_rec);
+		if (error)
+			goto out_error;
+		XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, found_rec == 1,
+				out_error);
+
+		error = xfs_refcount_delete(cur, &found_rec);
+		if (error)
+			goto out_error;
+		XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, found_rec == 1,
+				out_error);
+	}
+
+	error = xfs_refcount_lookup_le(cur, right->rc_startblock,
+			&found_rec);
+	if (error)
+		goto out_error;
+	XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, found_rec == 1, out_error);
+
+	right->rc_startblock -= cright->rc_blockcount;
+	right->rc_blockcount += cright->rc_blockcount;
+	error = xfs_refcount_update(cur, right);
+	if (error)
+		goto out_error;
+
+	*aglen -= cright->rc_blockcount;
+	return error;
+
+out_error:
+	trace_xfs_refcount_merge_right_extent_error(cur->bc_mp,
+			cur->bc_private.a.agno, error, _RET_IP_);
+	return error;
+}
+
+/*
+ * Find the left extent and the one after it (cleft).  This function assumes
+ * that we've already split any extent crossing agbno.
+ */
+STATIC int
+xfs_refcount_find_left_extents(
+	struct xfs_btree_cur		*cur,
+	struct xfs_refcount_irec	*left,
+	struct xfs_refcount_irec	*cleft,
+	xfs_agblock_t			agbno,
+	xfs_extlen_t			aglen)
+{
+	struct xfs_refcount_irec	tmp;
+	int				error;
+	int				found_rec;
+
+	left->rc_blockcount = cleft->rc_blockcount = 0;
+	error = xfs_refcount_lookup_le(cur, agbno - 1, &found_rec);
+	if (error)
+		goto out_error;
+	if (!found_rec)
+		return 0;
+
+	error = xfs_refcount_get_rec(cur, &tmp, &found_rec);
+	if (error)
+		goto out_error;
+	XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, found_rec == 1, out_error);
+
+	if (RCNEXT(tmp) != agbno)
+		return 0;
+	/* We have a left extent; retrieve (or invent) the next right one */
+	*left = tmp;
+
+	error = xfs_btree_increment(cur, 0, &found_rec);
+	if (error)
+		goto out_error;
+	if (found_rec) {
+		error = xfs_refcount_get_rec(cur, &tmp, &found_rec);
+		if (error)
+			goto out_error;
+		XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, found_rec == 1,
+				out_error);
+
+		/* if tmp starts at the end of our range, just use that */
+		if (tmp.rc_startblock == agbno)
+			*cleft = tmp;
+		else {
+			/*
+			 * There's a gap in the refcntbt at the start of the
+			 * range we're interested in (refcount == 1) so
+			 * create the implied extent and pass it back.
+			 */
+			cleft->rc_startblock = agbno;
+			cleft->rc_blockcount = min(aglen,
+					tmp.rc_startblock - agbno);
+			cleft->rc_refcount = 1;
+		}
+	} else {
+		/*
+		 * No extents, so pretend that there's one covering the whole
+		 * range.
+		 */
+		cleft->rc_startblock = agbno;
+		cleft->rc_blockcount = aglen;
+		cleft->rc_refcount = 1;
+	}
+	trace_xfs_refcount_find_left_extent(cur->bc_mp, cur->bc_private.a.agno,
+			left, cleft, agbno);
+	return error;
+
+out_error:
+	trace_xfs_refcount_find_left_extent_error(cur->bc_mp,
+			cur->bc_private.a.agno, error, _RET_IP_);
+	return error;
+}
+
+/*
+ * Find the right extent and the one before it (cright).  This function
+ * assumes that we've already split any extents crossing agbno + aglen.
+ */
+STATIC int
+xfs_refcount_find_right_extents(
+	struct xfs_btree_cur		*cur,
+	struct xfs_refcount_irec	*right,
+	struct xfs_refcount_irec	*cright,
+	xfs_agblock_t			agbno,
+	xfs_extlen_t			aglen)
+{
+	struct xfs_refcount_irec	tmp;
+	int				error;
+	int				found_rec;
+
+	right->rc_blockcount = cright->rc_blockcount = 0;
+	error = xfs_refcount_lookup_ge(cur, agbno + aglen, &found_rec);
+	if (error)
+		goto out_error;
+	if (!found_rec)
+		return 0;
+
+	error = xfs_refcount_get_rec(cur, &tmp, &found_rec);
+	if (error)
+		goto out_error;
+	XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, found_rec == 1, out_error);
+
+	if (tmp.rc_startblock != agbno + aglen)
+		return 0;
+	/* We have a right extent; retrieve (or invent) the next left one */
+	*right = tmp;
+
+	error = xfs_btree_decrement(cur, 0, &found_rec);
+	if (error)
+		goto out_error;
+	if (found_rec) {
+		error = xfs_refcount_get_rec(cur, &tmp, &found_rec);
+		if (error)
+			goto out_error;
+		XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, found_rec == 1,
+				out_error);
+
+		/* if tmp ends at the end of our range, just use that */
+		if (RCNEXT(tmp) == agbno + aglen)
+			*cright = tmp;
+		else {
+			/*
+			 * There's a gap in the refcntbt at the end of the
+			 * range we're interested in (refcount == 1) so
+			 * create the implied extent and pass it back.
+			 */
+			cright->rc_startblock = max(agbno, RCNEXT(tmp));
+			cright->rc_blockcount = right->rc_startblock -
+					cright->rc_startblock;
+			cright->rc_refcount = 1;
+		}
+	} else {
+		/*
+		 * No extents, so pretend that there's one covering the whole
+		 * range.
+		 */
+		cright->rc_startblock = agbno;
+		cright->rc_blockcount = aglen;
+		cright->rc_refcount = 1;
+	}
+	trace_xfs_refcount_find_right_extent(cur->bc_mp, cur->bc_private.a.agno,
+			cright, right, agbno + aglen);
+	return error;
+
+out_error:
+	trace_xfs_refcount_find_right_extent_error(cur->bc_mp,
+			cur->bc_private.a.agno, error, _RET_IP_);
+	return error;
+}
+#undef RCNEXT
+
+/*
+ * Try to merge with any extents on the boundaries of the adjustment range.
+ */
+STATIC int
+xfs_refcount_merge_extents(
+	struct xfs_btree_cur	*cur,
+	xfs_agblock_t		*agbno,
+	xfs_extlen_t		*aglen,
+	enum xfs_refc_adjust_op adjust,
+	bool			*shape_changed)
+{
+	struct xfs_refcount_irec	left = {0}, cleft = {0};
+	struct xfs_refcount_irec	cright = {0}, right = {0};
+	int				error;
+	unsigned long long		ulen;
+	bool				cequal;
+
+	*shape_changed = false;
+	/*
+	 * Find the extent just below agbno [left], just above agbno [cleft],
+	 * just below (agbno + aglen) [cright], and just above (agbno + aglen)
+	 * [right].
+	 */
+	error = xfs_refcount_find_left_extents(cur, &left, &cleft, *agbno,
+			*aglen);
+	if (error)
+		return error;
+	error = xfs_refcount_find_right_extents(cur, &right, &cright, *agbno,
+			*aglen);
+	if (error)
+		return error;
+
+	/* No left or right extent to merge; exit. */
+	if (left.rc_blockcount == 0 && right.rc_blockcount == 0)
+		return 0;
+
+	*shape_changed = true;
+	cequal = (cleft.rc_startblock == cright.rc_startblock) &&
+		 (cleft.rc_blockcount == cright.rc_blockcount);
+
+	/* Try to merge left, cleft, and right.  cleft must == cright. */
+	ulen = (unsigned long long)left.rc_blockcount + cleft.rc_blockcount +
+			right.rc_blockcount;
+	if (left.rc_blockcount != 0 && right.rc_blockcount != 0 &&
+	    cleft.rc_blockcount != 0 && cright.rc_blockcount != 0 &&
+	    cequal &&
+	    left.rc_refcount == cleft.rc_refcount + adjust &&
+	    right.rc_refcount == cleft.rc_refcount + adjust &&
+	    ulen < MAXREFCEXTLEN) {
+		trace_xfs_refcount_merge_center_extents(cur->bc_mp,
+				cur->bc_private.a.agno, &left, &cleft, &right);
+		return xfs_refcount_merge_center_extent(cur, &left, &cleft,
+				ulen, agbno, aglen);
+	}
+
+	/* Try to merge left and cleft. */
+	ulen = (unsigned long long)left.rc_blockcount + cleft.rc_blockcount;
+	if (left.rc_blockcount != 0 && cleft.rc_blockcount != 0 &&
+	    left.rc_refcount == cleft.rc_refcount + adjust &&
+	    ulen < MAXREFCEXTLEN) {
+		trace_xfs_refcount_merge_left_extent(cur->bc_mp,
+				cur->bc_private.a.agno, &left, &cleft);
+		error = xfs_refcount_merge_left_extent(cur, &left, &cleft,
+				agbno, aglen);
+		if (error)
+			return error;
+
+		/*
+		 * If we just merged left + cleft and cleft == cright,
+		 * we no longer have a cright to merge with right.  We're done.
+		 */
+		if (cequal)
+			return 0;
+	}
+
+	/* Try to merge cright and right. */
+	ulen = (unsigned long long)right.rc_blockcount + cright.rc_blockcount;
+	if (right.rc_blockcount != 0 && cright.rc_blockcount != 0 &&
+	    right.rc_refcount == cright.rc_refcount + adjust &&
+	    ulen < MAXREFCEXTLEN) {
+		trace_xfs_refcount_merge_right_extent(cur->bc_mp,
+				cur->bc_private.a.agno, &cright, &right);
+		return xfs_refcount_merge_right_extent(cur, &right, &cright,
+				agbno, aglen);
+	}
+
+	return error;
+}
+
+/*
+ * While we're adjusting the refcounts records of an extent, we have
+ * to keep an eye on the number of extents we're dirtying -- run too
+ * many in a single transaction and we'll exceed the transaction's
+ * reservation and crash the fs.  Each record adds 12 bytes to the
+ * log (plus any key updates) so we'll conservatively assume 24 bytes
+ * per record.  We must also leave space for btree splits on both ends
+ * of the range and space for the CUD and a new CUI.
+ *
+ * XXX: This is a pretty hand-wavy estimate.  The penalty for guessing
+ * true incorrectly is a shutdown FS; the penalty for guessing false
+ * incorrectly is more transaction rolls than might be necessary.
+ * Be conservative here.
+ */
+static bool
+xfs_refcount_still_have_space(
+	struct xfs_btree_cur		*cur)
+{
+	unsigned long			overhead;
+
+	overhead = cur->bc_private.a.priv.refc.shape_changes *
+			xfs_allocfree_log_count(cur->bc_mp, 1);
+	overhead *= cur->bc_mp->m_sb.sb_blocksize;
+
+	/*
+	 * Only allow 2 refcount extent updates per transaction if the
+	 * refcount continue update "error" has been injected.
+	 */
+	if (cur->bc_private.a.priv.refc.nr_ops > 2 &&
+	    XFS_TEST_ERROR(false, cur->bc_mp,
+			XFS_ERRTAG_REFCOUNT_CONTINUE_UPDATE,
+			XFS_RANDOM_REFCOUNT_CONTINUE_UPDATE))
+		return false;
+
+	if (cur->bc_private.a.priv.refc.nr_ops == 0)
+		return true;
+	else if (overhead > cur->bc_tp->t_log_res)
+		return false;
+	return  cur->bc_tp->t_log_res - overhead >
+		cur->bc_private.a.priv.refc.nr_ops * 32;
+}
+
+/*
+ * Adjust the refcounts of middle extents.  At this point we should have
+ * split extents that crossed the adjustment range; merged with adjacent
+ * extents; and updated agbno/aglen to reflect the merges.  Therefore,
+ * all we have to do is update the extents inside [agbno, agbno + aglen].
+ */
+STATIC int
+xfs_refcount_adjust_extents(
+	struct xfs_btree_cur	*cur,
+	xfs_agblock_t		agbno,
+	xfs_extlen_t		aglen,
+	xfs_extlen_t		*adjusted,
+	enum xfs_refc_adjust_op	adj,
+	struct xfs_defer_ops	*dfops,
+	struct xfs_owner_info	*oinfo)
+{
+	struct xfs_refcount_irec	ext, tmp;
+	int				error;
+	int				found_rec, found_tmp;
+	xfs_fsblock_t			fsbno;
+
+	/* Merging did all the work already. */
+	if (aglen == 0)
+		return 0;
+
+	error = xfs_refcount_lookup_ge(cur, agbno, &found_rec);
+	if (error)
+		goto out_error;
+
+	while (aglen > 0 && xfs_refcount_still_have_space(cur)) {
+		error = xfs_refcount_get_rec(cur, &ext, &found_rec);
+		if (error)
+			goto out_error;
+		if (!found_rec) {
+			ext.rc_startblock = cur->bc_mp->m_sb.sb_agblocks;
+			ext.rc_blockcount = 0;
+			ext.rc_refcount = 0;
+		}
+
+		/*
+		 * Deal with a hole in the refcount tree; if a file maps to
+		 * these blocks and there's no refcountbt recourd, pretend that
+		 * there is one with refcount == 1.
+		 */
+		if (ext.rc_startblock != agbno) {
+			tmp.rc_startblock = agbno;
+			tmp.rc_blockcount = min(aglen,
+					ext.rc_startblock - agbno);
+			tmp.rc_refcount = 1 + adj;
+			trace_xfs_refcount_modify_extent(cur->bc_mp,
+					cur->bc_private.a.agno, &tmp);
+
+			/*
+			 * Either cover the hole (increment) or
+			 * delete the range (decrement).
+			 */
+			if (tmp.rc_refcount) {
+				error = xfs_refcount_insert(cur, &tmp,
+						&found_tmp);
+				if (error)
+					goto out_error;
+				XFS_WANT_CORRUPTED_GOTO(cur->bc_mp,
+						found_tmp == 1, out_error);
+				cur->bc_private.a.priv.refc.nr_ops++;
+			} else {
+				fsbno = XFS_AGB_TO_FSB(cur->bc_mp,
+						cur->bc_private.a.agno,
+						tmp.rc_startblock);
+				xfs_bmap_add_free(cur->bc_mp, dfops, fsbno,
+						tmp.rc_blockcount, oinfo);
+			}
+
+			(*adjusted) += tmp.rc_blockcount;
+			agbno += tmp.rc_blockcount;
+			aglen -= tmp.rc_blockcount;
+
+			error = xfs_refcount_lookup_ge(cur, agbno,
+					&found_rec);
+			if (error)
+				goto out_error;
+		}
+
+		/* Stop if there's nothing left to modify */
+		if (aglen == 0 || !xfs_refcount_still_have_space(cur))
+			break;
+
+		/*
+		 * Adjust the reference count and either update the tree
+		 * (incr) or free the blocks (decr).
+		 */
+		if (ext.rc_refcount == MAXREFCOUNT)
+			goto skip;
+		ext.rc_refcount += adj;
+		trace_xfs_refcount_modify_extent(cur->bc_mp,
+				cur->bc_private.a.agno, &ext);
+		if (ext.rc_refcount > 1) {
+			error = xfs_refcount_update(cur, &ext);
+			if (error)
+				goto out_error;
+			cur->bc_private.a.priv.refc.nr_ops++;
+		} else if (ext.rc_refcount == 1) {
+			error = xfs_refcount_delete(cur, &found_rec);
+			if (error)
+				goto out_error;
+			XFS_WANT_CORRUPTED_GOTO(cur->bc_mp,
+					found_rec == 1, out_error);
+			cur->bc_private.a.priv.refc.nr_ops++;
+			goto advloop;
+		} else {
+			fsbno = XFS_AGB_TO_FSB(cur->bc_mp,
+					cur->bc_private.a.agno,
+					ext.rc_startblock);
+			xfs_bmap_add_free(cur->bc_mp, dfops, fsbno,
+					ext.rc_blockcount, oinfo);
+		}
+
+skip:
+		error = xfs_btree_increment(cur, 0, &found_rec);
+		if (error)
+			goto out_error;
+
+advloop:
+		(*adjusted) += ext.rc_blockcount;
+		agbno += ext.rc_blockcount;
+		aglen -= ext.rc_blockcount;
+	}
+
+	return error;
+out_error:
+	trace_xfs_refcount_modify_extent_error(cur->bc_mp,
+			cur->bc_private.a.agno, error, _RET_IP_);
+	return error;
+}
+
+/* Adjust the reference count of a range of AG blocks. */
+STATIC int
+xfs_refcount_adjust(
+	struct xfs_btree_cur	*cur,
+	xfs_agblock_t		agbno,
+	xfs_extlen_t		aglen,
+	xfs_extlen_t		*adjusted,
+	enum xfs_refc_adjust_op	adj,
+	struct xfs_defer_ops	*dfops,
+	struct xfs_owner_info	*oinfo)
+{
+	xfs_extlen_t		orig_aglen;
+	bool			shape_changed;
+	int			shape_changes = 0;
+	int			error;
+
+	*adjusted = 0;
+	switch (adj) {
+	case XFS_REFCOUNT_ADJUST_INCREASE:
+		trace_xfs_refcount_increase(cur->bc_mp, cur->bc_private.a.agno,
+				agbno, aglen);
+		break;
+	case XFS_REFCOUNT_ADJUST_DECREASE:
+		trace_xfs_refcount_decrease(cur->bc_mp, cur->bc_private.a.agno,
+				agbno, aglen);
+		break;
+	default:
+		ASSERT(0);
+	}
+
+	/*
+	 * Ensure that no rcextents cross the boundary of the adjustment range.
+	 */
+	error = xfs_refcount_split_extent(cur, agbno, &shape_changed);
+	if (error)
+		goto out_error;
+	if (shape_changed)
+		shape_changes++;
+
+	error = xfs_refcount_split_extent(cur, agbno + aglen, &shape_changed);
+	if (error)
+		goto out_error;
+	if (shape_changed)
+		shape_changes++;
+
+	/*
+	 * Try to merge with the left or right extents of the range.
+	 */
+	orig_aglen = aglen;
+	error = xfs_refcount_merge_extents(cur, &agbno, &aglen, adj,
+			&shape_changed);
+	if (error)
+		goto out_error;
+	if (shape_changed)
+		shape_changes++;
+	(*adjusted) += orig_aglen - aglen;
+	if (shape_changes)
+		cur->bc_private.a.priv.refc.shape_changes++;
+
+	/* Now that we've taken care of the ends, adjust the middle extents */
+	error = xfs_refcount_adjust_extents(cur, agbno, aglen, adjusted, adj,
+			dfops, oinfo);
+	if (error)
+		goto out_error;
+
+	return 0;
+
+out_error:
+	trace_xfs_refcount_adjust_error(cur->bc_mp, cur->bc_private.a.agno,
+			error, _RET_IP_);
+	return error;
+}

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 15/71] xfs: connect refcount adjust functions to upper layers

[Darrick J. Wong]

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>
---
 include/xfs_trans.h   |    1 
 libxfs/defer_item.c   |  129 +++++++++++++++++++++++++++++++++++++
 libxfs/init.c         |    1 
 libxfs/xfs_defer.h    |    1 
 libxfs/xfs_refcount.c |  170 +++++++++++++++++++++++++++++++++++++++++++++++++
 libxfs/xfs_refcount.h |   12 +++
 6 files changed, 314 insertions(+)


diff --git a/include/xfs_trans.h b/include/xfs_trans.h
index ab5d59b..739a792 100644
--- a/include/xfs_trans.h
+++ b/include/xfs_trans.h
@@ -147,5 +147,6 @@ libxfs_trans_read_buf(
 
 void xfs_extent_free_init_defer_op(void);
 void xfs_rmap_update_init_defer_op(void);
+void xfs_refcount_update_init_defer_op(void);
 
 #endif	/* __XFS_TRANS_H__ */
diff --git a/libxfs/defer_item.c b/libxfs/defer_item.c
index f60a11b..133e6e1 100644
--- a/libxfs/defer_item.c
+++ b/libxfs/defer_item.c
@@ -31,6 +31,7 @@
 #include "xfs_bmap.h"
 #include "xfs_alloc.h"
 #include "xfs_rmap.h"
+#include "xfs_refcount.h"
 
 /* Dummy defer item ops, since we don't do logging. */
 
@@ -257,3 +258,131 @@ xfs_rmap_update_init_defer_op(void)
 {
 	xfs_defer_init_op_type(&xfs_rmap_update_defer_type);
 }
+
+/* Reference Counting */
+
+/* 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)
+{
+	return NULL;
+}
+
+/* Log refcount updates in the intent item. */
+STATIC void
+xfs_refcount_update_log_item(
+	struct xfs_trans		*tp,
+	void				*intent,
+	struct list_head		*item)
+{
+}
+
+/* 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 NULL;
+}
+
+/* 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_extlen_t			adjusted;
+	int				error;
+
+	refc = container_of(item, struct xfs_refcount_intent, ri_list);
+	error = xfs_refcount_finish_one(tp, dop,
+			refc->ri_type,
+			refc->ri_startblock,
+			refc->ri_blockcount,
+			&adjusted,
+			(struct xfs_btree_cur **)state);
+	/* Did we run out of reservation?  Requeue what we didn't finish. */
+	if (!error && adjusted < refc->ri_blockcount) {
+		ASSERT(refc->ri_type == XFS_REFCOUNT_INCREASE ||
+		       refc->ri_type == XFS_REFCOUNT_DECREASE);
+		refc->ri_startblock += adjusted;
+		refc->ri_blockcount -= adjusted;
+		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)
+{
+}
+
+/* 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,
+	.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);
+}
diff --git a/libxfs/init.c b/libxfs/init.c
index 706925d..2d1bb58 100644
--- a/libxfs/init.c
+++ b/libxfs/init.c
@@ -266,6 +266,7 @@ libxfs_init(libxfs_init_t *a)
 
 	xfs_extent_free_init_defer_op();
 	xfs_rmap_update_init_defer_op();
+	xfs_refcount_update_init_defer_op();
 
 	radix_tree_init();
 
diff --git a/libxfs/xfs_defer.h b/libxfs/xfs_defer.h
index cc3981c..d47a482 100644
--- a/libxfs/xfs_defer.h
+++ b/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,
diff --git a/libxfs/xfs_refcount.c b/libxfs/xfs_refcount.c
index 08866f8..31028bf 100644
--- a/libxfs/xfs_refcount.c
+++ b/libxfs/xfs_refcount.c
@@ -957,3 +957,173 @@ 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_extlen_t			*adjusted,
+	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;
+	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, adjusted,
+			XFS_REFCOUNT_ADJUST_INCREASE, dfops, NULL);
+		break;
+	case XFS_REFCOUNT_DECREASE:
+		error = xfs_refcount_adjust(rcur, bno, blockcount, adjusted,
+			XFS_REFCOUNT_ADJUST_DECREASE, dfops, NULL);
+		break;
+	default:
+		ASSERT(0);
+		error = -EFSCORRUPTED;
+	}
+	if (!error && *adjusted != blockcount)
+		trace_xfs_refcount_finish_one_leftover(mp, agno, type,
+				bno, blockcount, *adjusted);
+	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);
+}
diff --git a/libxfs/xfs_refcount.h b/libxfs/xfs_refcount.h
index 2ef2b28..605d7d7 100644
--- a/libxfs/xfs_refcount.h
+++ b/libxfs/xfs_refcount.h
@@ -41,4 +41,16 @@ 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_extlen_t *adjusted, struct xfs_btree_cur **pcur);
+
 #endif	/* __XFS_REFCOUNT_H__ */

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 16/71] xfs: adjust refcount when unmapping file blocks

[Darrick J. Wong]

When we're unmapping blocks from a reflinked file, decrease the
refcount of the affected blocks and free the extents that are no
longer in use.

v2: Use deferred ops system to avoid deadlocks and running out of
transaction reservation.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_bmap.c |   14 +++++++++++---
 1 file changed, 11 insertions(+), 3 deletions(-)


diff --git a/libxfs/xfs_bmap.c b/libxfs/xfs_bmap.c
index a50a8af..4045340 100644
--- a/libxfs/xfs_bmap.c
+++ b/libxfs/xfs_bmap.c
@@ -40,6 +40,7 @@
 #include "xfs_quota_defs.h"
 #include "xfs_rmap.h"
 #include "xfs_ag_resv.h"
+#include "xfs_refcount.h"
 
 
 kmem_zone_t		*xfs_bmap_free_item_zone;
@@ -5054,9 +5055,16 @@ xfs_bmap_del_extent(
 	/*
 	 * If we need to, add to list of extents to delete.
 	 */
-	if (do_fx)
-		xfs_bmap_add_free(mp, dfops, del->br_startblock,
-				del->br_blockcount, NULL);
+	if (do_fx) {
+		if (xfs_is_reflink_inode(ip) && whichfork == XFS_DATA_FORK) {
+			error = xfs_refcount_decrease_extent(mp, dfops, del);
+			if (error)
+				goto done;
+		} else
+			xfs_bmap_add_free(mp, dfops, del->br_startblock,
+					del->br_blockcount, NULL);
+	}
+
 	/*
 	 * Adjust inode # blocks in the file.
 	 */

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 17/71] xfs: refcount btree requires more reserved space

[Darrick J. Wong]

The reference count btree is allocated from the free space, which
means that we have to ensure that an AG can't run out of free space
while performing a refcount operation.  In the pathological case each
AG block has its own refcntbt record, so we have to keep that much
space available.

v2: Calculate the maximum possible size of the rmap and refcount
btrees based on minimally-full btree blocks.  This increases the
per-AG block reservations to handle the worst case btree size.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_alloc.c          |    3 +++
 libxfs/xfs_refcount_btree.c |   23 +++++++++++++++++++++++
 libxfs/xfs_refcount_btree.h |    4 ++++
 3 files changed, 30 insertions(+)


diff --git a/libxfs/xfs_alloc.c b/libxfs/xfs_alloc.c
index e754647..3151fbb 100644
--- a/libxfs/xfs_alloc.c
+++ b/libxfs/xfs_alloc.c
@@ -34,6 +34,7 @@
 #include "xfs_trace.h"
 #include "xfs_trans.h"
 #include "xfs_ag_resv.h"
+#include "xfs_refcount_btree.h"
 
 struct workqueue_struct *xfs_alloc_wq;
 
@@ -124,6 +125,8 @@ xfs_alloc_ag_max_usable(
 		blocks++;		/* finobt root block */
 	if (xfs_sb_version_hasrmapbt(&mp->m_sb))
 		blocks++; 		/* rmap root block */
+	if (xfs_sb_version_hasreflink(&mp->m_sb))
+		blocks++;		/* refcount root block */
 
 	return mp->m_sb.sb_agblocks - blocks;
 }
diff --git a/libxfs/xfs_refcount_btree.c b/libxfs/xfs_refcount_btree.c
index 04f8db0..568a2f8 100644
--- a/libxfs/xfs_refcount_btree.c
+++ b/libxfs/xfs_refcount_btree.c
@@ -380,3 +380,26 @@ xfs_refcountbt_compute_maxlevels(
 	mp->m_refc_maxlevels = xfs_btree_compute_maxlevels(mp,
 			mp->m_refc_mnr, mp->m_sb.sb_agblocks);
 }
+
+/* Calculate the refcount btree size for some records. */
+xfs_extlen_t
+xfs_refcountbt_calc_size(
+	struct xfs_mount	*mp,
+	unsigned long long	len)
+{
+	return xfs_btree_calc_size(mp, mp->m_refc_mnr, len);
+}
+
+/*
+ * Calculate the maximum refcount btree size.
+ */
+xfs_extlen_t
+xfs_refcountbt_max_size(
+	struct xfs_mount	*mp)
+{
+	/* Bail out if we're uninitialized, which can happen in mkfs. */
+	if (mp->m_refc_mxr[0] == 0)
+		return 0;
+
+	return xfs_refcountbt_calc_size(mp, mp->m_sb.sb_agblocks);
+}
diff --git a/libxfs/xfs_refcount_btree.h b/libxfs/xfs_refcount_btree.h
index 9e9ad7c..780b02f 100644
--- a/libxfs/xfs_refcount_btree.h
+++ b/libxfs/xfs_refcount_btree.h
@@ -64,4 +64,8 @@ extern int xfs_refcountbt_maxrecs(struct xfs_mount *mp, int blocklen,
 		bool leaf);
 extern void xfs_refcountbt_compute_maxlevels(struct xfs_mount *mp);
 
+extern xfs_extlen_t xfs_refcountbt_calc_size(struct xfs_mount *mp,
+		unsigned long long len);
+extern xfs_extlen_t xfs_refcountbt_max_size(struct xfs_mount *mp);
+
 #endif	/* __XFS_REFCOUNT_BTREE_H__ */

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 18/71] xfs: introduce reflink utility functions

[Darrick J. Wong]

These functions will be used by the other reflink functions to find
the maximum length of a range of shared blocks.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.coM>
---
 include/xfs_trace.h   |    3 +
 libxfs/xfs_refcount.c |  100 +++++++++++++++++++++++++++++++++++++++++++++++++
 libxfs/xfs_refcount.h |    4 ++
 3 files changed, 107 insertions(+)


diff --git a/include/xfs_trace.h b/include/xfs_trace.h
index 5605c3c..0698280 100644
--- a/include/xfs_trace.h
+++ b/include/xfs_trace.h
@@ -255,6 +255,9 @@
 #define trace_xfs_refcount_deferred(...)	((void) 0)
 #define trace_xfs_refcount_defer(...)		((void) 0)
 #define trace_xfs_refcount_finish_one_leftover(...)	((void) 0)
+#define trace_xfs_refcount_find_shared(...)	((void) 0)
+#define trace_xfs_refcount_find_shared_result(...)	((void) 0)
+#define trace_xfs_refcount_find_shared_error(...)	((void) 0)
 
 /* set c = c to avoid unused var warnings */
 #define trace_xfs_perag_get(a,b,c,d)	((c) = (c))
diff --git a/libxfs/xfs_refcount.c b/libxfs/xfs_refcount.c
index 31028bf..20b4f68 100644
--- a/libxfs/xfs_refcount.c
+++ b/libxfs/xfs_refcount.c
@@ -1127,3 +1127,103 @@ xfs_refcount_decrease_extent(
 	return __xfs_refcount_add(mp, dfops, XFS_REFCOUNT_DECREASE,
 			PREV->br_startblock, PREV->br_blockcount);
 }
+
+/*
+ * Given an AG extent, find the lowest-numbered run of shared blocks within
+ * that range and return the range in fbno/flen.  If find_maximal is set,
+ * return the longest extent of shared blocks; if not, just return the first
+ * extent we find.  If no shared blocks are found, flen will be set to zero.
+ */
+int
+xfs_refcount_find_shared(
+	struct xfs_btree_cur		*cur,
+	xfs_agblock_t			agbno,
+	xfs_extlen_t			aglen,
+	xfs_agblock_t			*fbno,
+	xfs_extlen_t			*flen,
+	bool				find_maximal)
+{
+	struct xfs_refcount_irec	tmp;
+	int				i;
+	int				have;
+	int				error;
+
+	trace_xfs_refcount_find_shared(cur->bc_mp, cur->bc_private.a.agno,
+			agbno, aglen);
+
+	/* By default, skip the whole range */
+	*fbno = agbno + aglen;
+	*flen = 0;
+
+	/* Try to find a refcount extent that crosses the start */
+	error = xfs_refcount_lookup_le(cur, agbno, &have);
+	if (error)
+		goto out_error;
+	if (!have) {
+		/* No left extent, look at the next one */
+		error = xfs_btree_increment(cur, 0, &have);
+		if (error)
+			goto out_error;
+		if (!have)
+			goto done;
+	}
+	error = xfs_refcount_get_rec(cur, &tmp, &i);
+	if (error)
+		goto out_error;
+	XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, i == 1, out_error);
+
+	/* If the extent ends before the start, look at the next one */
+	if (tmp.rc_startblock + tmp.rc_blockcount <= agbno) {
+		error = xfs_btree_increment(cur, 0, &have);
+		if (error)
+			goto out_error;
+		if (!have)
+			goto done;
+		error = xfs_refcount_get_rec(cur, &tmp, &i);
+		if (error)
+			goto out_error;
+		XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, i == 1, out_error);
+	}
+
+	/* If the extent ends after the range we want, bail out */
+	if (tmp.rc_startblock >= agbno + aglen)
+		goto done;
+
+	/* We found the start of a shared extent! */
+	if (tmp.rc_startblock < agbno) {
+		tmp.rc_blockcount -= (agbno - tmp.rc_startblock);
+		tmp.rc_startblock = agbno;
+	}
+
+	*fbno = tmp.rc_startblock;
+	*flen = min(tmp.rc_blockcount, agbno + aglen - *fbno);
+	if (!find_maximal)
+		goto done;
+
+	/* Otherwise, find the end of this shared extent */
+	while (*fbno + *flen < agbno + aglen) {
+		error = xfs_btree_increment(cur, 0, &have);
+		if (error)
+			goto out_error;
+		if (!have)
+			break;
+		error = xfs_refcount_get_rec(cur, &tmp, &i);
+		if (error)
+			goto out_error;
+		XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, i == 1, out_error);
+		if (tmp.rc_startblock >= agbno + aglen ||
+		    tmp.rc_startblock != *fbno + *flen)
+			break;
+		*flen = min(*flen + tmp.rc_blockcount, agbno + aglen - *fbno);
+	}
+
+done:
+	trace_xfs_refcount_find_shared_result(cur->bc_mp,
+			cur->bc_private.a.agno, *fbno, *flen);
+
+out_error:
+	if (error)
+		trace_xfs_refcount_find_shared_error(cur->bc_mp,
+				cur->bc_private.a.agno, error, _RET_IP_);
+	return error;
+}
diff --git a/libxfs/xfs_refcount.h b/libxfs/xfs_refcount.h
index 605d7d7..0ec9a28 100644
--- a/libxfs/xfs_refcount.h
+++ b/libxfs/xfs_refcount.h
@@ -53,4 +53,8 @@ extern int xfs_refcount_finish_one(struct xfs_trans *tp,
 		xfs_fsblock_t startblock, xfs_extlen_t blockcount,
 		xfs_extlen_t *adjusted, struct xfs_btree_cur **pcur);
 
+extern int xfs_refcount_find_shared(struct xfs_btree_cur *cur,
+		xfs_agblock_t agbno, xfs_extlen_t aglen, xfs_agblock_t *fbno,
+		xfs_extlen_t *flen, bool find_maximal);
+
 #endif	/* __XFS_REFCOUNT_H__ */

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 19/71] xfs: create bmbt update intent log items

[Darrick J. Wong]

Create bmbt update intent/done log items to record redo information in
the log.  Because we need to roll transactions multiple times for
reflink operations, between we also have to track the status of the
metadata updates that will be recorded in the post-roll transactions,
just in case we crash before committing the final transaction.  This
mechanism enables log recovery to finish what was already started.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_log_format.h |   51 +++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 49 insertions(+), 2 deletions(-)


diff --git a/libxfs/xfs_log_format.h b/libxfs/xfs_log_format.h
index ebf5dc0..fffcc0f 100644
--- a/libxfs/xfs_log_format.h
+++ b/libxfs/xfs_log_format.h
@@ -114,7 +114,9 @@ static inline uint xlog_get_cycle(char *ptr)
 #define XLOG_REG_TYPE_RUD_FORMAT	22
 #define XLOG_REG_TYPE_CUI_FORMAT	23
 #define XLOG_REG_TYPE_CUD_FORMAT	24
-#define XLOG_REG_TYPE_MAX		24
+#define XLOG_REG_TYPE_BUI_FORMAT	25
+#define XLOG_REG_TYPE_BUD_FORMAT	26
+#define XLOG_REG_TYPE_MAX		26
 
 /*
  * Flags to log operation header
@@ -235,6 +237,8 @@ typedef struct xfs_trans_header {
 #define	XFS_LI_RUD		0x1241
 #define	XFS_LI_CUI		0x1242	/* refcount update intent */
 #define	XFS_LI_CUD		0x1243
+#define	XFS_LI_BUI		0x1244	/* bmbt update intent */
+#define	XFS_LI_BUD		0x1245
 
 #define XFS_LI_TYPE_DESC \
 	{ XFS_LI_EFI,		"XFS_LI_EFI" }, \
@@ -248,7 +252,9 @@ typedef struct xfs_trans_header {
 	{ XFS_LI_RUI,		"XFS_LI_RUI" }, \
 	{ XFS_LI_RUD,		"XFS_LI_RUD" }, \
 	{ XFS_LI_CUI,		"XFS_LI_CUI" }, \
-	{ XFS_LI_CUD,		"XFS_LI_CUD" }
+	{ XFS_LI_CUD,		"XFS_LI_CUD" }, \
+	{ XFS_LI_BUI,		"XFS_LI_BUI" }, \
+	{ XFS_LI_BUD,		"XFS_LI_BUD" }
 
 /*
  * Inode Log Item Format definitions.
@@ -712,6 +718,47 @@ struct xfs_cud_log_format {
 };
 
 /*
+ * BUI/BUD (inode block mapping) log format definitions
+ */
+
+/* bmbt me_flags: upper bits are flags, lower byte is type code */
+#define XFS_BMAP_EXTENT_MAP		1
+#define XFS_BMAP_EXTENT_UNMAP		2
+#define XFS_BMAP_EXTENT_TYPE_MASK	0xFF
+
+#define XFS_BMAP_EXTENT_ATTR_FORK	(1U << 31)
+#define XFS_BMAP_EXTENT_UNWRITTEN	(1U << 30)
+
+#define XFS_BMAP_EXTENT_FLAGS		(XFS_BMAP_EXTENT_TYPE_MASK | \
+					 XFS_BMAP_EXTENT_ATTR_FORK | \
+					 XFS_BMAP_EXTENT_UNWRITTEN)
+
+/*
+ * This is the structure used to lay out an bui log item in the
+ * log.  The bui_extents field is a variable size array whose
+ * size is given by bui_nextents.
+ */
+struct xfs_bui_log_format {
+	__uint16_t		bui_type;	/* bui log item type */
+	__uint16_t		bui_size;	/* size of this item */
+	__uint32_t		bui_nextents;	/* # extents to free */
+	__uint64_t		bui_id;		/* bui identifier */
+	struct xfs_map_extent	bui_extents[1];	/* array of extents to bmap */
+};
+
+/*
+ * This is the structure used to lay out an bud log item in the
+ * log.  The bud_extents array is a variable size array whose
+ * size is given by bud_nextents;
+ */
+struct xfs_bud_log_format {
+	__uint16_t		bud_type;	/* bud log item type */
+	__uint16_t		bud_size;	/* size of this item */
+	__uint32_t		__pad;
+	__uint64_t		bud_bui_id;	/* id of corresponding bui */
+};
+
+/*
  * Dquot Log format definitions.
  *
  * The first two fields must be the type and size fitting into

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 20/71] xfs: log bmap intent items

[Darrick J. Wong]

Provide a mechanism for higher levels to create BUI/BUD items, submit
them to the log, and a stub function to deal with recovered BUI items.
These parts will be connected to the rmapbt in a later patch.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_bmap.h |   12 ++++++++++++
 1 file changed, 12 insertions(+)


diff --git a/libxfs/xfs_bmap.h b/libxfs/xfs_bmap.h
index 9220b1d..c8160fe 100644
--- a/libxfs/xfs_bmap.h
+++ b/libxfs/xfs_bmap.h
@@ -209,5 +209,17 @@ struct xfs_bmbt_rec_host *
 				struct xfs_bmbt_irec *gotp,
 				struct xfs_bmbt_irec *prevp);
 
+enum xfs_bmap_intent_type {
+	XFS_BMAP_MAP,
+	XFS_BMAP_UNMAP,
+};
+
+struct xfs_bmap_intent {
+	struct list_head			bi_list;
+	enum xfs_bmap_intent_type		bi_type;
+	struct xfs_inode			*bi_owner;
+	int					bi_whichfork;
+	struct xfs_bmbt_irec			bi_bmap;
+};
 
 #endif	/* __XFS_BMAP_H__ */

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 21/71] xfs: map an inode's offset to an exact physical block

[Darrick J. Wong]

Teach the bmap routine to know how to map a range of file blocks to a
specific range of physical blocks, instead of simply allocating fresh
blocks.  This enables reflink to map a file to blocks that are already
in use.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 include/xfs_trace.h |    3 ++
 libxfs/xfs_bmap.c   |   63 +++++++++++++++++++++++++++++++++++++++++++++++++++
 libxfs/xfs_bmap.h   |   10 +++++++-
 3 files changed, 75 insertions(+), 1 deletion(-)


diff --git a/include/xfs_trace.h b/include/xfs_trace.h
index 0698280..cc0e960 100644
--- a/include/xfs_trace.h
+++ b/include/xfs_trace.h
@@ -259,6 +259,9 @@
 #define trace_xfs_refcount_find_shared_result(...)	((void) 0)
 #define trace_xfs_refcount_find_shared_error(...)	((void) 0)
 
+#define trace_xfs_bmap_remap_alloc(...)		((void) 0)
+#define trace_xfs_bmap_remap_alloc_error(...)	((void) 0)
+
 /* set c = c to avoid unused var warnings */
 #define trace_xfs_perag_get(a,b,c,d)	((c) = (c))
 #define trace_xfs_perag_get_tag(a,b,c,d) ((c) = (c))
diff --git a/libxfs/xfs_bmap.c b/libxfs/xfs_bmap.c
index 4045340..fab8182 100644
--- a/libxfs/xfs_bmap.c
+++ b/libxfs/xfs_bmap.c
@@ -3865,6 +3865,55 @@ xfs_bmap_btalloc(
 }
 
 /*
+ * For a remap operation, just "allocate" an extent at the address that the
+ * caller passed in, and ensure that the AGFL is the right size.  The caller
+ * will then map the "allocated" extent into the file somewhere.
+ */
+STATIC int
+xfs_bmap_remap_alloc(
+	struct xfs_bmalloca	*ap)
+{
+	struct xfs_trans	*tp = ap->tp;
+	struct xfs_mount	*mp = tp->t_mountp;
+	xfs_agblock_t		bno;
+	struct xfs_alloc_arg	args;
+	int			error;
+
+	/*
+	 * validate that the block number is legal - the enables us to detect
+	 * and handle a silent filesystem corruption rather than crashing.
+	 */
+	memset(&args, 0, sizeof(struct xfs_alloc_arg));
+	args.tp = ap->tp;
+	args.mp = ap->tp->t_mountp;
+	bno = *ap->firstblock;
+	args.agno = XFS_FSB_TO_AGNO(mp, bno);
+	ASSERT(args.agno < mp->m_sb.sb_agcount);
+	args.agbno = XFS_FSB_TO_AGBNO(mp, bno);
+	ASSERT(args.agbno < mp->m_sb.sb_agblocks);
+
+	/* "Allocate" the extent from the range we passed in. */
+	trace_xfs_bmap_remap_alloc(ap->ip, *ap->firstblock, ap->length);
+	ap->blkno = bno;
+	ap->ip->i_d.di_nblocks += ap->length;
+	xfs_trans_log_inode(ap->tp, ap->ip, XFS_ILOG_CORE);
+
+	/* Fix the freelist, like a real allocator does. */
+	args.userdata = 1;
+	args.pag = xfs_perag_get(args.mp, args.agno);
+	ASSERT(args.pag);
+
+	error = xfs_alloc_fix_freelist(&args, XFS_ALLOC_FLAG_FREEING);
+	if (error)
+		goto error0;
+error0:
+	xfs_perag_put(args.pag);
+	if (error)
+		trace_xfs_bmap_remap_alloc_error(ap->ip, error, _RET_IP_);
+	return error;
+}
+
+/*
  * xfs_bmap_alloc is called by xfs_bmapi to allocate an extent for a file.
  * It figures out where to ask the underlying allocator to put the new extent.
  */
@@ -3872,6 +3921,8 @@ STATIC int
 xfs_bmap_alloc(
 	struct xfs_bmalloca	*ap)	/* bmap alloc argument struct */
 {
+	if (ap->flags & XFS_BMAPI_REMAP)
+		return xfs_bmap_remap_alloc(ap);
 	if (XFS_IS_REALTIME_INODE(ap->ip) && ap->userdata)
 		return xfs_bmap_rtalloc(ap);
 	return xfs_bmap_btalloc(ap);
@@ -4508,6 +4559,12 @@ xfs_bmapi_write(
 	ASSERT(len > 0);
 	ASSERT(XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_LOCAL);
 	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
+	if (whichfork == XFS_ATTR_FORK)
+		ASSERT(!(flags & XFS_BMAPI_REMAP));
+	if (flags & XFS_BMAPI_REMAP) {
+		ASSERT(!(flags & XFS_BMAPI_PREALLOC));
+		ASSERT(!(flags & XFS_BMAPI_CONVERT));
+	}
 
 	/* zeroing is for currently only for data extents, not metadata */
 	ASSERT((flags & (XFS_BMAPI_METADATA | XFS_BMAPI_ZERO)) !=
@@ -4569,6 +4626,12 @@ xfs_bmapi_write(
 		wasdelay = !inhole && isnullstartblock(bma.got.br_startblock);
 
 		/*
+		 * Make sure we only reflink into a hole.
+		 */
+		if (flags & XFS_BMAPI_REMAP)
+			ASSERT(inhole);
+
+		/*
 		 * First, deal with the hole before the allocated space
 		 * that we found, if any.
 		 */
diff --git a/libxfs/xfs_bmap.h b/libxfs/xfs_bmap.h
index c8160fe..42dc49e 100644
--- a/libxfs/xfs_bmap.h
+++ b/libxfs/xfs_bmap.h
@@ -97,6 +97,13 @@ struct xfs_extent_free_item
  */
 #define XFS_BMAPI_ZERO		0x080
 
+/*
+ * Map the inode offset to the block given in ap->firstblock.  Primarily
+ * used for reflink.  The range must be in a hole, and this flag cannot be
+ * turned on with PREALLOC or CONVERT, and cannot be used on the attr fork.
+ */
+#define XFS_BMAPI_REMAP		0x100
+
 #define XFS_BMAPI_FLAGS \
 	{ XFS_BMAPI_ENTIRE,	"ENTIRE" }, \
 	{ XFS_BMAPI_METADATA,	"METADATA" }, \
@@ -105,7 +112,8 @@ struct xfs_extent_free_item
 	{ XFS_BMAPI_IGSTATE,	"IGSTATE" }, \
 	{ XFS_BMAPI_CONTIG,	"CONTIG" }, \
 	{ XFS_BMAPI_CONVERT,	"CONVERT" }, \
-	{ XFS_BMAPI_ZERO,	"ZERO" }
+	{ XFS_BMAPI_ZERO,	"ZERO" }, \
+	{ XFS_BMAPI_REMAP,	"REMAP" }
 
 
 static inline int xfs_bmapi_aflag(int w)

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 22/71] xfs: pass bmapi flags through to bmap_del_extent

[Darrick J. Wong]

Pass BMAPI_ flags from bunmapi into bmap_del_extent and extend
BMAPI_REMAP (which means "don't touch the allocator or the quota
accounting") to apply to bunmapi as well.  This will be used to
implement the unmap operation, which will be used by swapext.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_bmap.c |    9 +++++----
 libxfs/xfs_bmap.h |    3 +++
 2 files changed, 8 insertions(+), 4 deletions(-)


diff --git a/libxfs/xfs_bmap.c b/libxfs/xfs_bmap.c
index fab8182..ff99768 100644
--- a/libxfs/xfs_bmap.c
+++ b/libxfs/xfs_bmap.c
@@ -4826,7 +4826,8 @@ xfs_bmap_del_extent(
 	xfs_btree_cur_t		*cur,	/* if null, not a btree */
 	xfs_bmbt_irec_t		*del,	/* data to remove from extents */
 	int			*logflagsp, /* inode logging flags */
-	int			whichfork) /* data or attr fork */
+	int			whichfork, /* data or attr fork */
+	int			bflags)	/* bmapi flags */
 {
 	xfs_filblks_t		da_new;	/* new delay-alloc indirect blocks */
 	xfs_filblks_t		da_old;	/* old delay-alloc indirect blocks */
@@ -5118,7 +5119,7 @@ xfs_bmap_del_extent(
 	/*
 	 * If we need to, add to list of extents to delete.
 	 */
-	if (do_fx) {
+	if (do_fx && !(bflags & XFS_BMAPI_REMAP)) {
 		if (xfs_is_reflink_inode(ip) && whichfork == XFS_DATA_FORK) {
 			error = xfs_refcount_decrease_extent(mp, dfops, del);
 			if (error)
@@ -5136,7 +5137,7 @@ xfs_bmap_del_extent(
 	/*
 	 * Adjust quota data.
 	 */
-	if (qfield)
+	if (qfield && !(bflags & XFS_BMAPI_REMAP))
 		xfs_trans_mod_dquot_byino(tp, ip, qfield, (long)-nblks);
 
 	/*
@@ -5461,7 +5462,7 @@ xfs_bunmapi(
 			cur->bc_private.b.flags &= ~XFS_BTCUR_BPRV_WASDEL;
 
 		error = xfs_bmap_del_extent(ip, tp, &lastx, dfops, cur, &del,
-				&tmp_logflags, whichfork);
+				&tmp_logflags, whichfork, flags);
 		logflags |= tmp_logflags;
 		if (error)
 			goto error0;
diff --git a/libxfs/xfs_bmap.h b/libxfs/xfs_bmap.h
index 42dc49e..4d9b666 100644
--- a/libxfs/xfs_bmap.h
+++ b/libxfs/xfs_bmap.h
@@ -101,6 +101,9 @@ struct xfs_extent_free_item
  * Map the inode offset to the block given in ap->firstblock.  Primarily
  * used for reflink.  The range must be in a hole, and this flag cannot be
  * turned on with PREALLOC or CONVERT, and cannot be used on the attr fork.
+ *
+ * For bunmapi, this flag unmaps the range without adjusting quota, reducing
+ * refcount, or freeing the blocks.
  */
 #define XFS_BMAPI_REMAP		0x100
 

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 23/71] xfs: implement deferred bmbt map/unmap operations

[Darrick J. Wong]

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 <darrick.wong@oracle.com>
---
 include/xfs_trace.h |    2 +
 include/xfs_trans.h |    1 
 libxfs/defer_item.c |  107 +++++++++++++++++++++++++++++++++++++++
 libxfs/init.c       |    1 
 libxfs/xfs_bmap.c   |  141 +++++++++++++++++++++++++++++++++++++++++++++++++++
 libxfs/xfs_bmap.h   |   11 ++++
 libxfs/xfs_defer.h  |    1 
 7 files changed, 264 insertions(+)


diff --git a/include/xfs_trace.h b/include/xfs_trace.h
index cc0e960..506f23b 100644
--- a/include/xfs_trace.h
+++ b/include/xfs_trace.h
@@ -261,6 +261,8 @@
 
 #define trace_xfs_bmap_remap_alloc(...)		((void) 0)
 #define trace_xfs_bmap_remap_alloc_error(...)	((void) 0)
+#define trace_xfs_bmap_deferred(...)		((void) 0)
+#define trace_xfs_bmap_defer(...)		((void) 0)
 
 /* set c = c to avoid unused var warnings */
 #define trace_xfs_perag_get(a,b,c,d)	((c) = (c))
diff --git a/include/xfs_trans.h b/include/xfs_trans.h
index 739a792..44deebb 100644
--- a/include/xfs_trans.h
+++ b/include/xfs_trans.h
@@ -148,5 +148,6 @@ libxfs_trans_read_buf(
 void xfs_extent_free_init_defer_op(void);
 void xfs_rmap_update_init_defer_op(void);
 void xfs_refcount_update_init_defer_op(void);
+void xfs_bmap_update_init_defer_op(void);
 
 #endif	/* __XFS_TRANS_H__ */
diff --git a/libxfs/defer_item.c b/libxfs/defer_item.c
index 133e6e1..d4cd876 100644
--- a/libxfs/defer_item.c
+++ b/libxfs/defer_item.c
@@ -32,6 +32,8 @@
 #include "xfs_alloc.h"
 #include "xfs_rmap.h"
 #include "xfs_refcount.h"
+#include "xfs_bmap.h"
+#include "xfs_inode.h"
 
 /* Dummy defer item ops, since we don't do logging. */
 
@@ -386,3 +388,108 @@ xfs_refcount_update_init_defer_op(void)
 {
 	xfs_defer_init_op_type(&xfs_refcount_update_defer_type);
 }
+
+/* Inode Block Mapping */
+
+/* Sort bmap intents by inode. */
+static int
+xfs_bmap_update_diff_items(
+	void				*priv,
+	struct list_head		*a,
+	struct list_head		*b)
+{
+	struct xfs_bmap_intent		*ba;
+	struct xfs_bmap_intent		*bb;
+
+	ba = container_of(a, struct xfs_bmap_intent, bi_list);
+	bb = container_of(b, struct xfs_bmap_intent, bi_list);
+	return ba->bi_owner->i_ino - bb->bi_owner->i_ino;
+}
+
+/* Get an BUI. */
+STATIC void *
+xfs_bmap_update_create_intent(
+	struct xfs_trans		*tp,
+	unsigned int			count)
+{
+	return NULL;
+}
+
+/* Log bmap updates in the intent item. */
+STATIC void
+xfs_bmap_update_log_item(
+	struct xfs_trans		*tp,
+	void				*intent,
+	struct list_head		*item)
+{
+}
+
+/* Get an BUD so we can process all the deferred rmap updates. */
+STATIC void *
+xfs_bmap_update_create_done(
+	struct xfs_trans		*tp,
+	void				*intent,
+	unsigned int			count)
+{
+	return NULL;
+}
+
+/* Process a deferred rmap update. */
+STATIC int
+xfs_bmap_update_finish_item(
+	struct xfs_trans		*tp,
+	struct xfs_defer_ops		*dop,
+	struct list_head		*item,
+	void				*done_item,
+	void				**state)
+{
+	struct xfs_bmap_intent		*bmap;
+	int				error;
+
+	bmap = container_of(item, struct xfs_bmap_intent, bi_list);
+	error = xfs_bmap_finish_one(tp, dop,
+			bmap->bi_owner,
+			bmap->bi_type, bmap->bi_whichfork,
+			bmap->bi_bmap.br_startoff,
+			bmap->bi_bmap.br_startblock,
+			bmap->bi_bmap.br_blockcount,
+			bmap->bi_bmap.br_state);
+	kmem_free(bmap);
+	return error;
+}
+
+/* Abort all pending BUIs. */
+STATIC void
+xfs_bmap_update_abort_intent(
+	void				*intent)
+{
+}
+
+/* Cancel a deferred rmap update. */
+STATIC void
+xfs_bmap_update_cancel_item(
+	struct list_head		*item)
+{
+	struct xfs_bmap_intent		*bmap;
+
+	bmap = container_of(item, struct xfs_bmap_intent, bi_list);
+	kmem_free(bmap);
+}
+
+static const struct xfs_defer_op_type xfs_bmap_update_defer_type = {
+	.type		= XFS_DEFER_OPS_TYPE_BMAP,
+	.diff_items	= xfs_bmap_update_diff_items,
+	.create_intent	= xfs_bmap_update_create_intent,
+	.abort_intent	= xfs_bmap_update_abort_intent,
+	.log_item	= xfs_bmap_update_log_item,
+	.create_done	= xfs_bmap_update_create_done,
+	.finish_item	= xfs_bmap_update_finish_item,
+	.cancel_item	= xfs_bmap_update_cancel_item,
+};
+
+/* Register the deferred op type. */
+void
+xfs_bmap_update_init_defer_op(void)
+{
+	xfs_defer_init_op_type(&xfs_bmap_update_defer_type);
+}
diff --git a/libxfs/init.c b/libxfs/init.c
index 2d1bb58..9134879 100644
--- a/libxfs/init.c
+++ b/libxfs/init.c
@@ -267,6 +267,7 @@ libxfs_init(libxfs_init_t *a)
 	xfs_extent_free_init_defer_op();
 	xfs_rmap_update_init_defer_op();
 	xfs_refcount_update_init_defer_op();
+	xfs_bmap_update_init_defer_op();
 
 	radix_tree_init();
 
diff --git a/libxfs/xfs_bmap.c b/libxfs/xfs_bmap.c
index ff99768..a919371 100644
--- a/libxfs/xfs_bmap.c
+++ b/libxfs/xfs_bmap.c
@@ -6123,3 +6123,144 @@ out:
 	xfs_trans_cancel(tp);
 	return error;
 }
+
+/* Deferred mapping is only for real extents in the data fork. */
+static bool
+xfs_bmap_is_update_needed(
+	int			whichfork,
+	struct xfs_bmbt_irec	*bmap)
+{
+	ASSERT(whichfork == XFS_DATA_FORK);
+
+	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,
+	int			whichfork,
+	struct xfs_bmbt_irec	*PREV)
+{
+	if (!xfs_bmap_is_update_needed(whichfork, PREV))
+		return 0;
+
+	return __xfs_bmap_add(mp, dfops, XFS_BMAP_MAP, ip, whichfork, 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,
+	int			whichfork,
+	struct xfs_bmbt_irec	*PREV)
+{
+	if (!xfs_bmap_is_update_needed(whichfork, PREV))
+		return 0;
+
+	return __xfs_bmap_add(mp, dfops, XFS_BMAP_UNMAP, ip, whichfork, 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				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 (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,
+					XFS_BMAPI_REMAP, &firstfsb,
+					bmap.br_blockcount, &bmap, &nimaps,
+					dfops);
+		break;
+	case XFS_BMAP_UNMAP:
+		error = xfs_bunmapi(tp, ip, bmap.br_startoff,
+				bmap.br_blockcount, XFS_BMAPI_REMAP, 1, &firstfsb,
+				dfops, &done);
+		ASSERT(done);
+		break;
+	default:
+		ASSERT(0);
+		error = -EFSCORRUPTED;
+	}
+
+	return error;
+}
diff --git a/libxfs/xfs_bmap.h b/libxfs/xfs_bmap.h
index 4d9b666..858788b 100644
--- a/libxfs/xfs_bmap.h
+++ b/libxfs/xfs_bmap.h
@@ -233,4 +233,15 @@ 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, int whichfork,
+		struct xfs_bmbt_irec *imap);
+int	xfs_bmap_unmap_extent(struct xfs_mount *mp, struct xfs_defer_ops *dfops,
+		struct xfs_inode *ip, int whichfork,
+		struct xfs_bmbt_irec *imap);
+
 #endif	/* __XFS_BMAP_H__ */
diff --git a/libxfs/xfs_defer.h b/libxfs/xfs_defer.h
index d47a482..aa57eaa 100644
--- a/libxfs/xfs_defer.h
+++ b/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,

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 24/71] xfs: return work remaining at the end of a bunmapi operation

[Darrick J. Wong]

Return the range of file blocks that bunmapi didn't free.  This hint
is used by CoW and reflink to figure out what part of an extent
actually got freed so that it can set up the appropriate atomic
remapping of just the freed range.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_bmap.c |   36 ++++++++++++++++++++++++++++++------
 libxfs/xfs_bmap.h |    4 ++++
 2 files changed, 34 insertions(+), 6 deletions(-)


diff --git a/libxfs/xfs_bmap.c b/libxfs/xfs_bmap.c
index a919371..011ac9b 100644
--- a/libxfs/xfs_bmap.c
+++ b/libxfs/xfs_bmap.c
@@ -5159,17 +5159,16 @@ done:
  * *done is set.
  */
 int						/* error */
-xfs_bunmapi(
+__xfs_bunmapi(
 	xfs_trans_t		*tp,		/* transaction pointer */
 	struct xfs_inode	*ip,		/* incore inode */
 	xfs_fileoff_t		bno,		/* starting offset to unmap */
-	xfs_filblks_t		len,		/* length to unmap in file */
+	xfs_filblks_t		*rlen,		/* i/o: amount remaining */
 	int			flags,		/* misc flags */
 	xfs_extnum_t		nexts,		/* number of extents max */
 	xfs_fsblock_t		*firstblock,	/* first allocated block
 						   controls a.g. for allocs */
-	struct xfs_defer_ops	*dfops,		/* i/o: list extents to free */
-	int			*done)		/* set if not done yet */
+	struct xfs_defer_ops	*dfops)		/* i/o: deferred updates */
 {
 	xfs_btree_cur_t		*cur;		/* bmap btree cursor */
 	xfs_bmbt_irec_t		del;		/* extent being deleted */
@@ -5191,6 +5190,7 @@ xfs_bunmapi(
 	int			wasdel;		/* was a delayed alloc extent */
 	int			whichfork;	/* data or attribute fork */
 	xfs_fsblock_t		sum;
+	xfs_filblks_t		len = *rlen;	/* length to unmap in file */
 
 	trace_xfs_bunmap(ip, bno, len, flags, _RET_IP_);
 
@@ -5217,7 +5217,7 @@ xfs_bunmapi(
 		return error;
 	nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t);
 	if (nextents == 0) {
-		*done = 1;
+		*rlen = 0;
 		return 0;
 	}
 	XFS_STATS_INC(mp, xs_blk_unmap);
@@ -5488,7 +5488,10 @@ nodelete:
 			extno++;
 		}
 	}
-	*done = bno == (xfs_fileoff_t)-1 || bno < start || lastx < 0;
+	if (bno == (xfs_fileoff_t)-1 || bno < start || lastx < 0)
+		*rlen = 0;
+	else
+		*rlen = bno - start + 1;
 
 	/*
 	 * Convert to a btree if necessary.
@@ -5544,6 +5547,27 @@ error0:
 	return error;
 }
 
+/* Unmap a range of a file. */
+int
+xfs_bunmapi(
+	xfs_trans_t		*tp,
+	struct xfs_inode	*ip,
+	xfs_fileoff_t		bno,
+	xfs_filblks_t		len,
+	int			flags,
+	xfs_extnum_t		nexts,
+	xfs_fsblock_t		*firstblock,
+	struct xfs_defer_ops	*dfops,
+	int			*done)
+{
+	int			error;
+
+	error = __xfs_bunmapi(tp, ip, bno, &len, flags, nexts, firstblock,
+			dfops);
+	*done = (len == 0);
+	return error;
+}
+
 /*
  * Determine whether an extent shift can be accomplished by a merge with the
  * extent that precedes the target hole of the shift.
diff --git a/libxfs/xfs_bmap.h b/libxfs/xfs_bmap.h
index 858788b..4b68e96 100644
--- a/libxfs/xfs_bmap.h
+++ b/libxfs/xfs_bmap.h
@@ -200,6 +200,10 @@ int	xfs_bmapi_write(struct xfs_trans *tp, struct xfs_inode *ip,
 		xfs_fsblock_t *firstblock, xfs_extlen_t total,
 		struct xfs_bmbt_irec *mval, int *nmap,
 		struct xfs_defer_ops *dfops);
+int	__xfs_bunmapi(struct xfs_trans *tp, struct xfs_inode *ip,
+		xfs_fileoff_t bno, xfs_filblks_t *rlen, int flags,
+		xfs_extnum_t nexts, xfs_fsblock_t *firstblock,
+		struct xfs_defer_ops *dfops);
 int	xfs_bunmapi(struct xfs_trans *tp, struct xfs_inode *ip,
 		xfs_fileoff_t bno, xfs_filblks_t len, int flags,
 		xfs_extnum_t nexts, xfs_fsblock_t *firstblock,

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 25/71] xfs: add reflink feature flag to geometry

[Darrick J. Wong]

Report the reflink feature in the XFS geometry so that xfs_info and
friends know the filesystem has this feature.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_fs.h |    3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)


diff --git a/libxfs/xfs_fs.h b/libxfs/xfs_fs.h
index 085ea6f..f291a53 100644
--- a/libxfs/xfs_fs.h
+++ b/libxfs/xfs_fs.h
@@ -230,7 +230,8 @@ typedef struct xfs_fsop_resblks {
 #define XFS_FSOP_GEOM_FLAGS_FTYPE	0x10000	/* inode directory types */
 #define XFS_FSOP_GEOM_FLAGS_FINOBT	0x20000	/* free inode btree */
 #define XFS_FSOP_GEOM_FLAGS_SPINODES	0x40000	/* sparse inode chunks	*/
-#define XFS_FSOP_GEOM_FLAGS_RMAPBT	0x80000	/* Reverse mapping btree */
+#define XFS_FSOP_GEOM_FLAGS_RMAPBT	0x80000	/* reverse mapping btree */
+#define XFS_FSOP_GEOM_FLAGS_REFLINK	0x100000 /* files can share blocks */
 
 /*
  * Minimum and maximum sizes need for growth checks.

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 26/71] xfs: don't allow reflinked dir/dev/fifo/socket/pipe files

[Darrick J. Wong]

Only non-rt files can be reflinked, so check that when we load an
inode.  Also, don't leak the attr fork if there's a failure.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_inode_fork.c |   23 ++++++++++++++++++++++-
 1 file changed, 22 insertions(+), 1 deletion(-)


diff --git a/libxfs/xfs_inode_fork.c b/libxfs/xfs_inode_fork.c
index 75d448b..eda590f 100644
--- a/libxfs/xfs_inode_fork.c
+++ b/libxfs/xfs_inode_fork.c
@@ -117,6 +117,26 @@ xfs_iformat_fork(
 		return -EFSCORRUPTED;
 	}
 
+	if (unlikely(xfs_is_reflink_inode(ip) &&
+	    (VFS_I(ip)->i_mode & S_IFMT) != S_IFREG)) {
+		xfs_warn(ip->i_mount,
+			"corrupt dinode %llu, wrong file type for reflink.",
+			ip->i_ino);
+		XFS_CORRUPTION_ERROR("xfs_iformat(reflink)",
+				     XFS_ERRLEVEL_LOW, ip->i_mount, dip);
+		return -EFSCORRUPTED;
+	}
+
+	if (unlikely(xfs_is_reflink_inode(ip) &&
+	    (ip->i_d.di_flags & XFS_DIFLAG_REALTIME))) {
+		xfs_warn(ip->i_mount,
+			"corrupt dinode %llu, has reflink+realtime flag set.",
+			ip->i_ino);
+		XFS_CORRUPTION_ERROR("xfs_iformat(reflink)",
+				     XFS_ERRLEVEL_LOW, ip->i_mount, dip);
+		return -EFSCORRUPTED;
+	}
+
 	switch (VFS_I(ip)->i_mode & S_IFMT) {
 	case S_IFIFO:
 	case S_IFCHR:
@@ -204,7 +224,8 @@ xfs_iformat_fork(
 			XFS_CORRUPTION_ERROR("xfs_iformat(8)",
 					     XFS_ERRLEVEL_LOW,
 					     ip->i_mount, dip);
-			return -EFSCORRUPTED;
+			error = -EFSCORRUPTED;
+			break;
 		}
 
 		error = xfs_iformat_local(ip, dip, XFS_ATTR_FORK, size);

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 27/71] xfs: introduce the CoW fork

[Darrick J. Wong]

Introduce a new in-core fork for storing copy-on-write delalloc
reservations and allocated extents that are in the process of being
written out.

v2: fix up bmapi_read so that we can query the CoW fork, and have it
return a "hole" extent if there's no CoW fork.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 include/xfs_inode.h     |    3 +++
 libxfs/rdwr.c           |    2 ++
 libxfs/xfs_bmap.c       |   27 ++++++++++++++++++++-------
 libxfs/xfs_bmap.h       |   22 +++++++++++++++++++---
 libxfs/xfs_bmap_btree.c |    1 +
 libxfs/xfs_inode_fork.c |   47 ++++++++++++++++++++++++++++++++++++++++++++---
 libxfs/xfs_inode_fork.h |   28 ++++++++++++++++++++++------
 libxfs/xfs_rmap.c       |   15 ++++++++-------
 libxfs/xfs_types.h      |    1 +
 9 files changed, 120 insertions(+), 26 deletions(-)


diff --git a/include/xfs_inode.h b/include/xfs_inode.h
index 3876fa6..b7623c0 100644
--- a/include/xfs_inode.h
+++ b/include/xfs_inode.h
@@ -50,6 +50,7 @@ typedef struct xfs_inode {
 	struct xfs_imap		i_imap;		/* location for xfs_imap() */
 	struct xfs_buftarg	i_dev;		/* dev for this inode */
 	struct xfs_ifork	*i_afp;		/* attribute fork pointer */
+	struct xfs_ifork	*i_cowfp;	/* copy on write extents */
 	struct xfs_ifork	i_df;		/* data fork */
 	struct xfs_trans	*i_transp;	/* ptr to owning transaction */
 	struct xfs_inode_log_item *i_itemp;	/* logging information */
@@ -58,6 +59,8 @@ typedef struct xfs_inode {
 	xfs_fsize_t		i_size;		/* in-memory size */
 	const struct xfs_dir_ops *d_ops;	/* directory ops vector */
 	struct inode		i_vnode;
+	xfs_extnum_t		i_cnextents;	/* # of extents in cow fork */
+	unsigned int		i_cformat;	/* format of cow fork */
 } xfs_inode_t;
 
 static inline struct inode *VFS_I(struct xfs_inode *ip)
diff --git a/libxfs/rdwr.c b/libxfs/rdwr.c
index aa30522..533a064 100644
--- a/libxfs/rdwr.c
+++ b/libxfs/rdwr.c
@@ -1372,6 +1372,8 @@ libxfs_idestroy(xfs_inode_t *ip)
 	}
 	if (ip->i_afp)
 		libxfs_idestroy_fork(ip, XFS_ATTR_FORK);
+	if (ip->i_cowfp)
+		xfs_idestroy_fork(ip, XFS_COW_FORK);
 }
 
 void
diff --git a/libxfs/xfs_bmap.c b/libxfs/xfs_bmap.c
index 011ac9b..3121f7f 100644
--- a/libxfs/xfs_bmap.c
+++ b/libxfs/xfs_bmap.c
@@ -2916,6 +2916,7 @@ xfs_bmap_add_extent_hole_real(
 	ASSERT(!isnullstartblock(new->br_startblock));
 	ASSERT(!bma->cur ||
 	       !(bma->cur->bc_private.b.flags & XFS_BTCUR_BPRV_WASDEL));
+	ASSERT(whichfork != XFS_COW_FORK);
 
 	XFS_STATS_INC(mp, xs_add_exlist);
 
@@ -4051,12 +4052,11 @@ xfs_bmapi_read(
 	int			error;
 	int			eof;
 	int			n = 0;
-	int			whichfork = (flags & XFS_BMAPI_ATTRFORK) ?
-						XFS_ATTR_FORK : XFS_DATA_FORK;
+	int			whichfork = xfs_bmapi_whichfork(flags);
 
 	ASSERT(*nmap >= 1);
 	ASSERT(!(flags & ~(XFS_BMAPI_ATTRFORK|XFS_BMAPI_ENTIRE|
-			   XFS_BMAPI_IGSTATE)));
+			   XFS_BMAPI_IGSTATE|XFS_BMAPI_COWFORK)));
 	ASSERT(xfs_isilocked(ip, XFS_ILOCK_SHARED|XFS_ILOCK_EXCL));
 
 	if (unlikely(XFS_TEST_ERROR(
@@ -4074,6 +4074,16 @@ xfs_bmapi_read(
 
 	ifp = XFS_IFORK_PTR(ip, whichfork);
 
+	/* No CoW fork?  Return a hole. */
+	if (whichfork == XFS_COW_FORK && !ifp) {
+		mval->br_startoff = bno;
+		mval->br_startblock = HOLESTARTBLOCK;
+		mval->br_blockcount = len;
+		mval->br_state = XFS_EXT_NORM;
+		*nmap = 1;
+		return 0;
+	}
+
 	if (!(ifp->if_flags & XFS_IFEXTENTS)) {
 		error = xfs_iread_extents(NULL, ip, whichfork);
 		if (error)
@@ -4426,8 +4436,7 @@ xfs_bmapi_convert_unwritten(
 	xfs_filblks_t		len,
 	int			flags)
 {
-	int			whichfork = (flags & XFS_BMAPI_ATTRFORK) ?
-						XFS_ATTR_FORK : XFS_DATA_FORK;
+	int			whichfork = xfs_bmapi_whichfork(flags);
 	struct xfs_ifork	*ifp = XFS_IFORK_PTR(bma->ip, whichfork);
 	int			tmp_logflags = 0;
 	int			error;
@@ -4443,6 +4452,8 @@ xfs_bmapi_convert_unwritten(
 			(XFS_BMAPI_PREALLOC | XFS_BMAPI_CONVERT))
 		return 0;
 
+	ASSERT(whichfork != XFS_COW_FORK);
+
 	/*
 	 * Modify (by adding) the state flag, if writing.
 	 */
@@ -4856,6 +4867,8 @@ xfs_bmap_del_extent(
 
 	if (whichfork == XFS_ATTR_FORK)
 		state |= BMAP_ATTRFORK;
+	else if (whichfork == XFS_COW_FORK)
+		state |= BMAP_COWFORK;
 
 	ifp = XFS_IFORK_PTR(ip, whichfork);
 	ASSERT((*idx >= 0) && (*idx < ifp->if_bytes /
@@ -5194,8 +5207,8 @@ __xfs_bunmapi(
 
 	trace_xfs_bunmap(ip, bno, len, flags, _RET_IP_);
 
-	whichfork = (flags & XFS_BMAPI_ATTRFORK) ?
-		XFS_ATTR_FORK : XFS_DATA_FORK;
+	whichfork = xfs_bmapi_whichfork(flags);
+	ASSERT(whichfork != XFS_COW_FORK);
 	ifp = XFS_IFORK_PTR(ip, whichfork);
 	if (unlikely(
 	    XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_EXTENTS &&
diff --git a/libxfs/xfs_bmap.h b/libxfs/xfs_bmap.h
index 4b68e96..b7df7d4 100644
--- a/libxfs/xfs_bmap.h
+++ b/libxfs/xfs_bmap.h
@@ -107,6 +107,9 @@ struct xfs_extent_free_item
  */
 #define XFS_BMAPI_REMAP		0x100
 
+/* Map something in the CoW fork. */
+#define XFS_BMAPI_COWFORK	0x200
+
 #define XFS_BMAPI_FLAGS \
 	{ XFS_BMAPI_ENTIRE,	"ENTIRE" }, \
 	{ XFS_BMAPI_METADATA,	"METADATA" }, \
@@ -116,12 +119,23 @@ struct xfs_extent_free_item
 	{ XFS_BMAPI_CONTIG,	"CONTIG" }, \
 	{ XFS_BMAPI_CONVERT,	"CONVERT" }, \
 	{ XFS_BMAPI_ZERO,	"ZERO" }, \
-	{ XFS_BMAPI_REMAP,	"REMAP" }
+	{ XFS_BMAPI_REMAP,	"REMAP" }, \
+	{ XFS_BMAPI_COWFORK,	"COWFORK" }
 
 
 static inline int xfs_bmapi_aflag(int w)
 {
-	return (w == XFS_ATTR_FORK ? XFS_BMAPI_ATTRFORK : 0);
+	return (w == XFS_ATTR_FORK ? XFS_BMAPI_ATTRFORK :
+	       (w == XFS_COW_FORK ? XFS_BMAPI_COWFORK : 0));
+}
+
+static inline int xfs_bmapi_whichfork(int bmapi_flags)
+{
+	if (bmapi_flags & XFS_BMAPI_COWFORK)
+		return XFS_COW_FORK;
+	else if (bmapi_flags & XFS_BMAPI_ATTRFORK)
+		return XFS_ATTR_FORK;
+	return XFS_DATA_FORK;
 }
 
 /*
@@ -142,13 +156,15 @@ static inline int xfs_bmapi_aflag(int w)
 #define BMAP_LEFT_VALID		(1 << 6)
 #define BMAP_RIGHT_VALID	(1 << 7)
 #define BMAP_ATTRFORK		(1 << 8)
+#define BMAP_COWFORK		(1 << 9)
 
 #define XFS_BMAP_EXT_FLAGS \
 	{ BMAP_LEFT_CONTIG,	"LC" }, \
 	{ BMAP_RIGHT_CONTIG,	"RC" }, \
 	{ BMAP_LEFT_FILLING,	"LF" }, \
 	{ BMAP_RIGHT_FILLING,	"RF" }, \
-	{ BMAP_ATTRFORK,	"ATTR" }
+	{ BMAP_ATTRFORK,	"ATTR" }, \
+	{ BMAP_COWFORK,		"COW" }
 
 
 /*
diff --git a/libxfs/xfs_bmap_btree.c b/libxfs/xfs_bmap_btree.c
index 906983e..3d1a02e 100644
--- a/libxfs/xfs_bmap_btree.c
+++ b/libxfs/xfs_bmap_btree.c
@@ -774,6 +774,7 @@ xfs_bmbt_init_cursor(
 {
 	struct xfs_ifork	*ifp = XFS_IFORK_PTR(ip, whichfork);
 	struct xfs_btree_cur	*cur;
+	ASSERT(whichfork != XFS_COW_FORK);
 
 	cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_SLEEP);
 
diff --git a/libxfs/xfs_inode_fork.c b/libxfs/xfs_inode_fork.c
index eda590f..e65f633 100644
--- a/libxfs/xfs_inode_fork.c
+++ b/libxfs/xfs_inode_fork.c
@@ -202,9 +202,14 @@ xfs_iformat_fork(
 		XFS_ERROR_REPORT("xfs_iformat(7)", XFS_ERRLEVEL_LOW, ip->i_mount);
 		return -EFSCORRUPTED;
 	}
-	if (error) {
+	if (error)
 		return error;
+
+	if (xfs_is_reflink_inode(ip)) {
+		ASSERT(ip->i_cowfp == NULL);
+		xfs_ifork_init_cow(ip);
 	}
+
 	if (!XFS_DFORK_Q(dip))
 		return 0;
 
@@ -243,6 +248,9 @@ xfs_iformat_fork(
 	if (error) {
 		kmem_zone_free(xfs_ifork_zone, ip->i_afp);
 		ip->i_afp = NULL;
+		if (ip->i_cowfp)
+			kmem_zone_free(xfs_ifork_zone, ip->i_cowfp);
+		ip->i_cowfp = NULL;
 		xfs_idestroy_fork(ip, XFS_DATA_FORK);
 	}
 	return error;
@@ -757,6 +765,9 @@ xfs_idestroy_fork(
 	if (whichfork == XFS_ATTR_FORK) {
 		kmem_zone_free(xfs_ifork_zone, ip->i_afp);
 		ip->i_afp = NULL;
+	} else if (whichfork == XFS_COW_FORK) {
+		kmem_zone_free(xfs_ifork_zone, ip->i_cowfp);
+		ip->i_cowfp = NULL;
 	}
 }
 
@@ -944,6 +955,19 @@ xfs_iext_get_ext(
 	}
 }
 
+/* XFS_IEXT_STATE_TO_FORK() -- Convert BMAP state flags to an inode fork. */
+xfs_ifork_t *
+XFS_IEXT_STATE_TO_FORK(
+	struct xfs_inode	*ip,
+	int			state)
+{
+	if (state & BMAP_COWFORK)
+		return ip->i_cowfp;
+	else if (state & BMAP_ATTRFORK)
+		return ip->i_afp;
+	return &ip->i_df;
+}
+
 /*
  * Insert new item(s) into the extent records for incore inode
  * fork 'ifp'.  'count' new items are inserted at index 'idx'.
@@ -956,7 +980,7 @@ xfs_iext_insert(
 	xfs_bmbt_irec_t	*new,		/* items to insert */
 	int		state)		/* type of extent conversion */
 {
-	xfs_ifork_t	*ifp = (state & BMAP_ATTRFORK) ? ip->i_afp : &ip->i_df;
+	xfs_ifork_t	*ifp = XFS_IEXT_STATE_TO_FORK(ip, state);
 	xfs_extnum_t	i;		/* extent record index */
 
 	trace_xfs_iext_insert(ip, idx, new, state, _RET_IP_);
@@ -1206,7 +1230,7 @@ xfs_iext_remove(
 	int		ext_diff,	/* number of extents to remove */
 	int		state)		/* type of extent conversion */
 {
-	xfs_ifork_t	*ifp = (state & BMAP_ATTRFORK) ? ip->i_afp : &ip->i_df;
+	xfs_ifork_t	*ifp = XFS_IEXT_STATE_TO_FORK(ip, state);
 	xfs_extnum_t	nextents;	/* number of extents in file */
 	int		new_size;	/* size of extents after removal */
 
@@ -1953,3 +1977,20 @@ xfs_iext_irec_update_extoffs(
 		ifp->if_u1.if_ext_irec[i].er_extoff += ext_diff;
 	}
 }
+
+/*
+ * Initialize an inode's copy-on-write fork.
+ */
+void
+xfs_ifork_init_cow(
+	struct xfs_inode	*ip)
+{
+	if (ip->i_cowfp)
+		return;
+
+	ip->i_cowfp = kmem_zone_zalloc(xfs_ifork_zone,
+				       KM_SLEEP | KM_NOFS);
+	ip->i_cowfp->if_flags = XFS_IFEXTENTS;
+	ip->i_cformat = XFS_DINODE_FMT_EXTENTS;
+	ip->i_cnextents = 0;
+}
diff --git a/libxfs/xfs_inode_fork.h b/libxfs/xfs_inode_fork.h
index f95e072..44d38eb 100644
--- a/libxfs/xfs_inode_fork.h
+++ b/libxfs/xfs_inode_fork.h
@@ -92,7 +92,9 @@ typedef struct xfs_ifork {
 #define XFS_IFORK_PTR(ip,w)		\
 	((w) == XFS_DATA_FORK ? \
 		&(ip)->i_df : \
-		(ip)->i_afp)
+		((w) == XFS_ATTR_FORK ? \
+			(ip)->i_afp : \
+			(ip)->i_cowfp))
 #define XFS_IFORK_DSIZE(ip) \
 	(XFS_IFORK_Q(ip) ? \
 		XFS_IFORK_BOFF(ip) : \
@@ -105,26 +107,38 @@ typedef struct xfs_ifork {
 #define XFS_IFORK_SIZE(ip,w) \
 	((w) == XFS_DATA_FORK ? \
 		XFS_IFORK_DSIZE(ip) : \
-		XFS_IFORK_ASIZE(ip))
+		((w) == XFS_ATTR_FORK ? \
+			XFS_IFORK_ASIZE(ip) : \
+			0))
 #define XFS_IFORK_FORMAT(ip,w) \
 	((w) == XFS_DATA_FORK ? \
 		(ip)->i_d.di_format : \
-		(ip)->i_d.di_aformat)
+		((w) == XFS_ATTR_FORK ? \
+			(ip)->i_d.di_aformat : \
+			(ip)->i_cformat))
 #define XFS_IFORK_FMT_SET(ip,w,n) \
 	((w) == XFS_DATA_FORK ? \
 		((ip)->i_d.di_format = (n)) : \
-		((ip)->i_d.di_aformat = (n)))
+		((w) == XFS_ATTR_FORK ? \
+			((ip)->i_d.di_aformat = (n)) : \
+			((ip)->i_cformat = (n))))
 #define XFS_IFORK_NEXTENTS(ip,w) \
 	((w) == XFS_DATA_FORK ? \
 		(ip)->i_d.di_nextents : \
-		(ip)->i_d.di_anextents)
+		((w) == XFS_ATTR_FORK ? \
+			(ip)->i_d.di_anextents : \
+			(ip)->i_cnextents))
 #define XFS_IFORK_NEXT_SET(ip,w,n) \
 	((w) == XFS_DATA_FORK ? \
 		((ip)->i_d.di_nextents = (n)) : \
-		((ip)->i_d.di_anextents = (n)))
+		((w) == XFS_ATTR_FORK ? \
+			((ip)->i_d.di_anextents = (n)) : \
+			((ip)->i_cnextents = (n))))
 #define XFS_IFORK_MAXEXT(ip, w) \
 	(XFS_IFORK_SIZE(ip, w) / sizeof(xfs_bmbt_rec_t))
 
+xfs_ifork_t	*XFS_IEXT_STATE_TO_FORK(struct xfs_inode *ip, int state);
+
 int		xfs_iformat_fork(struct xfs_inode *, struct xfs_dinode *);
 void		xfs_iflush_fork(struct xfs_inode *, struct xfs_dinode *,
 				struct xfs_inode_log_item *, int);
@@ -169,4 +183,6 @@ void		xfs_iext_irec_update_extoffs(struct xfs_ifork *, int, int);
 
 extern struct kmem_zone	*xfs_ifork_zone;
 
+extern void xfs_ifork_init_cow(struct xfs_inode *ip);
+
 #endif	/* __XFS_INODE_FORK_H__ */
diff --git a/libxfs/xfs_rmap.c b/libxfs/xfs_rmap.c
index b752fed..82c2597 100644
--- a/libxfs/xfs_rmap.c
+++ b/libxfs/xfs_rmap.c
@@ -1261,9 +1261,10 @@ out_cur:
  */
 static bool
 xfs_rmap_update_is_needed(
-	struct xfs_mount	*mp)
+	struct xfs_mount	*mp,
+	int			whichfork)
 {
-	return xfs_sb_version_hasrmapbt(&mp->m_sb);
+	return xfs_sb_version_hasrmapbt(&mp->m_sb) && whichfork != XFS_COW_FORK;
 }
 
 /*
@@ -1309,7 +1310,7 @@ xfs_rmap_map_extent(
 	int			whichfork,
 	struct xfs_bmbt_irec	*PREV)
 {
-	if (!xfs_rmap_update_is_needed(mp))
+	if (!xfs_rmap_update_is_needed(mp, whichfork))
 		return 0;
 
 	return __xfs_rmap_add(mp, dfops, XFS_RMAP_MAP, ip->i_ino,
@@ -1325,7 +1326,7 @@ xfs_rmap_unmap_extent(
 	int			whichfork,
 	struct xfs_bmbt_irec	*PREV)
 {
-	if (!xfs_rmap_update_is_needed(mp))
+	if (!xfs_rmap_update_is_needed(mp, whichfork))
 		return 0;
 
 	return __xfs_rmap_add(mp, dfops, XFS_RMAP_UNMAP, ip->i_ino,
@@ -1341,7 +1342,7 @@ xfs_rmap_convert_extent(
 	int			whichfork,
 	struct xfs_bmbt_irec	*PREV)
 {
-	if (!xfs_rmap_update_is_needed(mp))
+	if (!xfs_rmap_update_is_needed(mp, whichfork))
 		return 0;
 
 	return __xfs_rmap_add(mp, dfops, XFS_RMAP_CONVERT, ip->i_ino,
@@ -1360,7 +1361,7 @@ xfs_rmap_alloc_extent(
 {
 	struct xfs_bmbt_irec	bmap;
 
-	if (!xfs_rmap_update_is_needed(mp))
+	if (!xfs_rmap_update_is_needed(mp, XFS_DATA_FORK))
 		return 0;
 
 	bmap.br_startblock = XFS_AGB_TO_FSB(mp, agno, bno);
@@ -1384,7 +1385,7 @@ xfs_rmap_free_extent(
 {
 	struct xfs_bmbt_irec	bmap;
 
-	if (!xfs_rmap_update_is_needed(mp))
+	if (!xfs_rmap_update_is_needed(mp, XFS_DATA_FORK))
 		return 0;
 
 	bmap.br_startblock = XFS_AGB_TO_FSB(mp, agno, bno);
diff --git a/libxfs/xfs_types.h b/libxfs/xfs_types.h
index 690d616..cf044c0 100644
--- a/libxfs/xfs_types.h
+++ b/libxfs/xfs_types.h
@@ -93,6 +93,7 @@ typedef __int64_t	xfs_sfiloff_t;	/* signed block number in a file */
  */
 #define	XFS_DATA_FORK	0
 #define	XFS_ATTR_FORK	1
+#define	XFS_COW_FORK	2
 
 /*
  * Min numbers of data/attr fork btree root pointers.

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 28/71] xfs: support bmapping delalloc extents in the CoW fork

[Darrick J. Wong]

Allow the creation of delayed allocation extents in the CoW fork.
In a subsequent patch we'll wire up write_begin and page_mkwrite to
actually do this.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_bmap.c |   29 ++++++++++++++++++-----------
 libxfs/xfs_bmap.h |    2 +-
 2 files changed, 19 insertions(+), 12 deletions(-)


diff --git a/libxfs/xfs_bmap.c b/libxfs/xfs_bmap.c
index 3121f7f..cdc1ed3 100644
--- a/libxfs/xfs_bmap.c
+++ b/libxfs/xfs_bmap.c
@@ -2752,6 +2752,7 @@ done:
 STATIC void
 xfs_bmap_add_extent_hole_delay(
 	xfs_inode_t		*ip,	/* incore inode pointer */
+	int			whichfork,
 	xfs_extnum_t		*idx,	/* extent number to update/insert */
 	xfs_bmbt_irec_t		*new)	/* new data to add to file extents */
 {
@@ -2763,8 +2764,10 @@ xfs_bmap_add_extent_hole_delay(
 	int			state;  /* state bits, accessed thru macros */
 	xfs_filblks_t		temp=0;	/* temp for indirect calculations */
 
-	ifp = XFS_IFORK_PTR(ip, XFS_DATA_FORK);
+	ifp = XFS_IFORK_PTR(ip, whichfork);
 	state = 0;
+	if (whichfork == XFS_COW_FORK)
+		state |= BMAP_COWFORK;
 	ASSERT(isnullstartblock(new->br_startblock));
 
 	/*
@@ -2782,7 +2785,7 @@ xfs_bmap_add_extent_hole_delay(
 	 * Check and set flags if the current (right) segment exists.
 	 * If it doesn't exist, we're converting the hole at end-of-file.
 	 */
-	if (*idx < ip->i_df.if_bytes / (uint)sizeof(xfs_bmbt_rec_t)) {
+	if (*idx < ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t)) {
 		state |= BMAP_RIGHT_VALID;
 		xfs_bmbt_get_all(xfs_iext_get_ext(ifp, *idx), &right);
 
@@ -4133,6 +4136,7 @@ xfs_bmapi_read(
 STATIC int
 xfs_bmapi_reserve_delalloc(
 	struct xfs_inode	*ip,
+	int			whichfork,
 	xfs_fileoff_t		aoff,
 	xfs_filblks_t		len,
 	struct xfs_bmbt_irec	*got,
@@ -4141,7 +4145,7 @@ xfs_bmapi_reserve_delalloc(
 	int			eof)
 {
 	struct xfs_mount	*mp = ip->i_mount;
-	struct xfs_ifork	*ifp = XFS_IFORK_PTR(ip, XFS_DATA_FORK);
+	struct xfs_ifork	*ifp = XFS_IFORK_PTR(ip, whichfork);
 	xfs_extlen_t		alen;
 	xfs_extlen_t		indlen;
 	char			rt = XFS_IS_REALTIME_INODE(ip);
@@ -4200,7 +4204,7 @@ xfs_bmapi_reserve_delalloc(
 	got->br_startblock = nullstartblock(indlen);
 	got->br_blockcount = alen;
 	got->br_state = XFS_EXT_NORM;
-	xfs_bmap_add_extent_hole_delay(ip, lastx, got);
+	xfs_bmap_add_extent_hole_delay(ip, whichfork, lastx, got);
 
 	/*
 	 * Update our extent pointer, given that xfs_bmap_add_extent_hole_delay
@@ -4232,6 +4236,7 @@ out_unreserve_quota:
 int
 xfs_bmapi_delay(
 	struct xfs_inode	*ip,	/* incore inode */
+	int			whichfork, /* data or cow fork? */
 	xfs_fileoff_t		bno,	/* starting file offs. mapped */
 	xfs_filblks_t		len,	/* length to map in file */
 	struct xfs_bmbt_irec	*mval,	/* output: map values */
@@ -4239,7 +4244,7 @@ xfs_bmapi_delay(
 	int			flags)	/* XFS_BMAPI_... */
 {
 	struct xfs_mount	*mp = ip->i_mount;
-	struct xfs_ifork	*ifp = XFS_IFORK_PTR(ip, XFS_DATA_FORK);
+	struct xfs_ifork	*ifp = XFS_IFORK_PTR(ip, whichfork);
 	struct xfs_bmbt_irec	got;	/* current file extent record */
 	struct xfs_bmbt_irec	prev;	/* previous file extent record */
 	xfs_fileoff_t		obno;	/* old block number (offset) */
@@ -4249,14 +4254,15 @@ xfs_bmapi_delay(
 	int			n = 0;	/* current extent index */
 	int			error = 0;
 
+	ASSERT(whichfork == XFS_DATA_FORK || whichfork == XFS_COW_FORK);
 	ASSERT(*nmap >= 1);
 	ASSERT(*nmap <= XFS_BMAP_MAX_NMAP);
 	ASSERT(!(flags & ~XFS_BMAPI_ENTIRE));
 	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
 
 	if (unlikely(XFS_TEST_ERROR(
-	    (XFS_IFORK_FORMAT(ip, XFS_DATA_FORK) != XFS_DINODE_FMT_EXTENTS &&
-	     XFS_IFORK_FORMAT(ip, XFS_DATA_FORK) != XFS_DINODE_FMT_BTREE),
+	    (XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_EXTENTS &&
+	     XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_BTREE),
 	     mp, XFS_ERRTAG_BMAPIFORMAT, XFS_RANDOM_BMAPIFORMAT))) {
 		XFS_ERROR_REPORT("xfs_bmapi_delay", XFS_ERRLEVEL_LOW, mp);
 		return -EFSCORRUPTED;
@@ -4267,19 +4273,20 @@ xfs_bmapi_delay(
 
 	XFS_STATS_INC(mp, xs_blk_mapw);
 
-	if (!(ifp->if_flags & XFS_IFEXTENTS)) {
-		error = xfs_iread_extents(NULL, ip, XFS_DATA_FORK);
+	if (whichfork == XFS_DATA_FORK && !(ifp->if_flags & XFS_IFEXTENTS)) {
+		error = xfs_iread_extents(NULL, ip, whichfork);
 		if (error)
 			return error;
 	}
 
-	xfs_bmap_search_extents(ip, bno, XFS_DATA_FORK, &eof, &lastx, &got, &prev);
+	xfs_bmap_search_extents(ip, bno, whichfork, &eof, &lastx, &got, &prev);
 	end = bno + len;
 	obno = bno;
 
 	while (bno < end && n < *nmap) {
 		if (eof || got.br_startoff > bno) {
-			error = xfs_bmapi_reserve_delalloc(ip, bno, len, &got,
+			error = xfs_bmapi_reserve_delalloc(ip, whichfork,
+							   bno, len, &got,
 							   &prev, &lastx, eof);
 			if (error) {
 				if (n == 0) {
diff --git a/libxfs/xfs_bmap.h b/libxfs/xfs_bmap.h
index b7df7d4..c9f29b3 100644
--- a/libxfs/xfs_bmap.h
+++ b/libxfs/xfs_bmap.h
@@ -208,7 +208,7 @@ int	xfs_bmap_read_extents(struct xfs_trans *tp, struct xfs_inode *ip,
 int	xfs_bmapi_read(struct xfs_inode *ip, xfs_fileoff_t bno,
 		xfs_filblks_t len, struct xfs_bmbt_irec *mval,
 		int *nmap, int flags);
-int	xfs_bmapi_delay(struct xfs_inode *ip, xfs_fileoff_t bno,
+int	xfs_bmapi_delay(struct xfs_inode *ip, int whichfork, xfs_fileoff_t bno,
 		xfs_filblks_t len, struct xfs_bmbt_irec *mval,
 		int *nmap, int flags);
 int	xfs_bmapi_write(struct xfs_trans *tp, struct xfs_inode *ip,

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 29/71] xfs: support allocating delayed extents in CoW fork

[Darrick J. Wong]

Modify xfs_bmap_add_extent_delay_real() so that we can convert delayed
allocation extents in the CoW fork to real allocations, and wire this
up all the way back to xfs_iomap_write_allocate().  In a subsequent
patch, we'll modify the writepage handler to call this.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_bmap.c |   51 +++++++++++++++++++++++++++++++++++----------------
 1 file changed, 35 insertions(+), 16 deletions(-)


diff --git a/libxfs/xfs_bmap.c b/libxfs/xfs_bmap.c
index cdc1ed3..6503241 100644
--- a/libxfs/xfs_bmap.c
+++ b/libxfs/xfs_bmap.c
@@ -133,7 +133,8 @@ xfs_bmbt_lookup_ge(
  */
 static inline bool xfs_bmap_needs_btree(struct xfs_inode *ip, int whichfork)
 {
-	return XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_EXTENTS &&
+	return whichfork != XFS_COW_FORK &&
+		XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_EXTENTS &&
 		XFS_IFORK_NEXTENTS(ip, whichfork) >
 			XFS_IFORK_MAXEXT(ip, whichfork);
 }
@@ -143,7 +144,8 @@ static inline bool xfs_bmap_needs_btree(struct xfs_inode *ip, int whichfork)
  */
 static inline bool xfs_bmap_wants_extents(struct xfs_inode *ip, int whichfork)
 {
-	return XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_BTREE &&
+	return whichfork != XFS_COW_FORK &&
+		XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_BTREE &&
 		XFS_IFORK_NEXTENTS(ip, whichfork) <=
 			XFS_IFORK_MAXEXT(ip, whichfork);
 }
@@ -633,6 +635,7 @@ xfs_bmap_btree_to_extents(
 
 	mp = ip->i_mount;
 	ifp = XFS_IFORK_PTR(ip, whichfork);
+	ASSERT(whichfork != XFS_COW_FORK);
 	ASSERT(ifp->if_flags & XFS_IFEXTENTS);
 	ASSERT(XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_BTREE);
 	rblock = ifp->if_broot;
@@ -699,6 +702,7 @@ xfs_bmap_extents_to_btree(
 	xfs_bmbt_ptr_t		*pp;		/* root block address pointer */
 
 	mp = ip->i_mount;
+	ASSERT(whichfork != XFS_COW_FORK);
 	ifp = XFS_IFORK_PTR(ip, whichfork);
 	ASSERT(XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_EXTENTS);
 
@@ -830,6 +834,7 @@ xfs_bmap_local_to_extents_empty(
 {
 	struct xfs_ifork	*ifp = XFS_IFORK_PTR(ip, whichfork);
 
+	ASSERT(whichfork != XFS_COW_FORK);
 	ASSERT(XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_LOCAL);
 	ASSERT(ifp->if_bytes == 0);
 	ASSERT(XFS_IFORK_NEXTENTS(ip, whichfork) == 0);
@@ -1663,7 +1668,8 @@ xfs_bmap_one_block(
  */
 STATIC int				/* error */
 xfs_bmap_add_extent_delay_real(
-	struct xfs_bmalloca	*bma)
+	struct xfs_bmalloca	*bma,
+	int			whichfork)
 {
 	struct xfs_bmbt_irec	*new = &bma->got;
 	int			diff;	/* temp value */
@@ -1682,10 +1688,13 @@ xfs_bmap_add_extent_delay_real(
 	xfs_filblks_t		temp2=0;/* value for da_new calculations */
 	int			tmp_rval;	/* partial logging flags */
 	struct xfs_mount	*mp;
-	int			whichfork = XFS_DATA_FORK;
+	xfs_extnum_t		*nextents;
 
 	mp = bma->ip->i_mount;
 	ifp = XFS_IFORK_PTR(bma->ip, whichfork);
+	ASSERT(whichfork != XFS_ATTR_FORK);
+	nextents = (whichfork == XFS_COW_FORK ? &bma->ip->i_cnextents :
+						&bma->ip->i_d.di_nextents);
 
 	ASSERT(bma->idx >= 0);
 	ASSERT(bma->idx <= ifp->if_bytes / sizeof(struct xfs_bmbt_rec));
@@ -1699,6 +1708,9 @@ xfs_bmap_add_extent_delay_real(
 #define	RIGHT		r[1]
 #define	PREV		r[2]
 
+	if (whichfork == XFS_COW_FORK)
+		state |= BMAP_COWFORK;
+
 	/*
 	 * Set up a bunch of variables to make the tests simpler.
 	 */
@@ -1785,7 +1797,7 @@ xfs_bmap_add_extent_delay_real(
 		trace_xfs_bmap_post_update(bma->ip, bma->idx, state, _THIS_IP_);
 
 		xfs_iext_remove(bma->ip, bma->idx + 1, 2, state);
-		bma->ip->i_d.di_nextents--;
+		(*nextents)--;
 		if (bma->cur == NULL)
 			rval = XFS_ILOG_CORE | XFS_ILOG_DEXT;
 		else {
@@ -1887,7 +1899,7 @@ xfs_bmap_add_extent_delay_real(
 		xfs_bmbt_set_startblock(ep, new->br_startblock);
 		trace_xfs_bmap_post_update(bma->ip, bma->idx, state, _THIS_IP_);
 
-		bma->ip->i_d.di_nextents++;
+		(*nextents)++;
 		if (bma->cur == NULL)
 			rval = XFS_ILOG_CORE | XFS_ILOG_DEXT;
 		else {
@@ -1957,7 +1969,7 @@ xfs_bmap_add_extent_delay_real(
 		temp = PREV.br_blockcount - new->br_blockcount;
 		xfs_bmbt_set_blockcount(ep, temp);
 		xfs_iext_insert(bma->ip, bma->idx, 1, new, state);
-		bma->ip->i_d.di_nextents++;
+		(*nextents)++;
 		if (bma->cur == NULL)
 			rval = XFS_ILOG_CORE | XFS_ILOG_DEXT;
 		else {
@@ -2041,7 +2053,7 @@ xfs_bmap_add_extent_delay_real(
 		trace_xfs_bmap_pre_update(bma->ip, bma->idx, state, _THIS_IP_);
 		xfs_bmbt_set_blockcount(ep, temp);
 		xfs_iext_insert(bma->ip, bma->idx + 1, 1, new, state);
-		bma->ip->i_d.di_nextents++;
+		(*nextents)++;
 		if (bma->cur == NULL)
 			rval = XFS_ILOG_CORE | XFS_ILOG_DEXT;
 		else {
@@ -2110,7 +2122,7 @@ xfs_bmap_add_extent_delay_real(
 		RIGHT.br_blockcount = temp2;
 		/* insert LEFT (r[0]) and RIGHT (r[1]) at the same time */
 		xfs_iext_insert(bma->ip, bma->idx + 1, 2, &LEFT, state);
-		bma->ip->i_d.di_nextents++;
+		(*nextents)++;
 		if (bma->cur == NULL)
 			rval = XFS_ILOG_CORE | XFS_ILOG_DEXT;
 		else {
@@ -2208,7 +2220,8 @@ xfs_bmap_add_extent_delay_real(
 
 	xfs_bmap_check_leaf_extents(bma->cur, bma->ip, whichfork);
 done:
-	bma->logflags |= rval;
+	if (whichfork != XFS_COW_FORK)
+		bma->logflags |= rval;
 	return error;
 #undef	LEFT
 #undef	RIGHT
@@ -3849,7 +3862,8 @@ xfs_bmap_btalloc(
 		ASSERT(nullfb || fb_agno == args.agno ||
 		       (ap->dfops->dop_low && fb_agno < args.agno));
 		ap->length = args.len;
-		ap->ip->i_d.di_nblocks += args.len;
+		if (!(ap->flags & XFS_BMAPI_COWFORK))
+			ap->ip->i_d.di_nblocks += args.len;
 		xfs_trans_log_inode(ap->tp, ap->ip, XFS_ILOG_CORE);
 		if (ap->wasdel)
 			ap->ip->i_delayed_blks -= args.len;
@@ -4323,8 +4337,7 @@ xfs_bmapi_allocate(
 	struct xfs_bmalloca	*bma)
 {
 	struct xfs_mount	*mp = bma->ip->i_mount;
-	int			whichfork = (bma->flags & XFS_BMAPI_ATTRFORK) ?
-						XFS_ATTR_FORK : XFS_DATA_FORK;
+	int			whichfork = xfs_bmapi_whichfork(bma->flags);
 	struct xfs_ifork	*ifp = XFS_IFORK_PTR(bma->ip, whichfork);
 	int			tmp_logflags = 0;
 	int			error;
@@ -4413,7 +4426,7 @@ xfs_bmapi_allocate(
 		bma->got.br_state = XFS_EXT_UNWRITTEN;
 
 	if (bma->wasdel)
-		error = xfs_bmap_add_extent_delay_real(bma);
+		error = xfs_bmap_add_extent_delay_real(bma, whichfork);
 	else
 		error = xfs_bmap_add_extent_hole_real(bma, whichfork);
 
@@ -4567,8 +4580,7 @@ xfs_bmapi_write(
 	orig_mval = mval;
 	orig_nmap = *nmap;
 #endif
-	whichfork = (flags & XFS_BMAPI_ATTRFORK) ?
-		XFS_ATTR_FORK : XFS_DATA_FORK;
+	whichfork = xfs_bmapi_whichfork(flags);
 
 	ASSERT(*nmap >= 1);
 	ASSERT(*nmap <= XFS_BMAP_MAX_NMAP);
@@ -4579,6 +4591,11 @@ xfs_bmapi_write(
 	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
 	if (whichfork == XFS_ATTR_FORK)
 		ASSERT(!(flags & XFS_BMAPI_REMAP));
+	if (whichfork == XFS_COW_FORK) {
+		ASSERT(!(flags & XFS_BMAPI_REMAP));
+		ASSERT(!(flags & XFS_BMAPI_PREALLOC));
+		ASSERT(!(flags & XFS_BMAPI_CONVERT));
+	}
 	if (flags & XFS_BMAPI_REMAP) {
 		ASSERT(!(flags & XFS_BMAPI_PREALLOC));
 		ASSERT(!(flags & XFS_BMAPI_CONVERT));
@@ -4648,6 +4665,8 @@ xfs_bmapi_write(
 		 */
 		if (flags & XFS_BMAPI_REMAP)
 			ASSERT(inhole);
+		if (flags & XFS_BMAPI_COWFORK)
+			ASSERT(!inhole);
 
 		/*
 		 * First, deal with the hole before the allocated space

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 30/71] xfs: support removing extents from CoW fork

[Darrick J. Wong]

Create a helper method to remove extents from the CoW fork without
any of the side effects (rmapbt/bmbt updates) of the regular extent
deletion routine.  We'll eventually use this to clear out the CoW fork
during ioend processing.

v2: Use bmapi_read to iterate and trim the CoW extents instead of
reading them raw via the iext code.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_bmap.c |  176 +++++++++++++++++++++++++++++++++++++++++++++++++++++
 libxfs/xfs_bmap.h |    1 
 2 files changed, 177 insertions(+)


diff --git a/libxfs/xfs_bmap.c b/libxfs/xfs_bmap.c
index 6503241..cb30099 100644
--- a/libxfs/xfs_bmap.c
+++ b/libxfs/xfs_bmap.c
@@ -4975,6 +4975,7 @@ xfs_bmap_del_extent(
 		/*
 		 * Matches the whole extent.  Delete the entry.
 		 */
+		trace_xfs_bmap_pre_update(ip, *idx, state, _THIS_IP_);
 		xfs_iext_remove(ip, *idx, 1,
 				whichfork == XFS_ATTR_FORK ? BMAP_ATTRFORK : 0);
 		--*idx;
@@ -5192,6 +5193,181 @@ done:
 }
 
 /*
+ * xfs_bunmapi_cow() -- Remove the relevant parts of the CoW fork.
+ *			See xfs_bmap_del_extent.
+ * @ip: XFS inode.
+ * @idx: Extent number to delete.
+ * @del: Extent to remove.
+ */
+int
+xfs_bunmapi_cow(
+	xfs_inode_t		*ip,
+	xfs_bmbt_irec_t		*del)
+{
+	xfs_filblks_t		da_new;	/* new delay-alloc indirect blocks */
+	xfs_filblks_t		da_old;	/* old delay-alloc indirect blocks */
+	xfs_fsblock_t		del_endblock = 0;/* first block past del */
+	xfs_fileoff_t		del_endoff;	/* first offset past del */
+	int			delay;	/* current block is delayed allocated */
+	xfs_bmbt_rec_host_t	*ep;	/* current extent entry pointer */
+	int			error;	/* error return value */
+	xfs_bmbt_irec_t		got;	/* current extent entry */
+	xfs_fileoff_t		got_endoff;	/* first offset past got */
+	xfs_ifork_t		*ifp;	/* inode fork pointer */
+	xfs_mount_t		*mp;	/* mount structure */
+	xfs_filblks_t		nblks;	/* quota/sb block count */
+	xfs_bmbt_irec_t		new;	/* new record to be inserted */
+	/* REFERENCED */
+	uint			qfield;	/* quota field to update */
+	xfs_filblks_t		temp;	/* for indirect length calculations */
+	xfs_filblks_t		temp2;	/* for indirect length calculations */
+	int			state = BMAP_COWFORK;
+	int			eof;
+	xfs_extnum_t		eidx;
+
+	mp = ip->i_mount;
+	XFS_STATS_INC(mp, xs_del_exlist);
+
+	ep = xfs_bmap_search_extents(ip, del->br_startoff, XFS_COW_FORK, &eof,
+			&eidx, &got, &new);
+
+	ifp = XFS_IFORK_PTR(ip, XFS_COW_FORK); ifp = ifp;
+	ASSERT((eidx >= 0) && (eidx < ifp->if_bytes /
+		(uint)sizeof(xfs_bmbt_rec_t)));
+	ASSERT(del->br_blockcount > 0);
+	ASSERT(got.br_startoff <= del->br_startoff);
+	del_endoff = del->br_startoff + del->br_blockcount;
+	got_endoff = got.br_startoff + got.br_blockcount;
+	ASSERT(got_endoff >= del_endoff);
+	delay = isnullstartblock(got.br_startblock);
+	ASSERT(isnullstartblock(del->br_startblock) == delay);
+	qfield = 0;
+	error = 0;
+	/*
+	 * If deleting a real allocation, must free up the disk space.
+	 */
+	if (!delay) {
+		nblks = del->br_blockcount;
+		qfield = XFS_TRANS_DQ_BCOUNT;
+		/*
+		 * Set up del_endblock and cur for later.
+		 */
+		del_endblock = del->br_startblock + del->br_blockcount;
+		da_old = da_new = 0;
+	} else {
+		da_old = startblockval(got.br_startblock);
+		da_new = 0;
+		nblks = 0;
+	}
+	qfield = qfield;
+	nblks = nblks;
+
+	/*
+	 * Set flag value to use in switch statement.
+	 * Left-contig is 2, right-contig is 1.
+	 */
+	switch (((got.br_startoff == del->br_startoff) << 1) |
+		(got_endoff == del_endoff)) {
+	case 3:
+		/*
+		 * Matches the whole extent.  Delete the entry.
+		 */
+		xfs_iext_remove(ip, eidx, 1, BMAP_COWFORK);
+		--eidx;
+		break;
+
+	case 2:
+		/*
+		 * Deleting the first part of the extent.
+		 */
+		trace_xfs_bmap_pre_update(ip, eidx, state, _THIS_IP_);
+		xfs_bmbt_set_startoff(ep, del_endoff);
+		temp = got.br_blockcount - del->br_blockcount;
+		xfs_bmbt_set_blockcount(ep, temp);
+		if (delay) {
+			temp = XFS_FILBLKS_MIN(xfs_bmap_worst_indlen(ip, temp),
+				da_old);
+			xfs_bmbt_set_startblock(ep, nullstartblock((int)temp));
+			trace_xfs_bmap_post_update(ip, eidx, state, _THIS_IP_);
+			da_new = temp;
+			break;
+		}
+		xfs_bmbt_set_startblock(ep, del_endblock);
+		trace_xfs_bmap_post_update(ip, eidx, state, _THIS_IP_);
+		break;
+
+	case 1:
+		/*
+		 * Deleting the last part of the extent.
+		 */
+		temp = got.br_blockcount - del->br_blockcount;
+		trace_xfs_bmap_pre_update(ip, eidx, state, _THIS_IP_);
+		xfs_bmbt_set_blockcount(ep, temp);
+		if (delay) {
+			temp = XFS_FILBLKS_MIN(xfs_bmap_worst_indlen(ip, temp),
+				da_old);
+			xfs_bmbt_set_startblock(ep, nullstartblock((int)temp));
+			trace_xfs_bmap_post_update(ip, eidx, state, _THIS_IP_);
+			da_new = temp;
+			break;
+		}
+		trace_xfs_bmap_post_update(ip, eidx, state, _THIS_IP_);
+		break;
+
+	case 0:
+		/*
+		 * Deleting the middle of the extent.
+		 */
+		temp = del->br_startoff - got.br_startoff;
+		trace_xfs_bmap_pre_update(ip, eidx, state, _THIS_IP_);
+		xfs_bmbt_set_blockcount(ep, temp);
+		new.br_startoff = del_endoff;
+		temp2 = got_endoff - del_endoff;
+		new.br_blockcount = temp2;
+		new.br_state = got.br_state;
+		if (!delay) {
+			new.br_startblock = del_endblock;
+		} else {
+			temp = xfs_bmap_worst_indlen(ip, temp);
+			xfs_bmbt_set_startblock(ep, nullstartblock((int)temp));
+			temp2 = xfs_bmap_worst_indlen(ip, temp2);
+			new.br_startblock = nullstartblock((int)temp2);
+			da_new = temp + temp2;
+			while (da_new > da_old) {
+				if (temp) {
+					temp--;
+					da_new--;
+					xfs_bmbt_set_startblock(ep,
+						nullstartblock((int)temp));
+				}
+				if (da_new == da_old)
+					break;
+				if (temp2) {
+					temp2--;
+					da_new--;
+					new.br_startblock =
+						nullstartblock((int)temp2);
+				}
+			}
+		}
+		trace_xfs_bmap_post_update(ip, eidx, state, _THIS_IP_);
+		xfs_iext_insert(ip, eidx + 1, 1, &new, state);
+		++eidx;
+		break;
+	}
+
+	/*
+	 * Account for change in delayed indirect blocks.
+	 * Nothing to do for disk quota accounting here.
+	 */
+	ASSERT(da_old >= da_new);
+	if (da_old > da_new)
+		xfs_mod_fdblocks(mp, (int64_t)(da_old - da_new), false);
+
+	return error;
+}
+
+/*
  * Unmap (remove) blocks from a file.
  * If nexts is nonzero then the number of extents to remove is limited to
  * that value.  If not all extents in the block range can be removed then
diff --git a/libxfs/xfs_bmap.h b/libxfs/xfs_bmap.h
index c9f29b3..e60be02 100644
--- a/libxfs/xfs_bmap.h
+++ b/libxfs/xfs_bmap.h
@@ -224,6 +224,7 @@ int	xfs_bunmapi(struct xfs_trans *tp, struct xfs_inode *ip,
 		xfs_fileoff_t bno, xfs_filblks_t len, int flags,
 		xfs_extnum_t nexts, xfs_fsblock_t *firstblock,
 		struct xfs_defer_ops *dfops, int *done);
+int	xfs_bunmapi_cow(struct xfs_inode *ip, struct xfs_bmbt_irec *del);
 int	xfs_check_nostate_extents(struct xfs_ifork *ifp, xfs_extnum_t idx,
 		xfs_extnum_t num);
 uint	xfs_default_attroffset(struct xfs_inode *ip);

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 31/71] xfs: store in-progress CoW allocations in the refcount btree

[Darrick J. Wong]

Due to the way the CoW algorithm in XFS works, there's an interval
during which blocks allocated to handle a CoW can be lost -- if the FS
goes down after the blocks are allocated but before the block
remapping takes place.  This is exacerbated by the cowextsz hint --
allocated reservations can sit around for a while, waiting to get
used.

Since the refcount btree doesn't normally store records with refcount
of 1, we can use it to record these in-progress extents.  In-progress
blocks cannot be shared because they're not user-visible, so there
shouldn't be any conflicts with other programs.  This is a better
solution than holding EFIs during writeback because (a) EFIs can't be
relogged currently, (b) even if they could, EFIs are bound by
available log space, which puts an unnecessary upper bound on how much
CoW we can have in flight, and (c) we already have a mechanism to
track blocks.

At mount time, read the refcount records and free anything we find
with a refcount of 1 because those were in-progress when the FS went
down.

v2: Use the deferred operations system to avoid deadlocks and blowing
out the transaction reservation.  This allows us to unmap a CoW
extent from the refcountbt and into a file atomically.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 include/xfs_trace.h   |    4 +
 libxfs/xfs_bmap.c     |   11 ++
 libxfs/xfs_format.h   |    3 
 libxfs/xfs_refcount.c |  317 ++++++++++++++++++++++++++++++++++++++++++++++++-
 libxfs/xfs_refcount.h |    7 +
 5 files changed, 336 insertions(+), 6 deletions(-)


diff --git a/include/xfs_trace.h b/include/xfs_trace.h
index 506f23b..b7bf3d0 100644
--- a/include/xfs_trace.h
+++ b/include/xfs_trace.h
@@ -264,6 +264,10 @@
 #define trace_xfs_bmap_deferred(...)		((void) 0)
 #define trace_xfs_bmap_defer(...)		((void) 0)
 
+#define trace_xfs_refcount_adjust_cow_error(...)	((void) 0)
+#define trace_xfs_refcount_cow_increase(...)	((void) 0)
+#define trace_xfs_refcount_cow_decrease(...)	((void) 0)
+
 /* set c = c to avoid unused var warnings */
 #define trace_xfs_perag_get(a,b,c,d)	((c) = (c))
 #define trace_xfs_perag_get_tag(a,b,c,d) ((c) = (c))
diff --git a/libxfs/xfs_bmap.c b/libxfs/xfs_bmap.c
index cb30099..a94f855 100644
--- a/libxfs/xfs_bmap.c
+++ b/libxfs/xfs_bmap.c
@@ -4698,6 +4698,17 @@ xfs_bmapi_write(
 				goto error0;
 			if (bma.blkno == NULLFSBLOCK)
 				break;
+
+			/*
+			 * If this is a CoW allocation, record the data in
+			 * the refcount btree for orphan recovery.
+			 */
+			if (whichfork == XFS_COW_FORK) {
+				error = xfs_refcount_alloc_cow_extent(mp, dfops,
+						bma.blkno, bma.length);
+				if (error)
+					goto error0;
+			}
 		}
 
 		/* Deal with the allocated space we found.  */
diff --git a/libxfs/xfs_format.h b/libxfs/xfs_format.h
index bbb9334..3991eaa 100644
--- a/libxfs/xfs_format.h
+++ b/libxfs/xfs_format.h
@@ -1374,7 +1374,8 @@ struct xfs_owner_info {
 #define XFS_RMAP_OWN_INOBT	(-6ULL)	/* Inode btree blocks */
 #define XFS_RMAP_OWN_INODES	(-7ULL)	/* Inode chunk */
 #define XFS_RMAP_OWN_REFC	(-8ULL) /* refcount tree */
-#define XFS_RMAP_OWN_MIN	(-9ULL) /* guard */
+#define XFS_RMAP_OWN_COW	(-9ULL) /* cow allocations */
+#define XFS_RMAP_OWN_MIN	(-10ULL) /* guard */
 
 #define XFS_RMAP_NON_INODE_OWNER(owner)	(!!((owner) & (1ULL << 63)))
 
diff --git a/libxfs/xfs_refcount.c b/libxfs/xfs_refcount.c
index 20b4f68..7d3735f 100644
--- a/libxfs/xfs_refcount.c
+++ b/libxfs/xfs_refcount.c
@@ -35,13 +35,23 @@
 #include "xfs_trans.h"
 #include "xfs_bit.h"
 #include "xfs_refcount.h"
+#include "xfs_rmap.h"
 
 /* Allowable refcount adjustment amounts. */
 enum xfs_refc_adjust_op {
 	XFS_REFCOUNT_ADJUST_INCREASE	= 1,
 	XFS_REFCOUNT_ADJUST_DECREASE	= -1,
+	XFS_REFCOUNT_ADJUST_COW_ALLOC	= 0,
+	XFS_REFCOUNT_ADJUST_COW_FREE	= -1,
 };
 
+STATIC int __xfs_refcount_cow_alloc(struct xfs_btree_cur *rcur,
+		xfs_agblock_t agbno, xfs_extlen_t aglen,
+		struct xfs_defer_ops *dfops);
+STATIC int __xfs_refcount_cow_free(struct xfs_btree_cur *rcur,
+		xfs_agblock_t agbno, xfs_extlen_t aglen,
+		struct xfs_defer_ops *dfops);
+
 /*
  * Look up the first record less than or equal to [bno, len] in the btree
  * given by cur.
@@ -467,6 +477,8 @@ out_error:
 	return error;
 }
 
+#define XFS_FIND_RCEXT_SHARED	1
+#define XFS_FIND_RCEXT_COW	2
 /*
  * Find the left extent and the one after it (cleft).  This function assumes
  * that we've already split any extent crossing agbno.
@@ -477,7 +489,8 @@ xfs_refcount_find_left_extents(
 	struct xfs_refcount_irec	*left,
 	struct xfs_refcount_irec	*cleft,
 	xfs_agblock_t			agbno,
-	xfs_extlen_t			aglen)
+	xfs_extlen_t			aglen,
+	int				flags)
 {
 	struct xfs_refcount_irec	tmp;
 	int				error;
@@ -497,6 +510,10 @@ xfs_refcount_find_left_extents(
 
 	if (RCNEXT(tmp) != agbno)
 		return 0;
+	if ((flags & XFS_FIND_RCEXT_SHARED) && tmp.rc_refcount < 2)
+		return 0;
+	if ((flags & XFS_FIND_RCEXT_COW) && tmp.rc_refcount > 1)
+		return 0;
 	/* We have a left extent; retrieve (or invent) the next right one */
 	*left = tmp;
 
@@ -553,7 +570,8 @@ xfs_refcount_find_right_extents(
 	struct xfs_refcount_irec	*right,
 	struct xfs_refcount_irec	*cright,
 	xfs_agblock_t			agbno,
-	xfs_extlen_t			aglen)
+	xfs_extlen_t			aglen,
+	int				flags)
 {
 	struct xfs_refcount_irec	tmp;
 	int				error;
@@ -573,6 +591,10 @@ xfs_refcount_find_right_extents(
 
 	if (tmp.rc_startblock != agbno + aglen)
 		return 0;
+	if ((flags & XFS_FIND_RCEXT_SHARED) && tmp.rc_refcount < 2)
+		return 0;
+	if ((flags & XFS_FIND_RCEXT_COW) && tmp.rc_refcount > 1)
+		return 0;
 	/* We have a right extent; retrieve (or invent) the next left one */
 	*right = tmp;
 
@@ -629,6 +651,7 @@ xfs_refcount_merge_extents(
 	xfs_agblock_t		*agbno,
 	xfs_extlen_t		*aglen,
 	enum xfs_refc_adjust_op adjust,
+	int			flags,
 	bool			*shape_changed)
 {
 	struct xfs_refcount_irec	left = {0}, cleft = {0};
@@ -644,11 +667,11 @@ xfs_refcount_merge_extents(
 	 * [right].
 	 */
 	error = xfs_refcount_find_left_extents(cur, &left, &cleft, *agbno,
-			*aglen);
+			*aglen, flags);
 	if (error)
 		return error;
 	error = xfs_refcount_find_right_extents(cur, &right, &cright, *agbno,
-			*aglen);
+			*aglen, flags);
 	if (error)
 		return error;
 
@@ -935,7 +958,7 @@ xfs_refcount_adjust(
 	 */
 	orig_aglen = aglen;
 	error = xfs_refcount_merge_extents(cur, &agbno, &aglen, adj,
-			&shape_changed);
+			XFS_FIND_RCEXT_SHARED, &shape_changed);
 	if (error)
 		goto out_error;
 	if (shape_changed)
@@ -1053,6 +1076,18 @@ xfs_refcount_finish_one(
 		error = xfs_refcount_adjust(rcur, bno, blockcount, adjusted,
 			XFS_REFCOUNT_ADJUST_DECREASE, dfops, NULL);
 		break;
+	case XFS_REFCOUNT_ALLOC_COW:
+		*adjusted = 0;
+		error = __xfs_refcount_cow_alloc(rcur, bno, blockcount, dfops);
+		if (!error)
+			*adjusted = blockcount;
+		break;
+	case XFS_REFCOUNT_FREE_COW:
+		*adjusted = 0;
+		error = __xfs_refcount_cow_free(rcur, bno, blockcount, dfops);
+		if (!error)
+			*adjusted = blockcount;
+		break;
 	default:
 		ASSERT(0);
 		error = -EFSCORRUPTED;
@@ -1227,3 +1262,275 @@ out_error:
 				cur->bc_private.a.agno, error, _RET_IP_);
 	return error;
 }
+
+/*
+ * Recovering CoW Blocks After a Crash
+ *
+ * Due to the way that the copy on write mechanism works, there's a window of
+ * opportunity in which we can lose track of allocated blocks during a crash.
+ * Because CoW uses delayed allocation in the in-core CoW fork, writeback
+ * causes blocks to be allocated and stored in the CoW fork.  The blocks are
+ * no longer in the free space btree but are not otherwise recorded anywhere
+ * until the write completes and the blocks are mapped into the file.  A crash
+ * in between allocation and remapping results in the replacement blocks being
+ * lost.  This situation is exacerbated by the CoW extent size hint because
+ * allocations can hang around for long time.
+ *
+ * However, there is a place where we can record these allocations before they
+ * become mappings -- the reference count btree.  The btree does not record
+ * extents with refcount == 1, so we can record allocations with a refcount of
+ * 1.  Blocks being used for CoW writeout cannot be shared, so there should be
+ * no conflict with shared block records.  These mappings should be created
+ * when we allocate blocks to the CoW fork and deleted when they're removed
+ * from the CoW fork.
+ *
+ * Minor nit: records for in-progress CoW allocations and records for shared
+ * extents must never be merged, to preserve the property that (except for CoW
+ * allocations) there are no refcount btree entries with refcount == 1.  The
+ * only time this could potentially happen is when unsharing a block that's
+ * adjacent to CoW allocations, so we must be careful to avoid this.
+ *
+ * At mount time we recover lost CoW allocations by searching the refcount
+ * btree for these refcount == 1 mappings.  These represent CoW allocations
+ * that were in progress at the time the filesystem went down, so we can free
+ * them to get the space back.
+ *
+ * This mechanism is superior to creating EFIs for unmapped CoW extents for
+ * several reasons -- first, EFIs pin the tail of the log and would have to be
+ * periodically relogged to avoid filling up the log.  Second, CoW completions
+ * will have to file an EFD and create new EFIs for whatever remains in the
+ * CoW fork; this partially takes care of (1) but extent-size reservations
+ * will have to periodically relog even if there's no writeout in progress.
+ * This can happen if the CoW extent size hint is set, which you really want.
+ * Third, EFIs cannot currently be automatically relogged into newer
+ * transactions to advance the log tail.  Fourth, stuffing the log full of
+ * EFIs places an upper bound on the number of CoW allocations that can be
+ * held filesystem-wide at any given time.  Recording them in the refcount
+ * btree doesn't require us to maintain any state in memory and doesn't pin
+ * the log.
+ */
+/*
+ * Adjust the refcounts of CoW allocations.  These allocations are "magic"
+ * in that they're not referenced anywhere else in the filesystem, so we
+ * stash them in the refcount btree with a refcount of 1 until either file
+ * remapping (or CoW cancellation) happens.
+ */
+STATIC int
+xfs_refcount_adjust_cow_extents(
+	struct xfs_btree_cur	*cur,
+	xfs_agblock_t		agbno,
+	xfs_extlen_t		aglen,
+	enum xfs_refc_adjust_op	adj,
+	struct xfs_defer_ops	*dfops,
+	struct xfs_owner_info	*oinfo)
+{
+	struct xfs_refcount_irec	ext, tmp;
+	int				error;
+	int				found_rec, found_tmp;
+
+	if (aglen == 0)
+		return 0;
+
+	/* Find any overlapping refcount records */
+	error = xfs_refcount_lookup_ge(cur, agbno, &found_rec);
+	if (error)
+		goto out_error;
+	error = xfs_refcount_get_rec(cur, &ext, &found_rec);
+	if (error)
+		goto out_error;
+	if (!found_rec) {
+		ext.rc_startblock = cur->bc_mp->m_sb.sb_agblocks;
+		ext.rc_blockcount = 0;
+		ext.rc_refcount = 0;
+	}
+
+	switch (adj) {
+	case XFS_REFCOUNT_ADJUST_COW_ALLOC:
+		/* Adding a CoW reservation, there should be nothing here. */
+		XFS_WANT_CORRUPTED_GOTO(cur->bc_mp,
+				ext.rc_startblock >= agbno + aglen, out_error);
+
+		tmp.rc_startblock = agbno;
+		tmp.rc_blockcount = aglen;
+		tmp.rc_refcount = 1;
+		trace_xfs_refcount_modify_extent(cur->bc_mp,
+				cur->bc_private.a.agno, &tmp);
+
+		error = xfs_refcount_insert(cur, &tmp,
+				&found_tmp);
+		if (error)
+			goto out_error;
+		XFS_WANT_CORRUPTED_GOTO(cur->bc_mp,
+				found_tmp == 1, out_error);
+		break;
+	case XFS_REFCOUNT_ADJUST_COW_FREE:
+		/* Removing a CoW reservation, there should be one extent. */
+		XFS_WANT_CORRUPTED_GOTO(cur->bc_mp,
+			ext.rc_startblock == agbno, out_error);
+		XFS_WANT_CORRUPTED_GOTO(cur->bc_mp,
+			ext.rc_blockcount == aglen, out_error);
+		XFS_WANT_CORRUPTED_GOTO(cur->bc_mp,
+			ext.rc_refcount == 1, out_error);
+
+		ext.rc_refcount = 0;
+		trace_xfs_refcount_modify_extent(cur->bc_mp,
+				cur->bc_private.a.agno, &ext);
+		error = xfs_refcount_delete(cur, &found_rec);
+		if (error)
+			goto out_error;
+		XFS_WANT_CORRUPTED_GOTO(cur->bc_mp,
+				found_rec == 1, out_error);
+		break;
+	default:
+		ASSERT(0);
+	}
+
+	return error;
+out_error:
+	trace_xfs_refcount_modify_extent_error(cur->bc_mp,
+			cur->bc_private.a.agno, error, _RET_IP_);
+	return error;
+}
+
+/*
+ * Add or remove refcount btree entries for CoW reservations.
+ */
+STATIC int
+xfs_refcount_adjust_cow(
+	struct xfs_btree_cur	*cur,
+	xfs_agblock_t		agbno,
+	xfs_extlen_t		aglen,
+	enum xfs_refc_adjust_op	adj,
+	struct xfs_defer_ops	*dfops)
+{
+	bool			shape_changed;
+	int			error;
+
+	/*
+	 * Ensure that no rcextents cross the boundary of the adjustment range.
+	 */
+	error = xfs_refcount_split_extent(cur, agbno, &shape_changed);
+	if (error)
+		goto out_error;
+
+	error = xfs_refcount_split_extent(cur, agbno + aglen, &shape_changed);
+	if (error)
+		goto out_error;
+
+	/*
+	 * Try to merge with the left or right extents of the range.
+	 */
+	error = xfs_refcount_merge_extents(cur, &agbno, &aglen, adj,
+			XFS_FIND_RCEXT_COW, &shape_changed);
+	if (error)
+		goto out_error;
+
+	/* Now that we've taken care of the ends, adjust the middle extents */
+	error = xfs_refcount_adjust_cow_extents(cur, agbno, aglen, adj,
+			dfops, NULL);
+	if (error)
+		goto out_error;
+
+	return 0;
+
+out_error:
+	trace_xfs_refcount_adjust_cow_error(cur->bc_mp, cur->bc_private.a.agno,
+			error, _RET_IP_);
+	return error;
+}
+
+/*
+ * Record a CoW allocation in the refcount btree.
+ */
+STATIC int
+__xfs_refcount_cow_alloc(
+	struct xfs_btree_cur	*rcur,
+	xfs_agblock_t		agbno,
+	xfs_extlen_t		aglen,
+	struct xfs_defer_ops	*dfops)
+{
+	int			error;
+
+	trace_xfs_refcount_cow_increase(rcur->bc_mp, rcur->bc_private.a.agno,
+			agbno, aglen);
+
+	/* Add refcount btree reservation */
+	error = xfs_refcount_adjust_cow(rcur, agbno, aglen,
+			XFS_REFCOUNT_ADJUST_COW_ALLOC, dfops);
+	if (error)
+		return error;
+
+	/* Add rmap entry */
+	if (xfs_sb_version_hasrmapbt(&rcur->bc_mp->m_sb)) {
+		error = xfs_rmap_alloc_extent(rcur->bc_mp, dfops,
+				rcur->bc_private.a.agno,
+				agbno, aglen, XFS_RMAP_OWN_COW);
+		if (error)
+			return error;
+	}
+
+	return error;
+}
+
+/*
+ * Remove a CoW allocation from the refcount btree.
+ */
+STATIC int
+__xfs_refcount_cow_free(
+	struct xfs_btree_cur	*rcur,
+	xfs_agblock_t		agbno,
+	xfs_extlen_t		aglen,
+	struct xfs_defer_ops	*dfops)
+{
+	int			error;
+
+	trace_xfs_refcount_cow_decrease(rcur->bc_mp, rcur->bc_private.a.agno,
+			agbno, aglen);
+
+	/* Remove refcount btree reservation */
+	error = xfs_refcount_adjust_cow(rcur, agbno, aglen,
+			XFS_REFCOUNT_ADJUST_COW_FREE, dfops);
+	if (error)
+		return error;
+
+	/* Remove rmap entry */
+	if (xfs_sb_version_hasrmapbt(&rcur->bc_mp->m_sb)) {
+		error = xfs_rmap_free_extent(rcur->bc_mp, dfops,
+				rcur->bc_private.a.agno,
+				agbno, aglen, XFS_RMAP_OWN_COW);
+		if (error)
+			return error;
+	}
+
+	return error;
+}
+
+/* Record a CoW staging extent in the refcount btree. */
+int
+xfs_refcount_alloc_cow_extent(
+	struct xfs_mount		*mp,
+	struct xfs_defer_ops		*dfops,
+	xfs_fsblock_t			fsb,
+	xfs_extlen_t			len)
+{
+	if (!xfs_sb_version_hasreflink(&mp->m_sb))
+		return 0;
+
+	return __xfs_refcount_add(mp, dfops, XFS_REFCOUNT_ALLOC_COW,
+			fsb, len);
+}
+
+/* Forget a CoW staging event in the refcount btree. */
+int
+xfs_refcount_free_cow_extent(
+	struct xfs_mount		*mp,
+	struct xfs_defer_ops		*dfops,
+	xfs_fsblock_t			fsb,
+	xfs_extlen_t			len)
+{
+	if (!xfs_sb_version_hasreflink(&mp->m_sb))
+		return 0;
+
+	return __xfs_refcount_add(mp, dfops, XFS_REFCOUNT_FREE_COW,
+			fsb, len);
+}
diff --git a/libxfs/xfs_refcount.h b/libxfs/xfs_refcount.h
index 0ec9a28..105c246 100644
--- a/libxfs/xfs_refcount.h
+++ b/libxfs/xfs_refcount.h
@@ -57,4 +57,11 @@ extern int xfs_refcount_find_shared(struct xfs_btree_cur *cur,
 		xfs_agblock_t agbno, xfs_extlen_t aglen, xfs_agblock_t *fbno,
 		xfs_extlen_t *flen, bool find_maximal);
 
+extern int xfs_refcount_alloc_cow_extent(struct xfs_mount *mp,
+		struct xfs_defer_ops *dfops, xfs_fsblock_t fsb,
+		xfs_extlen_t len);
+extern int xfs_refcount_free_cow_extent(struct xfs_mount *mp,
+		struct xfs_defer_ops *dfops, xfs_fsblock_t fsb,
+		xfs_extlen_t len);
+
 #endif	/* __XFS_REFCOUNT_H__ */

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 32/71] xfs: teach get_bmapx and fiemap about shared extents and the CoW fork

[Darrick J. Wong]

Teach xfs_getbmapx how to report shared extents and CoW fork contents,
then modify the FIEMAP formatters to set the appropriate flags.  A
previous version of this patch only modified the fiemap formatter,
which is insufficient.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_fs.h |    4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)


diff --git a/libxfs/xfs_fs.h b/libxfs/xfs_fs.h
index f291a53..bb70066 100644
--- a/libxfs/xfs_fs.h
+++ b/libxfs/xfs_fs.h
@@ -105,14 +105,16 @@ struct getbmapx {
 #define BMV_IF_PREALLOC		0x4	/* rtn status BMV_OF_PREALLOC if req */
 #define BMV_IF_DELALLOC		0x8	/* rtn status BMV_OF_DELALLOC if req */
 #define BMV_IF_NO_HOLES		0x10	/* Do not return holes */
+#define BMV_IF_COWFORK		0x20	/* return CoW fork rather than data */
 #define BMV_IF_VALID	\
 	(BMV_IF_ATTRFORK|BMV_IF_NO_DMAPI_READ|BMV_IF_PREALLOC|	\
-	 BMV_IF_DELALLOC|BMV_IF_NO_HOLES)
+	 BMV_IF_DELALLOC|BMV_IF_NO_HOLES|BMV_IF_COWFORK)
 
 /*	bmv_oflags values - returned for each non-header segment */
 #define BMV_OF_PREALLOC		0x1	/* segment = unwritten pre-allocation */
 #define BMV_OF_DELALLOC		0x2	/* segment = delayed allocation */
 #define BMV_OF_LAST		0x4	/* segment is the last in the file */
+#define BMV_OF_SHARED		0x8	/* segment shared with another file */
 
 /*
  * Structure for XFS_IOC_FSSETDM.

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 33/71] xfs: support FS_XFLAG_REFLINK on reflink filesystems

[Darrick J. Wong]

Add support for reporting the "reflink" inode flag in the XFS-specific
getxflags ioctl, and allow the user to clear the flag if file size is
zero.

v2: Move the reflink flag out of the way of the DAX flag, and add the
new cowextsize flag.

v3: do not report (or allow changes to) FL_NOCOW_FL, since we don't
support a flag to prevent CoWing and the reflink flag is a poor
proxy.  We'll try to design away the need for the NOCOW flag.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 include/darwin.h  |    8 ++++++++
 include/freebsd.h |    7 +++++++
 include/irix.h    |    7 +++++++
 include/linux.h   |    8 ++++++++
 4 files changed, 30 insertions(+)


diff --git a/include/darwin.h b/include/darwin.h
index 140386a..71a63a1 100644
--- a/include/darwin.h
+++ b/include/darwin.h
@@ -320,4 +320,12 @@ struct fsxattr {
 
 #endif
 
+#ifndef FS_XFLAG_REFLINK
+#define FS_XFLAG_REFLINK	0x00010000	/* file is reflinked */
+#endif
+
+#ifndef FS_XFLAG_COWEXTSIZE
+#define FS_XFLAG_COWEXTSIZE	0x00020000	/* CoW extent size allocator hint */
+#endif
+
 #endif	/* __XFS_DARWIN_H__ */
diff --git a/include/freebsd.h b/include/freebsd.h
index 668dcea..e8970a5 100644
--- a/include/freebsd.h
+++ b/include/freebsd.h
@@ -209,5 +209,12 @@ struct fsxattr {
 
 #endif
 
+#ifndef FS_XFLAG_REFLINK
+#define FS_XFLAG_REFLINK	0x00010000	/* file is reflinked */
+#endif
+
+#ifndef FS_XFLAG_COWEXTSIZE
+#define FS_XFLAG_COWEXTSIZE	0x00020000	/* CoW extent size allocator hint */
+#endif
 
 #endif	/* __XFS_FREEBSD_H__ */
diff --git a/include/irix.h b/include/irix.h
index 8b9d588..fea7c7c 100644
--- a/include/irix.h
+++ b/include/irix.h
@@ -454,6 +454,13 @@ struct fsxattr {
 
 #endif
 
+#ifndef FS_XFLAG_REFLINK
+#define FS_XFLAG_REFLINK	0x00010000	/* file is reflinked */
+#endif
+
+#ifndef FS_XFLAG_COWEXTSIZE
+#define FS_XFLAG_COWEXTSIZE	0x00020000	/* CoW extent size allocator hint */
+#endif
 
 /**
  * Abstraction of mountpoints.
diff --git a/include/linux.h b/include/linux.h
index 1663930..67337d6 100644
--- a/include/linux.h
+++ b/include/linux.h
@@ -209,4 +209,12 @@ struct fsxattr {
 
 #endif
 
+#ifndef FS_XFLAG_REFLINK
+#define FS_XFLAG_REFLINK	0x00010000	/* file is reflinked */
+#endif
+
+#ifndef FS_XFLAG_COWEXTSIZE
+#define FS_XFLAG_COWEXTSIZE	0x00020000	/* CoW extent size allocator hint */
+#endif
+
 #endif	/* __XFS_LINUX_H__ */

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 34/71] xfs: create a separate cow extent size hint for the allocator

[Darrick J. Wong]

Create a per-inode extent size allocator hint for copy-on-write.  This
hint is separate from the existing extent size hint so that CoW can
take advantage of the fragmentation-reducing properties of extent size
hints without disabling delalloc for regular writes.

The extent size hint that's fed to the allocator during a copy on
write operation is the greater of the cowextsize and regular extsize
hint.

During reflink, if we're sharing the entire source file to the entire
destination file and the destination file doesn't already have a
cowextsize hint, propagate the source file's cowextsize hint to the
destination file.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 include/darwin.h        |    3 ++-
 include/freebsd.h       |    3 ++-
 include/irix.h          |    3 ++-
 include/linux.h         |    3 ++-
 libxfs/libxfs_priv.h    |    1 +
 libxfs/xfs_bmap.c       |   13 +++++++++++--
 libxfs/xfs_format.h     |    3 ++-
 libxfs/xfs_fs.h         |    3 ++-
 libxfs/xfs_inode_buf.c  |    4 +++-
 libxfs/xfs_inode_buf.h  |    1 +
 libxfs/xfs_log_format.h |    3 ++-
 11 files changed, 30 insertions(+), 10 deletions(-)


diff --git a/include/darwin.h b/include/darwin.h
index 71a63a1..2b242e0 100644
--- a/include/darwin.h
+++ b/include/darwin.h
@@ -292,7 +292,8 @@ struct fsxattr {
 	__u32		fsx_extsize;	/* extsize field value (get/set)*/
 	__u32		fsx_nextents;	/* nextents field value (get)	*/
 	__u32		fsx_projid;	/* project identifier (get/set) */
-	unsigned char	fsx_pad[12];
+	__u32		fsx_cowextsize;	/* cow extsize field value (get/set) */
+	unsigned char	fsx_pad[8];
 };
 
 /*
diff --git a/include/freebsd.h b/include/freebsd.h
index e8970a5..8ee7279 100644
--- a/include/freebsd.h
+++ b/include/freebsd.h
@@ -181,7 +181,8 @@ struct fsxattr {
 	__u32		fsx_extsize;	/* extsize field value (get/set)*/
 	__u32		fsx_nextents;	/* nextents field value (get)	*/
 	__u32		fsx_projid;	/* project identifier (get/set) */
-	unsigned char	fsx_pad[12];
+	__u32		fsx_cowextsize;	/* cow extsize field value (get/set) */
+	unsigned char	fsx_pad[8];
 };
 
 /*
diff --git a/include/irix.h b/include/irix.h
index fea7c7c..d4f9467 100644
--- a/include/irix.h
+++ b/include/irix.h
@@ -426,7 +426,8 @@ struct fsxattr {
 	__u32		fsx_extsize;	/* extsize field value (get/set)*/
 	__u32		fsx_nextents;	/* nextents field value (get)	*/
 	__u32		fsx_projid;	/* project identifier (get/set) */
-	unsigned char	fsx_pad[12];
+	__u32		fsx_cowextsize;	/* cow extsize field value (get/set) */
+	unsigned char	fsx_pad[8];
 };
 
 /*
diff --git a/include/linux.h b/include/linux.h
index 67337d6..bc7af9f 100644
--- a/include/linux.h
+++ b/include/linux.h
@@ -181,7 +181,8 @@ struct fsxattr {
 	__u32		fsx_extsize;	/* extsize field value (get/set)*/
 	__u32		fsx_nextents;	/* nextents field value (get)	*/
 	__u32		fsx_projid;	/* project identifier (get/set) */
-	unsigned char	fsx_pad[12];
+	__u32		fsx_cowextsize;	/* cow extsize field value (get/set) */
+	unsigned char	fsx_pad[8];
 };
 
 /*
diff --git a/libxfs/libxfs_priv.h b/libxfs/libxfs_priv.h
index ba4503d..8e7cb33 100644
--- a/libxfs/libxfs_priv.h
+++ b/libxfs/libxfs_priv.h
@@ -426,6 +426,7 @@ do { \
 #define xfs_rotorstep				1
 #define xfs_bmap_rtalloc(a)			(-ENOSYS)
 #define xfs_get_extsz_hint(ip)			(0)
+#define xfs_get_cowextsz_hint(ip)		(0)
 #define xfs_inode_is_filestream(ip)		(0)
 #define xfs_filestream_lookup_ag(ip)		(0)
 #define xfs_filestream_new_ag(ip,ag)		(0)
diff --git a/libxfs/xfs_bmap.c b/libxfs/xfs_bmap.c
index a94f855..875a15f 100644
--- a/libxfs/xfs_bmap.c
+++ b/libxfs/xfs_bmap.c
@@ -3657,7 +3657,13 @@ xfs_bmap_btalloc(
 	else if (mp->m_dalign)
 		stripe_align = mp->m_dalign;
 
-	align = ap->userdata ? xfs_get_extsz_hint(ap->ip) : 0;
+	if (ap->userdata) {
+		if (ap->flags & XFS_BMAPI_COWFORK)
+			align = xfs_get_cowextsz_hint(ap->ip);
+		else
+			align = xfs_get_extsz_hint(ap->ip);
+	} else
+		align = 0;
 	if (unlikely(align)) {
 		error = xfs_bmap_extsize_align(mp, &ap->got, &ap->prev,
 						align, 0, ap->eof, 0, ap->conv,
@@ -4171,7 +4177,10 @@ xfs_bmapi_reserve_delalloc(
 		alen = XFS_FILBLKS_MIN(alen, got->br_startoff - aoff);
 
 	/* Figure out the extent size, adjust alen */
-	extsz = xfs_get_extsz_hint(ip);
+	if (whichfork == XFS_COW_FORK)
+		extsz = xfs_get_cowextsz_hint(ip);
+	else
+		extsz = xfs_get_extsz_hint(ip);
 	if (extsz) {
 		error = xfs_bmap_extsize_align(mp, got, prev, extsz, rt, eof,
 					       1, 0, &aoff, &alen);
diff --git a/libxfs/xfs_format.h b/libxfs/xfs_format.h
index 3991eaa..e1551f1 100644
--- a/libxfs/xfs_format.h
+++ b/libxfs/xfs_format.h
@@ -900,7 +900,8 @@ typedef struct xfs_dinode {
 	__be64		di_changecount;	/* number of attribute changes */
 	__be64		di_lsn;		/* flush sequence */
 	__be64		di_flags2;	/* more random flags */
-	__u8		di_pad2[16];	/* more padding for future expansion */
+	__be32		di_cowextsize;	/* basic cow extent size for file */
+	__u8		di_pad2[12];	/* more padding for future expansion */
 
 	/* fields only written to during inode creation */
 	xfs_timestamp_t	di_crtime;	/* time created */
diff --git a/libxfs/xfs_fs.h b/libxfs/xfs_fs.h
index bb70066..df58c1c 100644
--- a/libxfs/xfs_fs.h
+++ b/libxfs/xfs_fs.h
@@ -302,7 +302,8 @@ typedef struct xfs_bstat {
 #define	bs_projid	bs_projid_lo	/* (previously just bs_projid)	*/
 	__u16		bs_forkoff;	/* inode fork offset in bytes	*/
 	__u16		bs_projid_hi;	/* higher part of project id	*/
-	unsigned char	bs_pad[10];	/* pad space, unused		*/
+	unsigned char	bs_pad[6];	/* pad space, unused		*/
+	__u32		bs_cowextsize;	/* cow extent size		*/
 	__u32		bs_dmevmask;	/* DMIG event mask		*/
 	__u16		bs_dmstate;	/* DMIG state info		*/
 	__u16		bs_aextents;	/* attribute number of extents	*/
diff --git a/libxfs/xfs_inode_buf.c b/libxfs/xfs_inode_buf.c
index 572c101..8a804e2 100644
--- a/libxfs/xfs_inode_buf.c
+++ b/libxfs/xfs_inode_buf.c
@@ -265,6 +265,7 @@ xfs_inode_from_disk(
 		to->di_crtime.t_sec = be32_to_cpu(from->di_crtime.t_sec);
 		to->di_crtime.t_nsec = be32_to_cpu(from->di_crtime.t_nsec);
 		to->di_flags2 = be64_to_cpu(from->di_flags2);
+		to->di_cowextsize = be32_to_cpu(from->di_cowextsize);
 	}
 }
 
@@ -314,7 +315,7 @@ xfs_inode_to_disk(
 		to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec);
 		to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec);
 		to->di_flags2 = cpu_to_be64(from->di_flags2);
-
+		to->di_cowextsize = cpu_to_be32(from->di_cowextsize);
 		to->di_ino = cpu_to_be64(ip->i_ino);
 		to->di_lsn = cpu_to_be64(lsn);
 		memset(to->di_pad2, 0, sizeof(to->di_pad2));
@@ -366,6 +367,7 @@ xfs_log_dinode_to_disk(
 		to->di_crtime.t_sec = cpu_to_be32(from->di_crtime.t_sec);
 		to->di_crtime.t_nsec = cpu_to_be32(from->di_crtime.t_nsec);
 		to->di_flags2 = cpu_to_be64(from->di_flags2);
+		to->di_cowextsize = cpu_to_be32(from->di_cowextsize);
 		to->di_ino = cpu_to_be64(from->di_ino);
 		to->di_lsn = cpu_to_be64(from->di_lsn);
 		memcpy(to->di_pad2, from->di_pad2, sizeof(to->di_pad2));
diff --git a/libxfs/xfs_inode_buf.h b/libxfs/xfs_inode_buf.h
index 958c543..6848a0a 100644
--- a/libxfs/xfs_inode_buf.h
+++ b/libxfs/xfs_inode_buf.h
@@ -47,6 +47,7 @@ struct xfs_icdinode {
 	__uint16_t	di_flags;	/* random flags, XFS_DIFLAG_... */
 
 	__uint64_t	di_flags2;	/* more random flags */
+	__uint32_t	di_cowextsize;	/* basic cow extent size for file */
 
 	xfs_ictimestamp_t di_crtime;	/* time created */
 };
diff --git a/libxfs/xfs_log_format.h b/libxfs/xfs_log_format.h
index fffcc0f..d788d01 100644
--- a/libxfs/xfs_log_format.h
+++ b/libxfs/xfs_log_format.h
@@ -423,7 +423,8 @@ struct xfs_log_dinode {
 	__uint64_t	di_changecount;	/* number of attribute changes */
 	xfs_lsn_t	di_lsn;		/* flush sequence */
 	__uint64_t	di_flags2;	/* more random flags */
-	__uint8_t	di_pad2[16];	/* more padding for future expansion */
+	__uint32_t	di_cowextsize;	/* basic cow extent size for file */
+	__uint8_t	di_pad2[12];	/* more padding for future expansion */
 
 	/* fields only written to during inode creation */
 	xfs_ictimestamp_t di_crtime;	/* time created */

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 35/71] xfs: preallocate blocks for worst-case btree expansion

[Darrick J. Wong]

To gracefully handle the situation where a CoW operation turns a
single refcount extent into a lot of tiny ones and then run out of
space when a tree split has to happen, use the per-AG reserved block
pool to pre-allocate all the space we'll ever need for a maximal
btree.  For a 4K block size, this only costs an overhead of 0.3% of
available disk space.

When reflink is enabled, we have an unfortunate problem with rmap --
since we can share a block billions of times, this means that the
reverse mapping btree can expand basically infinitely.  When an AG is
so full that there are no free blocks with which to expand the rmapbt,
the filesystem will shut down hard.

This is rather annoying to the user, so use the AG reservation code to
reserve a "reasonable" amount of space for rmap.  We'll prevent
reflinks and CoW operations if we think we're getting close to
exhausting an AG's free space rather than shutting down, but this
permanent reservation should be enough for "most" users.  Hopefully.

v2: Simplify the return value from xfs_perag_pool_free_block to a bool
so that we can easily call xfs_trans_binval for both the per-AG pool
and the real freeing case.  Without this we fail to invalidate the
btree buffer and will trip over the write verifier on a shrinking
refcount btree.

v3: Convert to the new per-AG reservation code.

v4: Combine this patch with the one that adds the rmapbt reservation,
since the rmapbt reservation is only needed for reflink filesystems.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
[hch@lst.de: ensure that we invalidate the freed btree buffer]
Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 libxfs/xfs_ag_resv.c        |   11 ++++++++
 libxfs/xfs_refcount_btree.c |   45 ++++++++++++++++++++++++++++++--
 libxfs/xfs_refcount_btree.h |    3 ++
 libxfs/xfs_rmap_btree.c     |   60 +++++++++++++++++++++++++++++++++++++++++++
 libxfs/xfs_rmap_btree.h     |    7 +++++
 5 files changed, 123 insertions(+), 3 deletions(-)


diff --git a/libxfs/xfs_ag_resv.c b/libxfs/xfs_ag_resv.c
index af69235..9d338bd 100644
--- a/libxfs/xfs_ag_resv.c
+++ b/libxfs/xfs_ag_resv.c
@@ -37,6 +37,7 @@
 #include "xfs_trans_space.h"
 #include "xfs_rmap_btree.h"
 #include "xfs_btree.h"
+#include "xfs_refcount_btree.h"
 
 /*
  * Per-AG Block Reservations
@@ -229,6 +230,11 @@ xfs_ag_resv_init(
 	/* Create the metadata reservation. */
 	ask = used = 0;
 
+	err2 = xfs_refcountbt_calc_reserves(pag->pag_mount, pag->pag_agno,
+			&ask, &used);
+	if (err2 && !error)
+		error = err2;
+
 	err2 = __xfs_ag_resv_init(pag, XFS_AG_RESV_METADATA, ask, used);
 	if (err2 && !error)
 		error = err2;
@@ -240,6 +246,11 @@ init_agfl:
 	/* Create the AGFL metadata reservation */
 	ask = used = 0;
 
+	err2 = xfs_rmapbt_calc_reserves(pag->pag_mount, pag->pag_agno,
+			&ask, &used);
+	if (err2 && !error)
+		error = err2;
+
 	err2 = __xfs_ag_resv_init(pag, XFS_AG_RESV_AGFL, ask, used);
 	if (err2 && !error)
 		error = err2;
diff --git a/libxfs/xfs_refcount_btree.c b/libxfs/xfs_refcount_btree.c
index 568a2f8..d2dbdbd 100644
--- a/libxfs/xfs_refcount_btree.c
+++ b/libxfs/xfs_refcount_btree.c
@@ -78,6 +78,8 @@ xfs_refcountbt_alloc_block(
 	struct xfs_alloc_arg	args;		/* block allocation args */
 	int			error;		/* error return value */
 
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
+
 	memset(&args, 0, sizeof(args));
 	args.tp = cur->bc_tp;
 	args.mp = cur->bc_mp;
@@ -87,6 +89,7 @@ xfs_refcountbt_alloc_block(
 	args.firstblock = args.fsbno;
 	xfs_rmap_ag_owner(&args.oinfo, XFS_RMAP_OWN_REFC);
 	args.minlen = args.maxlen = args.prod = 1;
+	args.resv = XFS_AG_RESV_METADATA;
 
 	error = xfs_alloc_vextent(&args);
 	if (error)
@@ -124,16 +127,19 @@ xfs_refcountbt_free_block(
 	struct xfs_agf		*agf = XFS_BUF_TO_AGF(agbp);
 	xfs_fsblock_t		fsbno = XFS_DADDR_TO_FSB(mp, XFS_BUF_ADDR(bp));
 	struct xfs_owner_info	oinfo;
+	int			error;
 
 	trace_xfs_refcountbt_free_block(cur->bc_mp, cur->bc_private.a.agno,
 			XFS_FSB_TO_AGBNO(cur->bc_mp, fsbno), 1);
 	xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_REFC);
 	be32_add_cpu(&agf->agf_refcount_blocks, -1);
 	xfs_alloc_log_agf(cur->bc_tp, agbp, XFS_AGF_REFCOUNT_BLOCKS);
-	xfs_bmap_add_free(mp, cur->bc_private.a.dfops, fsbno, 1,
-			&oinfo);
+	error = xfs_free_extent(cur->bc_tp, fsbno, 1, &oinfo,
+			XFS_AG_RESV_METADATA);
+	if (error)
+		return error;
 
-	return 0;
+	return error;
 }
 
 STATIC int
@@ -403,3 +409,36 @@ xfs_refcountbt_max_size(
 
 	return xfs_refcountbt_calc_size(mp, mp->m_sb.sb_agblocks);
 }
+
+/*
+ * Figure out how many blocks to reserve and how many are used by this btree.
+ */
+int
+xfs_refcountbt_calc_reserves(
+	struct xfs_mount	*mp,
+	xfs_agnumber_t		agno,
+	xfs_extlen_t		*ask,
+	xfs_extlen_t		*used)
+{
+	struct xfs_buf		*agbp;
+	struct xfs_agf		*agf;
+	xfs_extlen_t		tree_len;
+	int			error;
+
+	if (!xfs_sb_version_hasreflink(&mp->m_sb))
+		return 0;
+
+	*ask += xfs_refcountbt_max_size(mp);
+
+	error = xfs_alloc_read_agf(mp, NULL, agno, 0, &agbp);
+	if (error)
+		return error;
+
+	agf = XFS_BUF_TO_AGF(agbp);
+	tree_len = be32_to_cpu(agf->agf_refcount_blocks);
+	xfs_buf_relse(agbp);
+
+	*used += tree_len;
+
+	return error;
+}
diff --git a/libxfs/xfs_refcount_btree.h b/libxfs/xfs_refcount_btree.h
index 780b02f..3be7768 100644
--- a/libxfs/xfs_refcount_btree.h
+++ b/libxfs/xfs_refcount_btree.h
@@ -68,4 +68,7 @@ extern xfs_extlen_t xfs_refcountbt_calc_size(struct xfs_mount *mp,
 		unsigned long long len);
 extern xfs_extlen_t xfs_refcountbt_max_size(struct xfs_mount *mp);
 
+extern int xfs_refcountbt_calc_reserves(struct xfs_mount *mp,
+		xfs_agnumber_t agno, xfs_extlen_t *ask, xfs_extlen_t *used);
+
 #endif	/* __XFS_REFCOUNT_BTREE_H__ */
diff --git a/libxfs/xfs_rmap_btree.c b/libxfs/xfs_rmap_btree.c
index 29a4dd6..02ceace 100644
--- a/libxfs/xfs_rmap_btree.c
+++ b/libxfs/xfs_rmap_btree.c
@@ -33,6 +33,7 @@
 #include "xfs_rmap_btree.h"
 #include "xfs_trace.h"
 #include "xfs_cksum.h"
+#include "xfs_ag_resv.h"
 
 /*
  * Reverse map btree.
@@ -516,3 +517,62 @@ xfs_rmapbt_compute_maxlevels(
 		mp->m_rmap_maxlevels = xfs_btree_compute_maxlevels(mp,
 				mp->m_rmap_mnr, mp->m_sb.sb_agblocks);
 }
+
+/* Calculate the refcount btree size for some records. */
+xfs_extlen_t
+xfs_rmapbt_calc_size(
+	struct xfs_mount	*mp,
+	unsigned long long	len)
+{
+	return xfs_btree_calc_size(mp, mp->m_rmap_mnr, len);
+}
+
+/*
+ * Calculate the maximum refcount btree size.
+ */
+xfs_extlen_t
+xfs_rmapbt_max_size(
+	struct xfs_mount	*mp)
+{
+	/* Bail out if we're uninitialized, which can happen in mkfs. */
+	if (mp->m_rmap_mxr[0] == 0)
+		return 0;
+
+	return xfs_rmapbt_calc_size(mp, mp->m_sb.sb_agblocks);
+}
+
+/*
+ * Figure out how many blocks to reserve and how many are used by this btree.
+ */
+int
+xfs_rmapbt_calc_reserves(
+	struct xfs_mount	*mp,
+	xfs_agnumber_t		agno,
+	xfs_extlen_t		*ask,
+	xfs_extlen_t		*used)
+{
+	struct xfs_buf		*agbp;
+	struct xfs_agf		*agf;
+	xfs_extlen_t		pool_len;
+	xfs_extlen_t		tree_len;
+	int			error;
+
+	if (!xfs_sb_version_hasrmapbt(&mp->m_sb))
+		return 0;
+
+	/* Reserve 1% of the AG or enough for 1 block per record. */
+	pool_len = max(mp->m_sb.sb_agblocks / 100, xfs_rmapbt_max_size(mp));
+	*ask += pool_len;
+
+	error = xfs_alloc_read_agf(mp, NULL, agno, 0, &agbp);
+	if (error)
+		return error;
+
+	agf = XFS_BUF_TO_AGF(agbp);
+	tree_len = be32_to_cpu(agf->agf_rmap_blocks);
+	xfs_buf_relse(agbp);
+
+	*used += tree_len;
+
+	return error;
+}
diff --git a/libxfs/xfs_rmap_btree.h b/libxfs/xfs_rmap_btree.h
index 5ff9cfa..f3137a3 100644
--- a/libxfs/xfs_rmap_btree.h
+++ b/libxfs/xfs_rmap_btree.h
@@ -58,4 +58,11 @@ struct xfs_btree_cur *xfs_rmapbt_init_cursor(struct xfs_mount *mp,
 int xfs_rmapbt_maxrecs(struct xfs_mount *mp, int blocklen, int leaf);
 extern void xfs_rmapbt_compute_maxlevels(struct xfs_mount *mp);
 
+extern xfs_extlen_t xfs_rmapbt_calc_size(struct xfs_mount *mp,
+		unsigned long long len);
+extern xfs_extlen_t xfs_rmapbt_max_size(struct xfs_mount *mp);
+
+extern int xfs_rmapbt_calc_reserves(struct xfs_mount *mp,
+		xfs_agnumber_t agno, xfs_extlen_t *ask, xfs_extlen_t *used);
+
 #endif /* __XFS_RMAP_BTREE_H__ */

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 36/71] xfs: try other AGs to allocate a BMBT block

[Darrick J. Wong]

Prior to the introduction of reflink, allocating a block and mapping
it into a file was performed in a single transaction with a single
block reservation, and the allocator was supposed to find enough
blocks to allocate the extent and any BMBT blocks that might be
necessary (unless we're low on space).

However, due to the way copy on write works, allocation and mapping
have been split into two transactions, which means that we must be
able to handle the case where we allocate an extent for CoW but that
AG runs out of free space before the blocks can be mapped into a file,
and the mapping requires a new BMBT block.  When this happens, look in
one of the other AGs for a BMBT block instead of taking the FS down.

The same applies to the functions that convert a data fork to extents
and later btree format.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_bmap.c       |   30 ++++++++++++++++++++++++++++++
 libxfs/xfs_bmap_btree.c |   17 +++++++++++++++++
 2 files changed, 47 insertions(+)


diff --git a/libxfs/xfs_bmap.c b/libxfs/xfs_bmap.c
index 875a15f..4721cff 100644
--- a/libxfs/xfs_bmap.c
+++ b/libxfs/xfs_bmap.c
@@ -745,6 +745,7 @@ xfs_bmap_extents_to_btree(
 		args.type = XFS_ALLOCTYPE_START_BNO;
 		args.fsbno = XFS_INO_TO_FSB(mp, ip->i_ino);
 	} else if (dfops->dop_low) {
+try_another_ag:
 		args.type = XFS_ALLOCTYPE_START_BNO;
 		args.fsbno = *firstblock;
 	} else {
@@ -759,6 +760,21 @@ xfs_bmap_extents_to_btree(
 		xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
 		return error;
 	}
+
+	/*
+	 * During a CoW operation, the allocation and bmbt updates occur in
+	 * different transactions.  The mapping code tries to put new bmbt
+	 * blocks near extents being mapped, but the only way to guarantee this
+	 * is if the alloc and the mapping happen in a single transaction that
+	 * has a block reservation.  That isn't the case here, so if we run out
+	 * of space we'll try again with another AG.
+	 */
+	if (xfs_sb_version_hasreflink(&cur->bc_mp->m_sb) &&
+	    args.fsbno == NULLFSBLOCK &&
+	    args.type == XFS_ALLOCTYPE_NEAR_BNO) {
+		dfops->dop_low = true;
+		goto try_another_ag;
+	}
 	/*
 	 * Allocation can't fail, the space was reserved.
 	 */
@@ -894,6 +910,7 @@ xfs_bmap_local_to_extents(
 	 * file currently fits in an inode.
 	 */
 	if (*firstblock == NULLFSBLOCK) {
+try_another_ag:
 		args.fsbno = XFS_INO_TO_FSB(args.mp, ip->i_ino);
 		args.type = XFS_ALLOCTYPE_START_BNO;
 	} else {
@@ -906,6 +923,19 @@ xfs_bmap_local_to_extents(
 	if (error)
 		goto done;
 
+	/*
+	 * During a CoW operation, the allocation and bmbt updates occur in
+	 * different transactions.  The mapping code tries to put new bmbt
+	 * blocks near extents being mapped, but the only way to guarantee this
+	 * is if the alloc and the mapping happen in a single transaction that
+	 * has a block reservation.  That isn't the case here, so if we run out
+	 * of space we'll try again with another AG.
+	 */
+	if (xfs_sb_version_hasreflink(&ip->i_mount->m_sb) &&
+	    args.fsbno == NULLFSBLOCK &&
+	    args.type == XFS_ALLOCTYPE_NEAR_BNO) {
+		goto try_another_ag;
+	}
 	/* Can't fail, the space was reserved. */
 	ASSERT(args.fsbno != NULLFSBLOCK);
 	ASSERT(args.len == 1);
diff --git a/libxfs/xfs_bmap_btree.c b/libxfs/xfs_bmap_btree.c
index 3d1a02e..601385d 100644
--- a/libxfs/xfs_bmap_btree.c
+++ b/libxfs/xfs_bmap_btree.c
@@ -450,6 +450,7 @@ xfs_bmbt_alloc_block(
 
 	if (args.fsbno == NULLFSBLOCK) {
 		args.fsbno = be64_to_cpu(start->l);
+try_another_ag:
 		args.type = XFS_ALLOCTYPE_START_BNO;
 		/*
 		 * Make sure there is sufficient room left in the AG to
@@ -479,6 +480,22 @@ xfs_bmbt_alloc_block(
 	if (error)
 		goto error0;
 
+	/*
+	 * During a CoW operation, the allocation and bmbt updates occur in
+	 * different transactions.  The mapping code tries to put new bmbt
+	 * blocks near extents being mapped, but the only way to guarantee this
+	 * is if the alloc and the mapping happen in a single transaction that
+	 * has a block reservation.  That isn't the case here, so if we run out
+	 * of space we'll try again with another AG.
+	 */
+	if (xfs_sb_version_hasreflink(&cur->bc_mp->m_sb) &&
+	    args.fsbno == NULLFSBLOCK &&
+	    args.type == XFS_ALLOCTYPE_NEAR_BNO) {
+		cur->bc_private.b.dfops->dop_low = true;
+		args.fsbno = cur->bc_private.b.firstblock;
+		goto try_another_ag;
+	}
+
 	if (args.fsbno == NULLFSBLOCK && args.minleft) {
 		/*
 		 * Could not find an AG with enough free space to satisfy

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 37/71] xfs: increase log reservations for reflink

[Darrick J. Wong]

Increase the log reservations to handle the increased rolling that
happens at the end of copy-on-write operations.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_trans_resv.c |   16 +++++++++++++---
 libxfs/xfs_trans_resv.h |    2 ++
 2 files changed, 15 insertions(+), 3 deletions(-)


diff --git a/libxfs/xfs_trans_resv.c b/libxfs/xfs_trans_resv.c
index 5b6bbcd..5152a5b 100644
--- a/libxfs/xfs_trans_resv.c
+++ b/libxfs/xfs_trans_resv.c
@@ -811,11 +811,18 @@ xfs_trans_resv_calc(
 	 * require a permanent reservation on space.
 	 */
 	resp->tr_write.tr_logres = xfs_calc_write_reservation(mp);
-	resp->tr_write.tr_logcount = XFS_WRITE_LOG_COUNT;
+	if (xfs_sb_version_hasreflink(&mp->m_sb))
+		resp->tr_write.tr_logcount = XFS_WRITE_LOG_COUNT_REFLINK;
+	else
+		resp->tr_write.tr_logcount = XFS_WRITE_LOG_COUNT;
 	resp->tr_write.tr_logflags |= XFS_TRANS_PERM_LOG_RES;
 
 	resp->tr_itruncate.tr_logres = xfs_calc_itruncate_reservation(mp);
-	resp->tr_itruncate.tr_logcount = XFS_ITRUNCATE_LOG_COUNT;
+	if (xfs_sb_version_hasreflink(&mp->m_sb))
+		resp->tr_itruncate.tr_logcount =
+				XFS_ITRUNCATE_LOG_COUNT_REFLINK;
+	else
+		resp->tr_itruncate.tr_logcount = XFS_ITRUNCATE_LOG_COUNT;
 	resp->tr_itruncate.tr_logflags |= XFS_TRANS_PERM_LOG_RES;
 
 	resp->tr_rename.tr_logres = xfs_calc_rename_reservation(mp);
@@ -872,7 +879,10 @@ xfs_trans_resv_calc(
 	resp->tr_growrtalloc.tr_logflags |= XFS_TRANS_PERM_LOG_RES;
 
 	resp->tr_qm_dqalloc.tr_logres = xfs_calc_qm_dqalloc_reservation(mp);
-	resp->tr_qm_dqalloc.tr_logcount = XFS_WRITE_LOG_COUNT;
+	if (xfs_sb_version_hasreflink(&mp->m_sb))
+		resp->tr_qm_dqalloc.tr_logcount = XFS_WRITE_LOG_COUNT_REFLINK;
+	else
+		resp->tr_qm_dqalloc.tr_logcount = XFS_WRITE_LOG_COUNT;
 	resp->tr_qm_dqalloc.tr_logflags |= XFS_TRANS_PERM_LOG_RES;
 
 	/*
diff --git a/libxfs/xfs_trans_resv.h b/libxfs/xfs_trans_resv.h
index 36a1511..b7e5357 100644
--- a/libxfs/xfs_trans_resv.h
+++ b/libxfs/xfs_trans_resv.h
@@ -87,6 +87,7 @@ struct xfs_trans_resv {
 #define	XFS_DEFAULT_LOG_COUNT		1
 #define	XFS_DEFAULT_PERM_LOG_COUNT	2
 #define	XFS_ITRUNCATE_LOG_COUNT		2
+#define	XFS_ITRUNCATE_LOG_COUNT_REFLINK	8
 #define XFS_INACTIVE_LOG_COUNT		2
 #define	XFS_CREATE_LOG_COUNT		2
 #define	XFS_CREATE_TMPFILE_LOG_COUNT	2
@@ -96,6 +97,7 @@ struct xfs_trans_resv {
 #define	XFS_LINK_LOG_COUNT		2
 #define	XFS_RENAME_LOG_COUNT		2
 #define	XFS_WRITE_LOG_COUNT		2
+#define	XFS_WRITE_LOG_COUNT_REFLINK	8
 #define	XFS_ADDAFORK_LOG_COUNT		2
 #define	XFS_ATTRINVAL_LOG_COUNT		1
 #define	XFS_ATTRSET_LOG_COUNT		3

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 38/71] xfs: add shared rmap map/unmap/convert log item types

[Darrick J. Wong]

Wire up some rmap log redo item type codes to map, unmap, or convert
shared data block extents.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_log_format.h |    3 +++
 1 file changed, 3 insertions(+)


diff --git a/libxfs/xfs_log_format.h b/libxfs/xfs_log_format.h
index d788d01..75f9890 100644
--- a/libxfs/xfs_log_format.h
+++ b/libxfs/xfs_log_format.h
@@ -635,8 +635,11 @@ struct xfs_map_extent {
 
 /* rmap me_flags: upper bits are flags, lower byte is type code */
 #define XFS_RMAP_EXTENT_MAP		1
+#define XFS_RMAP_EXTENT_MAP_SHARED	2
 #define XFS_RMAP_EXTENT_UNMAP		3
+#define XFS_RMAP_EXTENT_UNMAP_SHARED	4
 #define XFS_RMAP_EXTENT_CONVERT		5
+#define XFS_RMAP_EXTENT_CONVERT_SHARED	6
 #define XFS_RMAP_EXTENT_ALLOC		7
 #define XFS_RMAP_EXTENT_FREE		8
 #define XFS_RMAP_EXTENT_TYPE_MASK	0xFF

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 39/71] xfs: use interval query for rmap map and unmap operations on shared files

[Darrick J. Wong]

When it's possible for reverse mappings to overlap (data fork extents
of files on reflink filesystems), use the interval query function to
find the left neighbor of an extent we're trying to add; and be
careful to use the lookup functions to update the neighbors and/or
add new extents.

v2: xfs_rmap_find_left_neighbor() needs to calculate the high key of a
query range correctly.  We can also add a few shortcuts -- there are
no left neighbors of a query at offset zero.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 include/xfs_trace.h |   14 +
 libxfs/xfs_rmap.c   |  514 +++++++++++++++++++++++++++++++++++++++++++++++++++
 libxfs/xfs_rmap.h   |    7 +
 3 files changed, 532 insertions(+), 3 deletions(-)


diff --git a/include/xfs_trace.h b/include/xfs_trace.h
index b7bf3d0..24b99dd 100644
--- a/include/xfs_trace.h
+++ b/include/xfs_trace.h
@@ -209,7 +209,6 @@
 #define trace_xfs_rmap_defer(...)		((void) 0)
 #define trace_xfs_rmap_deferred(...)		((void) 0)
 #define trace_xfs_rmap_find_right_neighbor_result(...)	((void) 0)
-#define trace_xfs_rmap_find_left_neighbor_result(...)	((void) 0)
 #define trace_xfs_rmap_lookup_le_range_result(...)	((void) 0)
 
 #define trace_xfs_rmapbt_free_block(...)	((void) 0)
@@ -268,6 +267,19 @@
 #define trace_xfs_refcount_cow_increase(...)	((void) 0)
 #define trace_xfs_refcount_cow_decrease(...)	((void) 0)
 
+#define trace_xfs_rmap_find_left_neighbor_candidate(...)	((void) 0)
+#define trace_xfs_rmap_find_left_neighbor_query(...)		((void) 0)
+#define trace_xfs_rmap_find_left_neighbor_result(...)		((void) 0)
+#define trace_xfs_rmap_lookup_le_range_candidate(...)		((void) 0)
+#define trace_xfs_rmap_lookup_le_range(...)	((void) 0)
+#define trace_xfs_rmap_unmap(...)		((void) 0)
+#define trace_xfs_rmap_unmap_done(...)		((void) 0)
+#define trace_xfs_rmap_unmap_error(...)		((void) 0)
+#define trace_xfs_rmap_map(...)			((void) 0)
+#define trace_xfs_rmap_map_done(...)		((void) 0)
+#define trace_xfs_rmap_map_error(...)		((void) 0)
+#define trace_xfs_rmap_delete_error(...)	((void) 0)
+
 /* set c = c to avoid unused var warnings */
 #define trace_xfs_perag_get(a,b,c,d)	((c) = (c))
 #define trace_xfs_perag_get_tag(a,b,c,d) ((c) = (c))
diff --git a/libxfs/xfs_rmap.c b/libxfs/xfs_rmap.c
index 82c2597..17dc7d6 100644
--- a/libxfs/xfs_rmap.c
+++ b/libxfs/xfs_rmap.c
@@ -146,6 +146,37 @@ done:
 	return error;
 }
 
+STATIC int
+xfs_rmap_delete(
+	struct xfs_btree_cur	*rcur,
+	xfs_agblock_t		agbno,
+	xfs_extlen_t		len,
+	uint64_t		owner,
+	uint64_t		offset,
+	unsigned int		flags)
+{
+	int			i;
+	int			error;
+
+	trace_xfs_rmap_delete(rcur->bc_mp, rcur->bc_private.a.agno, agbno,
+			len, owner, offset, flags);
+
+	error = xfs_rmap_lookup_eq(rcur, agbno, len, owner, offset, flags, &i);
+	if (error)
+		goto done;
+	XFS_WANT_CORRUPTED_GOTO(rcur->bc_mp, i == 1, done);
+
+	error = xfs_btree_delete(rcur, &i);
+	if (error)
+		goto done;
+	XFS_WANT_CORRUPTED_GOTO(rcur->bc_mp, i == 1, done);
+done:
+	if (error)
+		trace_xfs_rmap_delete_error(rcur->bc_mp,
+				rcur->bc_private.a.agno, error, _RET_IP_);
+	return error;
+}
+
 static int
 xfs_rmap_btrec_to_irec(
 	union xfs_btree_rec	*rec,
@@ -178,6 +209,160 @@ xfs_rmap_get_rec(
 	return xfs_rmap_btrec_to_irec(rec, irec);
 }
 
+struct xfs_find_left_neighbor_info {
+	struct xfs_rmap_irec	high;
+	struct xfs_rmap_irec	*irec;
+	int			*stat;
+};
+
+/* For each rmap given, figure out if it matches the key we want. */
+STATIC int
+xfs_rmap_find_left_neighbor_helper(
+	struct xfs_btree_cur	*cur,
+	struct xfs_rmap_irec	*rec,
+	void			*priv)
+{
+	struct xfs_find_left_neighbor_info	*info = priv;
+
+	trace_xfs_rmap_find_left_neighbor_candidate(cur->bc_mp,
+			cur->bc_private.a.agno, rec->rm_startblock,
+			rec->rm_blockcount, rec->rm_owner, rec->rm_offset,
+			rec->rm_flags);
+
+	if (rec->rm_owner != info->high.rm_owner)
+		return XFS_BTREE_QUERY_RANGE_CONTINUE;
+	if (!XFS_RMAP_NON_INODE_OWNER(rec->rm_owner) &&
+	    !(rec->rm_flags & XFS_RMAP_BMBT_BLOCK) &&
+	    rec->rm_offset + rec->rm_blockcount - 1 != info->high.rm_offset)
+		return XFS_BTREE_QUERY_RANGE_CONTINUE;
+
+	*info->irec = *rec;
+	*info->stat = 1;
+	return XFS_BTREE_QUERY_RANGE_ABORT;
+}
+
+/*
+ * Find the record to the left of the given extent, being careful only to
+ * return a match with the same owner and adjacent physical and logical
+ * block ranges.
+ */
+int
+xfs_rmap_find_left_neighbor(
+	struct xfs_btree_cur	*cur,
+	xfs_agblock_t		bno,
+	uint64_t		owner,
+	uint64_t		offset,
+	unsigned int		flags,
+	struct xfs_rmap_irec	*irec,
+	int			*stat)
+{
+	struct xfs_find_left_neighbor_info	info;
+	int			error;
+
+	*stat = 0;
+	if (bno == 0)
+		return 0;
+	info.high.rm_startblock = bno - 1;
+	info.high.rm_owner = owner;
+	if (!XFS_RMAP_NON_INODE_OWNER(owner) &&
+	    !(flags & XFS_RMAP_BMBT_BLOCK)) {
+		if (offset == 0)
+			return 0;
+		info.high.rm_offset = offset - 1;
+	} else
+		info.high.rm_offset = 0;
+	info.high.rm_flags = flags;
+	info.high.rm_blockcount = 0;
+	info.irec = irec;
+	info.stat = stat;
+
+	trace_xfs_rmap_find_left_neighbor_query(cur->bc_mp,
+			cur->bc_private.a.agno, bno, 0, owner, offset, flags);
+
+	error = xfs_rmap_query_range(cur, &info.high, &info.high,
+			xfs_rmap_find_left_neighbor_helper, &info);
+	if (error == XFS_BTREE_QUERY_RANGE_ABORT)
+		error = 0;
+	if (*stat)
+		trace_xfs_rmap_find_left_neighbor_result(cur->bc_mp,
+				cur->bc_private.a.agno, irec->rm_startblock,
+				irec->rm_blockcount, irec->rm_owner,
+				irec->rm_offset, irec->rm_flags);
+	return error;
+}
+
+/* For each rmap given, figure out if it matches the key we want. */
+STATIC int
+xfs_rmap_lookup_le_range_helper(
+	struct xfs_btree_cur	*cur,
+	struct xfs_rmap_irec	*rec,
+	void			*priv)
+{
+	struct xfs_find_left_neighbor_info	*info = priv;
+
+	trace_xfs_rmap_lookup_le_range_candidate(cur->bc_mp,
+			cur->bc_private.a.agno, rec->rm_startblock,
+			rec->rm_blockcount, rec->rm_owner, rec->rm_offset,
+			rec->rm_flags);
+
+	if (rec->rm_owner != info->high.rm_owner)
+		return XFS_BTREE_QUERY_RANGE_CONTINUE;
+	if (!XFS_RMAP_NON_INODE_OWNER(rec->rm_owner) &&
+	    !(rec->rm_flags & XFS_RMAP_BMBT_BLOCK) &&
+	    (rec->rm_offset > info->high.rm_offset ||
+	     rec->rm_offset + rec->rm_blockcount <= info->high.rm_offset))
+		return XFS_BTREE_QUERY_RANGE_CONTINUE;
+
+	*info->irec = *rec;
+	*info->stat = 1;
+	return XFS_BTREE_QUERY_RANGE_ABORT;
+}
+
+/*
+ * Find the record to the left of the given extent, being careful only to
+ * return a match with the same owner and overlapping physical and logical
+ * block ranges.  This is the overlapping-interval version of
+ * xfs_rmap_lookup_le.
+ */
+int
+xfs_rmap_lookup_le_range(
+	struct xfs_btree_cur	*cur,
+	xfs_agblock_t		bno,
+	uint64_t		owner,
+	uint64_t		offset,
+	unsigned int		flags,
+	struct xfs_rmap_irec	*irec,
+	int			*stat)
+{
+	struct xfs_find_left_neighbor_info	info;
+	int			error;
+
+	info.high.rm_startblock = bno;
+	info.high.rm_owner = owner;
+	if (!XFS_RMAP_NON_INODE_OWNER(owner) && !(flags & XFS_RMAP_BMBT_BLOCK))
+		info.high.rm_offset = offset;
+	else
+		info.high.rm_offset = 0;
+	info.high.rm_flags = flags;
+	info.high.rm_blockcount = 0;
+	*stat = 0;
+	info.irec = irec;
+	info.stat = stat;
+
+	trace_xfs_rmap_lookup_le_range(cur->bc_mp,
+			cur->bc_private.a.agno, bno, 0, owner, offset, flags);
+	error = xfs_rmap_query_range(cur, &info.high, &info.high,
+			xfs_rmap_lookup_le_range_helper, &info);
+	if (error == XFS_BTREE_QUERY_RANGE_ABORT)
+		error = 0;
+	if (*stat)
+		trace_xfs_rmap_lookup_le_range_result(cur->bc_mp,
+				cur->bc_private.a.agno, irec->rm_startblock,
+				irec->rm_blockcount, irec->rm_owner,
+				irec->rm_offset, irec->rm_flags);
+	return error;
+}
+
 /*
  * Find the extent in the rmap btree and remove it.
  *
@@ -1096,6 +1281,321 @@ done:
 #undef	RIGHT
 #undef	PREV
 
+/*
+ * Find an extent in the rmap btree and unmap it.  For rmap extent types that
+ * can overlap (data fork rmaps on reflink filesystems) we must be careful
+ * that the prev/next records in the btree might belong to another owner.
+ * Therefore we must use delete+insert to alter any of the key fields.
+ *
+ * For every other situation there can only be one owner for a given extent,
+ * so we can call the regular _free function.
+ */
+STATIC int
+xfs_rmap_unmap_shared(
+	struct xfs_btree_cur	*cur,
+	xfs_agblock_t		bno,
+	xfs_extlen_t		len,
+	bool			unwritten,
+	struct xfs_owner_info	*oinfo)
+{
+	struct xfs_mount	*mp = cur->bc_mp;
+	struct xfs_rmap_irec	ltrec;
+	uint64_t		ltoff;
+	int			error = 0;
+	int			i;
+	uint64_t		owner;
+	uint64_t		offset;
+	unsigned int		flags;
+
+	xfs_owner_info_unpack(oinfo, &owner, &offset, &flags);
+	if (unwritten)
+		flags |= XFS_RMAP_UNWRITTEN;
+	trace_xfs_rmap_unmap(mp, cur->bc_private.a.agno, bno, len,
+			unwritten, oinfo);
+
+	/*
+	 * We should always have a left record because there's a static record
+	 * for the AG headers at rm_startblock == 0 created by mkfs/growfs that
+	 * will not ever be removed from the tree.
+	 */
+	error = xfs_rmap_lookup_le_range(cur, bno, owner, offset, flags,
+			&ltrec, &i);
+	if (error)
+		goto out_error;
+	XFS_WANT_CORRUPTED_GOTO(mp, i == 1, out_error);
+	ltoff = ltrec.rm_offset;
+
+	/* Make sure the extent we found covers the entire freeing range. */
+	XFS_WANT_CORRUPTED_GOTO(mp, ltrec.rm_startblock <= bno &&
+		ltrec.rm_startblock + ltrec.rm_blockcount >=
+		bno + len, out_error);
+
+	/* Make sure the owner matches what we expect to find in the tree. */
+	XFS_WANT_CORRUPTED_GOTO(mp, owner == ltrec.rm_owner, out_error);
+
+	/* Make sure the unwritten flag matches. */
+	XFS_WANT_CORRUPTED_GOTO(mp, (flags & XFS_RMAP_UNWRITTEN) ==
+			(ltrec.rm_flags & XFS_RMAP_UNWRITTEN), out_error);
+
+	/* Check the offset. */
+	XFS_WANT_CORRUPTED_GOTO(mp, ltrec.rm_offset <= offset, out_error);
+	XFS_WANT_CORRUPTED_GOTO(mp, offset <= ltoff + ltrec.rm_blockcount,
+			out_error);
+
+	if (ltrec.rm_startblock == bno && ltrec.rm_blockcount == len) {
+		/* Exact match, simply remove the record from rmap tree. */
+		error = xfs_rmap_delete(cur, ltrec.rm_startblock,
+				ltrec.rm_blockcount, ltrec.rm_owner,
+				ltrec.rm_offset, ltrec.rm_flags);
+		if (error)
+			goto out_error;
+	} else if (ltrec.rm_startblock == bno) {
+		/*
+		 * Overlap left hand side of extent: move the start, trim the
+		 * length and update the current record.
+		 *
+		 *       ltbno                ltlen
+		 * Orig:    |oooooooooooooooooooo|
+		 * Freeing: |fffffffff|
+		 * Result:            |rrrrrrrrrr|
+		 *         bno       len
+		 */
+
+		/* Delete prev rmap. */
+		error = xfs_rmap_delete(cur, ltrec.rm_startblock,
+				ltrec.rm_blockcount, ltrec.rm_owner,
+				ltrec.rm_offset, ltrec.rm_flags);
+		if (error)
+			goto out_error;
+
+		/* Add an rmap at the new offset. */
+		ltrec.rm_startblock += len;
+		ltrec.rm_blockcount -= len;
+		ltrec.rm_offset += len;
+		error = xfs_rmap_insert(cur, ltrec.rm_startblock,
+				ltrec.rm_blockcount, ltrec.rm_owner,
+				ltrec.rm_offset, ltrec.rm_flags);
+		if (error)
+			goto out_error;
+	} else if (ltrec.rm_startblock + ltrec.rm_blockcount == bno + len) {
+		/*
+		 * Overlap right hand side of extent: trim the length and
+		 * update the current record.
+		 *
+		 *       ltbno                ltlen
+		 * Orig:    |oooooooooooooooooooo|
+		 * Freeing:            |fffffffff|
+		 * Result:  |rrrrrrrrrr|
+		 *                    bno       len
+		 */
+		error = xfs_rmap_lookup_eq(cur, ltrec.rm_startblock,
+				ltrec.rm_blockcount, ltrec.rm_owner,
+				ltrec.rm_offset, ltrec.rm_flags, &i);
+		if (error)
+			goto out_error;
+		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, out_error);
+		ltrec.rm_blockcount -= len;
+		error = xfs_rmap_update(cur, &ltrec);
+		if (error)
+			goto out_error;
+	} else {
+		/*
+		 * Overlap middle of extent: trim the length of the existing
+		 * record to the length of the new left-extent size, increment
+		 * the insertion position so we can insert a new record
+		 * containing the remaining right-extent space.
+		 *
+		 *       ltbno                ltlen
+		 * Orig:    |oooooooooooooooooooo|
+		 * Freeing:       |fffffffff|
+		 * Result:  |rrrrr|         |rrrr|
+		 *               bno       len
+		 */
+		xfs_extlen_t	orig_len = ltrec.rm_blockcount;
+
+		/* Shrink the left side of the rmap */
+		error = xfs_rmap_lookup_eq(cur, ltrec.rm_startblock,
+				ltrec.rm_blockcount, ltrec.rm_owner,
+				ltrec.rm_offset, ltrec.rm_flags, &i);
+		if (error)
+			goto out_error;
+		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, out_error);
+		ltrec.rm_blockcount = bno - ltrec.rm_startblock;
+		error = xfs_rmap_update(cur, &ltrec);
+		if (error)
+			goto out_error;
+
+		/* Add an rmap at the new offset */
+		error = xfs_rmap_insert(cur, bno + len,
+				orig_len - len - ltrec.rm_blockcount,
+				ltrec.rm_owner, offset + len,
+				ltrec.rm_flags);
+		if (error)
+			goto out_error;
+	}
+
+	trace_xfs_rmap_unmap_done(mp, cur->bc_private.a.agno, bno, len,
+			unwritten, oinfo);
+out_error:
+	if (error)
+		trace_xfs_rmap_unmap_error(cur->bc_mp,
+				cur->bc_private.a.agno, error, _RET_IP_);
+	return error;
+}
+
+/*
+ * Find an extent in the rmap btree and map it.  For rmap extent types that
+ * can overlap (data fork rmaps on reflink filesystems) we must be careful
+ * that the prev/next records in the btree might belong to another owner.
+ * Therefore we must use delete+insert to alter any of the key fields.
+ *
+ * For every other situation there can only be one owner for a given extent,
+ * so we can call the regular _alloc function.
+ */
+STATIC int
+xfs_rmap_map_shared(
+	struct xfs_btree_cur	*cur,
+	xfs_agblock_t		bno,
+	xfs_extlen_t		len,
+	bool			unwritten,
+	struct xfs_owner_info	*oinfo)
+{
+	struct xfs_mount	*mp = cur->bc_mp;
+	struct xfs_rmap_irec	ltrec;
+	struct xfs_rmap_irec	gtrec;
+	int			have_gt;
+	int			have_lt;
+	int			error = 0;
+	int			i;
+	uint64_t		owner;
+	uint64_t		offset;
+	unsigned int		flags = 0;
+
+	xfs_owner_info_unpack(oinfo, &owner, &offset, &flags);
+	if (unwritten)
+		flags |= XFS_RMAP_UNWRITTEN;
+	trace_xfs_rmap_map(mp, cur->bc_private.a.agno, bno, len,
+			unwritten, oinfo);
+
+	/* Is there a left record that abuts our range? */
+	error = xfs_rmap_find_left_neighbor(cur, bno, owner, offset, flags,
+			&ltrec, &have_lt);
+	if (error)
+		goto out_error;
+	if (have_lt &&
+	    !xfs_rmap_is_mergeable(&ltrec, owner, flags))
+		have_lt = 0;
+
+	/* Is there a right record that abuts our range? */
+	error = xfs_rmap_lookup_eq(cur, bno + len, len, owner, offset + len,
+			flags, &have_gt);
+	if (error)
+		goto out_error;
+	if (have_gt) {
+		error = xfs_rmap_get_rec(cur, &gtrec, &have_gt);
+		if (error)
+			goto out_error;
+		XFS_WANT_CORRUPTED_GOTO(mp, have_gt == 1, out_error);
+		trace_xfs_rmap_find_right_neighbor_result(cur->bc_mp,
+			cur->bc_private.a.agno, gtrec.rm_startblock,
+			gtrec.rm_blockcount, gtrec.rm_owner,
+			gtrec.rm_offset, gtrec.rm_flags);
+
+		if (!xfs_rmap_is_mergeable(&gtrec, owner, flags))
+			have_gt = 0;
+	}
+
+	if (have_lt &&
+	    ltrec.rm_startblock + ltrec.rm_blockcount == bno &&
+	    ltrec.rm_offset + ltrec.rm_blockcount == offset) {
+		/*
+		 * Left edge contiguous, merge into left record.
+		 *
+		 *       ltbno     ltlen
+		 * orig:   |ooooooooo|
+		 * adding:           |aaaaaaaaa|
+		 * result: |rrrrrrrrrrrrrrrrrrr|
+		 *                  bno       len
+		 */
+		ltrec.rm_blockcount += len;
+		if (have_gt &&
+		    bno + len == gtrec.rm_startblock &&
+		    offset + len == gtrec.rm_offset) {
+			/*
+			 * Right edge also contiguous, delete right record
+			 * and merge into left record.
+			 *
+			 *       ltbno     ltlen    gtbno     gtlen
+			 * orig:   |ooooooooo|         |ooooooooo|
+			 * adding:           |aaaaaaaaa|
+			 * result: |rrrrrrrrrrrrrrrrrrrrrrrrrrrrr|
+			 */
+			ltrec.rm_blockcount += gtrec.rm_blockcount;
+			error = xfs_rmap_delete(cur, gtrec.rm_startblock,
+					gtrec.rm_blockcount, gtrec.rm_owner,
+					gtrec.rm_offset, gtrec.rm_flags);
+			if (error)
+				goto out_error;
+		}
+
+		/* Point the cursor back to the left record and update. */
+		error = xfs_rmap_lookup_eq(cur, ltrec.rm_startblock,
+				ltrec.rm_blockcount, ltrec.rm_owner,
+				ltrec.rm_offset, ltrec.rm_flags, &i);
+		if (error)
+			goto out_error;
+		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, out_error);
+
+		error = xfs_rmap_update(cur, &ltrec);
+		if (error)
+			goto out_error;
+	} else if (have_gt &&
+		   bno + len == gtrec.rm_startblock &&
+		   offset + len == gtrec.rm_offset) {
+		/*
+		 * Right edge contiguous, merge into right record.
+		 *
+		 *                 gtbno     gtlen
+		 * Orig:             |ooooooooo|
+		 * adding: |aaaaaaaaa|
+		 * Result: |rrrrrrrrrrrrrrrrrrr|
+		 *        bno       len
+		 */
+		/* Delete the old record. */
+		error = xfs_rmap_delete(cur, gtrec.rm_startblock,
+				gtrec.rm_blockcount, gtrec.rm_owner,
+				gtrec.rm_offset, gtrec.rm_flags);
+		if (error)
+			goto out_error;
+
+		/* Move the start and re-add it. */
+		gtrec.rm_startblock = bno;
+		gtrec.rm_blockcount += len;
+		gtrec.rm_offset = offset;
+		error = xfs_rmap_insert(cur, gtrec.rm_startblock,
+				gtrec.rm_blockcount, gtrec.rm_owner,
+				gtrec.rm_offset, gtrec.rm_flags);
+		if (error)
+			goto out_error;
+	} else {
+		/*
+		 * No contiguous edge with identical owner, insert
+		 * new record at current cursor position.
+		 */
+		error = xfs_rmap_insert(cur, bno, len, owner, offset, flags);
+		if (error)
+			goto out_error;
+	}
+
+	trace_xfs_rmap_map_done(mp, cur->bc_private.a.agno, bno, len,
+			unwritten, oinfo);
+out_error:
+	if (error)
+		trace_xfs_rmap_map_error(cur->bc_mp,
+				cur->bc_private.a.agno, error, _RET_IP_);
+	return error;
+}
+
 struct xfs_rmap_query_range_info {
 	xfs_rmap_query_range_fn	fn;
 	void				*priv;
@@ -1235,11 +1735,19 @@ xfs_rmap_finish_one(
 	case XFS_RMAP_MAP:
 		error = xfs_rmap_map(rcur, bno, blockcount, unwritten, &oinfo);
 		break;
+	case XFS_RMAP_MAP_SHARED:
+		error = xfs_rmap_map_shared(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_UNMAP_SHARED:
+		error = xfs_rmap_unmap_shared(rcur, bno, blockcount, unwritten,
+				&oinfo);
+		break;
 	case XFS_RMAP_CONVERT:
 		error = xfs_rmap_convert(rcur, bno, blockcount, !unwritten,
 				&oinfo);
@@ -1313,7 +1821,8 @@ xfs_rmap_map_extent(
 	if (!xfs_rmap_update_is_needed(mp, whichfork))
 		return 0;
 
-	return __xfs_rmap_add(mp, dfops, XFS_RMAP_MAP, ip->i_ino,
+	return __xfs_rmap_add(mp, dfops, xfs_is_reflink_inode(ip) ?
+			XFS_RMAP_MAP_SHARED : XFS_RMAP_MAP, ip->i_ino,
 			whichfork, PREV);
 }
 
@@ -1329,7 +1838,8 @@ xfs_rmap_unmap_extent(
 	if (!xfs_rmap_update_is_needed(mp, whichfork))
 		return 0;
 
-	return __xfs_rmap_add(mp, dfops, XFS_RMAP_UNMAP, ip->i_ino,
+	return __xfs_rmap_add(mp, dfops, xfs_is_reflink_inode(ip) ?
+			XFS_RMAP_UNMAP_SHARED : XFS_RMAP_UNMAP, ip->i_ino,
 			whichfork, PREV);
 }
 
diff --git a/libxfs/xfs_rmap.h b/libxfs/xfs_rmap.h
index 71cf99a..7899305 100644
--- a/libxfs/xfs_rmap.h
+++ b/libxfs/xfs_rmap.h
@@ -206,4 +206,11 @@ int xfs_rmap_finish_one(struct xfs_trans *tp, enum xfs_rmap_intent_type type,
 		xfs_fsblock_t startblock, xfs_filblks_t blockcount,
 		xfs_exntst_t state, struct xfs_btree_cur **pcur);
 
+int xfs_rmap_find_left_neighbor(struct xfs_btree_cur *cur, xfs_agblock_t bno,
+		uint64_t owner, uint64_t offset, unsigned int flags,
+		struct xfs_rmap_irec *irec, int	*stat);
+int xfs_rmap_lookup_le_range(struct xfs_btree_cur *cur, xfs_agblock_t bno,
+		uint64_t owner, uint64_t offset, unsigned int flags,
+		struct xfs_rmap_irec *irec, int	*stat);
+
 #endif	/* __XFS_RMAP_H__ */

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 40/71] xfs: convert unwritten status of shared-extent reverse mappings on shared files

[Darrick J. Wong]

Upgrade the rmap extent conversion function to handle shared extents.

v2: Move unwritten bit to rm_offset.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_rmap.c |  385 +++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 384 insertions(+), 1 deletion(-)


diff --git a/libxfs/xfs_rmap.c b/libxfs/xfs_rmap.c
index 17dc7d6..7a75e26 100644
--- a/libxfs/xfs_rmap.c
+++ b/libxfs/xfs_rmap.c
@@ -1276,6 +1276,384 @@ done:
 	return error;
 }
 
+/*
+ * Convert an unwritten extent to a real extent or vice versa.  If there is no
+ * possibility of overlapping extents, delegate to the simpler convert
+ * function.
+ */
+STATIC int
+xfs_rmap_convert_shared(
+	struct xfs_btree_cur	*cur,
+	xfs_agblock_t		bno,
+	xfs_extlen_t		len,
+	bool			unwritten,
+	struct xfs_owner_info	*oinfo)
+{
+	struct xfs_mount	*mp = cur->bc_mp;
+	struct xfs_rmap_irec	r[4];	/* neighbor extent entries */
+					/* left is 0, right is 1, prev is 2 */
+					/* new is 3 */
+	uint64_t		owner;
+	uint64_t		offset;
+	uint64_t		new_endoff;
+	unsigned int		oldext;
+	unsigned int		newext;
+	unsigned int		flags = 0;
+	int			i;
+	int			state = 0;
+	int			error;
+
+	xfs_owner_info_unpack(oinfo, &owner, &offset, &flags);
+	ASSERT(!(XFS_RMAP_NON_INODE_OWNER(owner) ||
+			(flags & (XFS_RMAP_ATTR_FORK | XFS_RMAP_BMBT_BLOCK))));
+	oldext = unwritten ? XFS_RMAP_UNWRITTEN : 0;
+	new_endoff = offset + len;
+	trace_xfs_rmap_convert(mp, cur->bc_private.a.agno, bno, len,
+			unwritten, oinfo);
+
+	/*
+	 * For the initial lookup, look for and exact match or the left-adjacent
+	 * record for our insertion point. This will also give us the record for
+	 * start block contiguity tests.
+	 */
+	error = xfs_rmap_lookup_le_range(cur, bno, owner, offset, flags,
+			&PREV, &i);
+	XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done);
+
+	ASSERT(PREV.rm_offset <= offset);
+	ASSERT(PREV.rm_offset + PREV.rm_blockcount >= new_endoff);
+	ASSERT((PREV.rm_flags & XFS_RMAP_UNWRITTEN) == oldext);
+	newext = ~oldext & XFS_RMAP_UNWRITTEN;
+
+	/*
+	 * Set flags determining what part of the previous oldext allocation
+	 * extent is being replaced by a newext allocation.
+	 */
+	if (PREV.rm_offset == offset)
+		state |= RMAP_LEFT_FILLING;
+	if (PREV.rm_offset + PREV.rm_blockcount == new_endoff)
+		state |= RMAP_RIGHT_FILLING;
+
+	/* Is there a left record that abuts our range? */
+	error = xfs_rmap_find_left_neighbor(cur, bno, owner, offset, newext,
+			&LEFT, &i);
+	if (error)
+		goto done;
+	if (i) {
+		state |= RMAP_LEFT_VALID;
+		XFS_WANT_CORRUPTED_GOTO(mp,
+				LEFT.rm_startblock + LEFT.rm_blockcount <= bno,
+				done);
+		if (xfs_rmap_is_mergeable(&LEFT, owner, newext))
+			state |= RMAP_LEFT_CONTIG;
+	}
+
+	/* Is there a right record that abuts our range? */
+	error = xfs_rmap_lookup_eq(cur, bno + len, len, owner, offset + len,
+			newext, &i);
+	if (error)
+		goto done;
+	if (i) {
+		state |= RMAP_RIGHT_VALID;
+		error = xfs_rmap_get_rec(cur, &RIGHT, &i);
+		if (error)
+			goto done;
+		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done);
+		XFS_WANT_CORRUPTED_GOTO(mp, bno + len <= RIGHT.rm_startblock,
+				done);
+		trace_xfs_rmap_find_right_neighbor_result(cur->bc_mp,
+				cur->bc_private.a.agno, RIGHT.rm_startblock,
+				RIGHT.rm_blockcount, RIGHT.rm_owner,
+				RIGHT.rm_offset, RIGHT.rm_flags);
+		if (xfs_rmap_is_mergeable(&RIGHT, owner, newext))
+			state |= RMAP_RIGHT_CONTIG;
+	}
+
+	/* check that left + prev + right is not too long */
+	if ((state & (RMAP_LEFT_FILLING | RMAP_LEFT_CONTIG |
+			 RMAP_RIGHT_FILLING | RMAP_RIGHT_CONTIG)) ==
+	    (RMAP_LEFT_FILLING | RMAP_LEFT_CONTIG |
+	     RMAP_RIGHT_FILLING | RMAP_RIGHT_CONTIG) &&
+	    (unsigned long)LEFT.rm_blockcount + len +
+	     RIGHT.rm_blockcount > XFS_RMAP_LEN_MAX)
+		state &= ~RMAP_RIGHT_CONTIG;
+
+	trace_xfs_rmap_convert_state(mp, cur->bc_private.a.agno, state,
+			_RET_IP_);
+	/*
+	 * Switch out based on the FILLING and CONTIG state bits.
+	 */
+	switch (state & (RMAP_LEFT_FILLING | RMAP_LEFT_CONTIG |
+			 RMAP_RIGHT_FILLING | RMAP_RIGHT_CONTIG)) {
+	case RMAP_LEFT_FILLING | RMAP_LEFT_CONTIG |
+	     RMAP_RIGHT_FILLING | RMAP_RIGHT_CONTIG:
+		/*
+		 * Setting all of a previous oldext extent to newext.
+		 * The left and right neighbors are both contiguous with new.
+		 */
+		error = xfs_rmap_delete(cur, RIGHT.rm_startblock,
+				RIGHT.rm_blockcount, RIGHT.rm_owner,
+				RIGHT.rm_offset, RIGHT.rm_flags);
+		if (error)
+			goto done;
+		error = xfs_rmap_delete(cur, PREV.rm_startblock,
+				PREV.rm_blockcount, PREV.rm_owner,
+				PREV.rm_offset, PREV.rm_flags);
+		if (error)
+			goto done;
+		NEW = LEFT;
+		error = xfs_rmap_lookup_eq(cur, NEW.rm_startblock,
+				NEW.rm_blockcount, NEW.rm_owner,
+				NEW.rm_offset, NEW.rm_flags, &i);
+		if (error)
+			goto done;
+		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done);
+		NEW.rm_blockcount += PREV.rm_blockcount + RIGHT.rm_blockcount;
+		error = xfs_rmap_update(cur, &NEW);
+		if (error)
+			goto done;
+		break;
+
+	case RMAP_LEFT_FILLING | RMAP_RIGHT_FILLING | RMAP_LEFT_CONTIG:
+		/*
+		 * Setting all of a previous oldext extent to newext.
+		 * The left neighbor is contiguous, the right is not.
+		 */
+		error = xfs_rmap_delete(cur, PREV.rm_startblock,
+				PREV.rm_blockcount, PREV.rm_owner,
+				PREV.rm_offset, PREV.rm_flags);
+		if (error)
+			goto done;
+		NEW = LEFT;
+		error = xfs_rmap_lookup_eq(cur, NEW.rm_startblock,
+				NEW.rm_blockcount, NEW.rm_owner,
+				NEW.rm_offset, NEW.rm_flags, &i);
+		if (error)
+			goto done;
+		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done);
+		NEW.rm_blockcount += PREV.rm_blockcount;
+		error = xfs_rmap_update(cur, &NEW);
+		if (error)
+			goto done;
+		break;
+
+	case RMAP_LEFT_FILLING | RMAP_RIGHT_FILLING | RMAP_RIGHT_CONTIG:
+		/*
+		 * Setting all of a previous oldext extent to newext.
+		 * The right neighbor is contiguous, the left is not.
+		 */
+		error = xfs_rmap_delete(cur, RIGHT.rm_startblock,
+				RIGHT.rm_blockcount, RIGHT.rm_owner,
+				RIGHT.rm_offset, RIGHT.rm_flags);
+		if (error)
+			goto done;
+		NEW = PREV;
+		error = xfs_rmap_lookup_eq(cur, NEW.rm_startblock,
+				NEW.rm_blockcount, NEW.rm_owner,
+				NEW.rm_offset, NEW.rm_flags, &i);
+		if (error)
+			goto done;
+		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done);
+		NEW.rm_blockcount += RIGHT.rm_blockcount;
+		NEW.rm_flags = RIGHT.rm_flags;
+		error = xfs_rmap_update(cur, &NEW);
+		if (error)
+			goto done;
+		break;
+
+	case RMAP_LEFT_FILLING | RMAP_RIGHT_FILLING:
+		/*
+		 * Setting all of a previous oldext extent to newext.
+		 * Neither the left nor right neighbors are contiguous with
+		 * the new one.
+		 */
+		NEW = PREV;
+		error = xfs_rmap_lookup_eq(cur, NEW.rm_startblock,
+				NEW.rm_blockcount, NEW.rm_owner,
+				NEW.rm_offset, NEW.rm_flags, &i);
+		if (error)
+			goto done;
+		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done);
+		NEW.rm_flags = newext;
+		error = xfs_rmap_update(cur, &NEW);
+		if (error)
+			goto done;
+		break;
+
+	case RMAP_LEFT_FILLING | RMAP_LEFT_CONTIG:
+		/*
+		 * Setting the first part of a previous oldext extent to newext.
+		 * The left neighbor is contiguous.
+		 */
+		NEW = PREV;
+		error = xfs_rmap_delete(cur, NEW.rm_startblock,
+				NEW.rm_blockcount, NEW.rm_owner,
+				NEW.rm_offset, NEW.rm_flags);
+		if (error)
+			goto done;
+		NEW.rm_offset += len;
+		NEW.rm_startblock += len;
+		NEW.rm_blockcount -= len;
+		error = xfs_rmap_insert(cur, NEW.rm_startblock,
+				NEW.rm_blockcount, NEW.rm_owner,
+				NEW.rm_offset, NEW.rm_flags);
+		if (error)
+			goto done;
+		NEW = LEFT;
+		error = xfs_rmap_lookup_eq(cur, NEW.rm_startblock,
+				NEW.rm_blockcount, NEW.rm_owner,
+				NEW.rm_offset, NEW.rm_flags, &i);
+		if (error)
+			goto done;
+		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done);
+		NEW.rm_blockcount += len;
+		error = xfs_rmap_update(cur, &NEW);
+		if (error)
+			goto done;
+		break;
+
+	case RMAP_LEFT_FILLING:
+		/*
+		 * Setting the first part of a previous oldext extent to newext.
+		 * The left neighbor is not contiguous.
+		 */
+		NEW = PREV;
+		error = xfs_rmap_delete(cur, NEW.rm_startblock,
+				NEW.rm_blockcount, NEW.rm_owner,
+				NEW.rm_offset, NEW.rm_flags);
+		if (error)
+			goto done;
+		NEW.rm_offset += len;
+		NEW.rm_startblock += len;
+		NEW.rm_blockcount -= len;
+		error = xfs_rmap_insert(cur, NEW.rm_startblock,
+				NEW.rm_blockcount, NEW.rm_owner,
+				NEW.rm_offset, NEW.rm_flags);
+		if (error)
+			goto done;
+		error = xfs_rmap_insert(cur, bno, len, owner, offset, newext);
+		if (error)
+			goto done;
+		break;
+
+	case RMAP_RIGHT_FILLING | RMAP_RIGHT_CONTIG:
+		/*
+		 * Setting the last part of a previous oldext extent to newext.
+		 * The right neighbor is contiguous with the new allocation.
+		 */
+		NEW = PREV;
+		error = xfs_rmap_lookup_eq(cur, NEW.rm_startblock,
+				NEW.rm_blockcount, NEW.rm_owner,
+				NEW.rm_offset, NEW.rm_flags, &i);
+		if (error)
+			goto done;
+		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done);
+		NEW.rm_blockcount = offset - NEW.rm_offset;
+		error = xfs_rmap_update(cur, &NEW);
+		if (error)
+			goto done;
+		NEW = RIGHT;
+		error = xfs_rmap_delete(cur, NEW.rm_startblock,
+				NEW.rm_blockcount, NEW.rm_owner,
+				NEW.rm_offset, NEW.rm_flags);
+		if (error)
+			goto done;
+		NEW.rm_offset = offset;
+		NEW.rm_startblock = bno;
+		NEW.rm_blockcount += len;
+		error = xfs_rmap_insert(cur, NEW.rm_startblock,
+				NEW.rm_blockcount, NEW.rm_owner,
+				NEW.rm_offset, NEW.rm_flags);
+		if (error)
+			goto done;
+		break;
+
+	case RMAP_RIGHT_FILLING:
+		/*
+		 * Setting the last part of a previous oldext extent to newext.
+		 * The right neighbor is not contiguous.
+		 */
+		NEW = PREV;
+		error = xfs_rmap_lookup_eq(cur, NEW.rm_startblock,
+				NEW.rm_blockcount, NEW.rm_owner,
+				NEW.rm_offset, NEW.rm_flags, &i);
+		if (error)
+			goto done;
+		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done);
+		NEW.rm_blockcount -= len;
+		error = xfs_rmap_update(cur, &NEW);
+		if (error)
+			goto done;
+		error = xfs_rmap_insert(cur, bno, len, owner, offset, newext);
+		if (error)
+			goto done;
+		break;
+
+	case 0:
+		/*
+		 * Setting the middle part of a previous oldext extent to
+		 * newext.  Contiguity is impossible here.
+		 * One extent becomes three extents.
+		 */
+		/* new right extent - oldext */
+		NEW.rm_startblock = bno + len;
+		NEW.rm_owner = owner;
+		NEW.rm_offset = new_endoff;
+		NEW.rm_blockcount = PREV.rm_offset + PREV.rm_blockcount -
+				new_endoff;
+		NEW.rm_flags = PREV.rm_flags;
+		error = xfs_rmap_insert(cur, NEW.rm_startblock,
+				NEW.rm_blockcount, NEW.rm_owner, NEW.rm_offset,
+				NEW.rm_flags);
+		if (error)
+			goto done;
+		/* new left extent - oldext */
+		NEW = PREV;
+		error = xfs_rmap_lookup_eq(cur, NEW.rm_startblock,
+				NEW.rm_blockcount, NEW.rm_owner,
+				NEW.rm_offset, NEW.rm_flags, &i);
+		if (error)
+			goto done;
+		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done);
+		NEW.rm_blockcount = offset - NEW.rm_offset;
+		error = xfs_rmap_update(cur, &NEW);
+		if (error)
+			goto done;
+		/* new middle extent - newext */
+		NEW.rm_startblock = bno;
+		NEW.rm_blockcount = len;
+		NEW.rm_owner = owner;
+		NEW.rm_offset = offset;
+		NEW.rm_flags = newext;
+		error = xfs_rmap_insert(cur, NEW.rm_startblock,
+				NEW.rm_blockcount, NEW.rm_owner, NEW.rm_offset,
+				NEW.rm_flags);
+		if (error)
+			goto done;
+		break;
+
+	case RMAP_LEFT_FILLING | RMAP_LEFT_CONTIG | RMAP_RIGHT_CONTIG:
+	case RMAP_RIGHT_FILLING | RMAP_LEFT_CONTIG | RMAP_RIGHT_CONTIG:
+	case RMAP_LEFT_FILLING | RMAP_RIGHT_CONTIG:
+	case RMAP_RIGHT_FILLING | RMAP_LEFT_CONTIG:
+	case RMAP_LEFT_CONTIG | RMAP_RIGHT_CONTIG:
+	case RMAP_LEFT_CONTIG:
+	case RMAP_RIGHT_CONTIG:
+		/*
+		 * These cases are all impossible.
+		 */
+		ASSERT(0);
+	}
+
+	trace_xfs_rmap_convert_done(mp, cur->bc_private.a.agno, bno, len,
+			unwritten, oinfo);
+done:
+	if (error)
+		trace_xfs_rmap_convert_error(cur->bc_mp,
+				cur->bc_private.a.agno, error, _RET_IP_);
+	return error;
+}
+
 #undef	NEW
 #undef	LEFT
 #undef	RIGHT
@@ -1752,6 +2130,10 @@ xfs_rmap_finish_one(
 		error = xfs_rmap_convert(rcur, bno, blockcount, !unwritten,
 				&oinfo);
 		break;
+	case XFS_RMAP_CONVERT_SHARED:
+		error = xfs_rmap_convert_shared(rcur, bno, blockcount,
+				!unwritten, &oinfo);
+		break;
 	default:
 		ASSERT(0);
 		error = -EFSCORRUPTED;
@@ -1855,7 +2237,8 @@ xfs_rmap_convert_extent(
 	if (!xfs_rmap_update_is_needed(mp, whichfork))
 		return 0;
 
-	return __xfs_rmap_add(mp, dfops, XFS_RMAP_CONVERT, ip->i_ino,
+	return __xfs_rmap_add(mp, dfops, xfs_is_reflink_inode(ip) ?
+			XFS_RMAP_CONVERT_SHARED : XFS_RMAP_CONVERT, ip->i_ino,
 			whichfork, PREV);
 }
 

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 41/71] xfs: don't allow realtime and reflinked files to mix

[Darrick J. Wong]

We don't support sharing blocks on the realtime device.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_inode_buf.c |   16 ++++++++++++++++
 1 file changed, 16 insertions(+)


diff --git a/libxfs/xfs_inode_buf.c b/libxfs/xfs_inode_buf.c
index 8a804e2..2966d54 100644
--- a/libxfs/xfs_inode_buf.c
+++ b/libxfs/xfs_inode_buf.c
@@ -384,6 +384,9 @@ xfs_dinode_verify(
 	xfs_ino_t		ino,
 	struct xfs_dinode	*dip)
 {
+	uint16_t		flags;
+	uint64_t		flags2;
+
 	if (dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC))
 		return false;
 
@@ -400,6 +403,19 @@ xfs_dinode_verify(
 		return false;
 	if (!uuid_equal(&dip->di_uuid, &mp->m_sb.sb_meta_uuid))
 		return false;
+
+	flags = be16_to_cpu(dip->di_flags);
+	flags2 = be64_to_cpu(dip->di_flags2);
+
+	/* don't allow reflink/cowextsize if we don't have reflink */
+	if ((flags2 & (XFS_DIFLAG2_REFLINK | XFS_DIFLAG2_COWEXTSIZE)) &&
+            !xfs_sb_version_hasreflink(&mp->m_sb))
+		return false;
+
+	/* don't let reflink and realtime mix */
+	if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags & XFS_DIFLAG_REALTIME))
+		return false;
+
 	return true;
 }
 

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 42/71] xfs: don't mix reflink and DAX mode for now

[Darrick J. Wong]

Since we don't have a strategy for handling both DAX and reflink,
for now we'll just prohibit both being set at the same time.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_inode_buf.c |    4 ++++
 1 file changed, 4 insertions(+)


diff --git a/libxfs/xfs_inode_buf.c b/libxfs/xfs_inode_buf.c
index 2966d54..3f2049b 100644
--- a/libxfs/xfs_inode_buf.c
+++ b/libxfs/xfs_inode_buf.c
@@ -416,6 +416,10 @@ xfs_dinode_verify(
 	if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags & XFS_DIFLAG_REALTIME))
 		return false;
 
+	/* don't let reflink and dax mix */
+	if ((flags2 & XFS_DIFLAG2_REFLINK) && (flags2 & XFS_DIFLAG2_DAX))
+		return false;
+
 	return true;
 }
 

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 43/71] xfs: recognize the reflink feature bit

[Darrick J. Wong]

Add the reflink feature flag to the set of recognized feature flags.
This enables users to write to reflink filesystems.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/xfs_format.h |    3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)


diff --git a/libxfs/xfs_format.h b/libxfs/xfs_format.h
index e1551f1..f2b9737 100644
--- a/libxfs/xfs_format.h
+++ b/libxfs/xfs_format.h
@@ -459,7 +459,8 @@ xfs_sb_has_compat_feature(
 #define XFS_SB_FEAT_RO_COMPAT_REFLINK  (1 << 2)		/* reflinked files */
 #define XFS_SB_FEAT_RO_COMPAT_ALL \
 		(XFS_SB_FEAT_RO_COMPAT_FINOBT | \
-		 XFS_SB_FEAT_RO_COMPAT_RMAPBT)
+		 XFS_SB_FEAT_RO_COMPAT_RMAPBT | \
+		 XFS_SB_FEAT_RO_COMPAT_REFLINK)
 #define XFS_SB_FEAT_RO_COMPAT_UNKNOWN	~XFS_SB_FEAT_RO_COMPAT_ALL
 static inline bool
 xfs_sb_has_ro_compat_feature(

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 44/71] xfs_db: dump refcount btree data

[Darrick J. Wong]

Add the ability to walk and dump the refcount btree in xfs_db.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 db/agf.c          |   13 +++++++++++--
 db/btblock.c      |   50 ++++++++++++++++++++++++++++++++++++++++++++++++++
 db/btblock.h      |    5 +++++
 db/field.c        |    9 +++++++++
 db/field.h        |    4 ++++
 db/inode.c        |    3 +++
 db/sb.c           |    2 ++
 db/type.c         |    5 +++++
 db/type.h         |    2 +-
 man/man8/xfs_db.8 |   47 +++++++++++++++++++++++++++++++++++++++++++++--
 10 files changed, 135 insertions(+), 5 deletions(-)


diff --git a/db/agf.c b/db/agf.c
index 467dd4c..275f407 100644
--- a/db/agf.c
+++ b/db/agf.c
@@ -47,7 +47,7 @@ const field_t	agf_flds[] = {
 	{ "versionnum", FLDT_UINT32D, OI(OFF(versionnum)), C1, 0, TYP_NONE },
 	{ "seqno", FLDT_AGNUMBER, OI(OFF(seqno)), C1, 0, TYP_NONE },
 	{ "length", FLDT_AGBLOCK, OI(OFF(length)), C1, 0, TYP_NONE },
-	{ "roots", FLDT_AGBLOCK, OI(OFF(roots)), CI(XFS_BTNUM_AGF),
+	{ "roots", FLDT_AGBLOCK, OI(OFF(roots)), CI(XFS_BTNUM_AGF) + 1,
 	  FLD_ARRAY|FLD_SKIPALL, TYP_NONE },
 	{ "bnoroot", FLDT_AGBLOCK,
 	  OI(OFF(roots) + XFS_BTNUM_BNO * SZ(roots[XFS_BTNUM_BNO])), C1, 0,
@@ -58,7 +58,10 @@ const field_t	agf_flds[] = {
 	{ "rmaproot", FLDT_AGBLOCKNZ,
 	  OI(OFF(roots) + XFS_BTNUM_RMAP * SZ(roots[XFS_BTNUM_RMAP])), C1, 0,
 	  TYP_RMAPBT },
-	{ "levels", FLDT_UINT32D, OI(OFF(levels)), CI(XFS_BTNUM_AGF),
+	{ "refcntroot", FLDT_AGBLOCKNZ,
+	  OI(OFF(refcount_root)), C1, 0,
+	  TYP_REFCBT },
+	{ "levels", FLDT_UINT32D, OI(OFF(levels)), CI(XFS_BTNUM_AGF) + 1,
 	  FLD_ARRAY|FLD_SKIPALL, TYP_NONE },
 	{ "bnolevel", FLDT_UINT32D,
 	  OI(OFF(levels) + XFS_BTNUM_BNO * SZ(levels[XFS_BTNUM_BNO])), C1, 0,
@@ -69,9 +72,15 @@ const field_t	agf_flds[] = {
 	{ "rmaplevel", FLDT_UINT32D,
 	  OI(OFF(levels) + XFS_BTNUM_RMAP * SZ(levels[XFS_BTNUM_RMAP])), C1, 0,
 	  TYP_NONE },
+	{ "refcntlevel", FLDT_UINT32D,
+	  OI(OFF(refcount_level)), C1, 0,
+	  TYP_NONE },
 	{ "rmapblocks", FLDT_UINT32D,
 	  OI(OFF(rmap_blocks)), C1, 0,
 	  TYP_NONE },
+	{ "refcntblocks", FLDT_UINT32D,
+	  OI(OFF(refcount_blocks)), C1, 0,
+	  TYP_NONE },
 	{ "flfirst", FLDT_UINT32D, OI(OFF(flfirst)), C1, 0, TYP_NONE },
 	{ "fllast", FLDT_UINT32D, OI(OFF(fllast)), C1, 0, TYP_NONE },
 	{ "flcount", FLDT_UINT32D, OI(OFF(flcount)), C1, 0, TYP_NONE },
diff --git a/db/btblock.c b/db/btblock.c
index ce59d18..e0c896b 100644
--- a/db/btblock.c
+++ b/db/btblock.c
@@ -102,6 +102,12 @@ struct xfs_db_btree {
 		sizeof(struct xfs_rmap_rec),
 		sizeof(__be32),
 	},
+	{	XFS_REFC_CRC_MAGIC,
+		XFS_BTREE_SBLOCK_CRC_LEN,
+		sizeof(struct xfs_refcount_key),
+		sizeof(struct xfs_refcount_rec),
+		sizeof(__be32),
+	},
 	{	0,
 	},
 };
@@ -707,3 +713,47 @@ const field_t	rmapbt_rec_flds[] = {
 	{ NULL }
 };
 #undef ROFF
+
+/* refcount btree blocks */
+const field_t	refcbt_crc_hfld[] = {
+	{ "", FLDT_REFCBT_CRC, OI(0), C1, 0, TYP_NONE },
+	{ NULL }
+};
+
+#define	OFF(f)	bitize(offsetof(struct xfs_btree_block, bb_ ## f))
+const field_t	refcbt_crc_flds[] = {
+	{ "magic", FLDT_UINT32X, OI(OFF(magic)), C1, 0, TYP_NONE },
+	{ "level", FLDT_UINT16D, OI(OFF(level)), C1, 0, TYP_NONE },
+	{ "numrecs", FLDT_UINT16D, OI(OFF(numrecs)), C1, 0, TYP_NONE },
+	{ "leftsib", FLDT_AGBLOCK, OI(OFF(u.s.bb_leftsib)), C1, 0, TYP_REFCBT },
+	{ "rightsib", FLDT_AGBLOCK, OI(OFF(u.s.bb_rightsib)), C1, 0, TYP_REFCBT },
+	{ "bno", FLDT_DFSBNO, OI(OFF(u.s.bb_blkno)), C1, 0, TYP_REFCBT },
+	{ "lsn", FLDT_UINT64X, OI(OFF(u.s.bb_lsn)), C1, 0, TYP_NONE },
+	{ "uuid", FLDT_UUID, OI(OFF(u.s.bb_uuid)), C1, 0, TYP_NONE },
+	{ "owner", FLDT_AGNUMBER, OI(OFF(u.s.bb_owner)), C1, 0, TYP_NONE },
+	{ "crc", FLDT_CRC, OI(OFF(u.s.bb_crc)), C1, 0, TYP_NONE },
+	{ "recs", FLDT_REFCBTREC, btblock_rec_offset, btblock_rec_count,
+	  FLD_ARRAY|FLD_ABASE1|FLD_COUNT|FLD_OFFSET, TYP_NONE },
+	{ "keys", FLDT_REFCBTKEY, btblock_key_offset, btblock_key_count,
+	  FLD_ARRAY|FLD_ABASE1|FLD_COUNT|FLD_OFFSET, TYP_NONE },
+	{ "ptrs", FLDT_REFCBTPTR, btblock_ptr_offset, btblock_key_count,
+	  FLD_ARRAY|FLD_ABASE1|FLD_COUNT|FLD_OFFSET, TYP_REFCBT },
+	{ NULL }
+};
+#undef OFF
+
+#define	KOFF(f)	bitize(offsetof(struct xfs_refcount_key, rc_ ## f))
+const field_t	refcbt_key_flds[] = {
+	{ "startblock", FLDT_AGBLOCK, OI(KOFF(startblock)), C1, 0, TYP_DATA },
+	{ NULL }
+};
+#undef KOFF
+
+#define	ROFF(f)	bitize(offsetof(struct xfs_refcount_rec, rc_ ## f))
+const field_t	refcbt_rec_flds[] = {
+	{ "startblock", FLDT_AGBLOCK, OI(ROFF(startblock)), C1, 0, TYP_DATA },
+	{ "blockcount", FLDT_EXTLEN, OI(ROFF(blockcount)), C1, 0, TYP_NONE },
+	{ "refcount", FLDT_UINT32D, OI(ROFF(refcount)), C1, 0, TYP_DATA },
+	{ NULL }
+};
+#undef ROFF
diff --git a/db/btblock.h b/db/btblock.h
index 35299b4..fead2f1 100644
--- a/db/btblock.h
+++ b/db/btblock.h
@@ -59,4 +59,9 @@ extern const struct field	rmapbt_crc_hfld[];
 extern const struct field	rmapbt_key_flds[];
 extern const struct field	rmapbt_rec_flds[];
 
+extern const struct field	refcbt_crc_flds[];
+extern const struct field	refcbt_crc_hfld[];
+extern const struct field	refcbt_key_flds[];
+extern const struct field	refcbt_rec_flds[];
+
 extern int	btblock_size(void *obj, int startoff, int idx);
diff --git a/db/field.c b/db/field.c
index ca7642f..e0e7066 100644
--- a/db/field.c
+++ b/db/field.c
@@ -183,6 +183,15 @@ const ftattr_t	ftattrtab[] = {
 	{ FLDT_RMAPBTREC, "rmapbtrec", fp_sarray, (char *)rmapbt_rec_flds,
 	  SI(bitsz(struct xfs_rmap_rec)), 0, NULL, rmapbt_rec_flds },
 
+	{ FLDT_REFCBT_CRC, "refcntbt", NULL, (char *)refcbt_crc_flds, btblock_size,
+	  FTARG_SIZE, NULL, refcbt_crc_flds },
+	{ FLDT_REFCBTKEY, "refcntbtkey", fp_sarray, (char *)refcbt_key_flds,
+	  SI(bitsz(struct xfs_refcount_key)), 0, NULL, refcbt_key_flds },
+	{ FLDT_REFCBTPTR, "refcntbtptr", fp_num, "%u", SI(bitsz(xfs_refcount_ptr_t)),
+	  0, fa_agblock, NULL },
+	{ FLDT_REFCBTREC, "refcntbtrec", fp_sarray, (char *)refcbt_rec_flds,
+	  SI(bitsz(struct xfs_refcount_rec)), 0, NULL, refcbt_rec_flds },
+
 /* CRC field */
 	{ FLDT_CRC, "crc", fp_crc, "%#x (%s)", SI(bitsz(__uint32_t)),
 	  0, NULL, NULL },
diff --git a/db/field.h b/db/field.h
index 47f562a..ae5f490 100644
--- a/db/field.h
+++ b/db/field.h
@@ -89,6 +89,10 @@ typedef enum fldt	{
 	FLDT_RMAPBTKEY,
 	FLDT_RMAPBTPTR,
 	FLDT_RMAPBTREC,
+	FLDT_REFCBT_CRC,
+	FLDT_REFCBTKEY,
+	FLDT_REFCBTPTR,
+	FLDT_REFCBTREC,
 
 	/* CRC field type */
 	FLDT_CRC,
diff --git a/db/inode.c b/db/inode.c
index 442e6ea..702cdf8 100644
--- a/db/inode.c
+++ b/db/inode.c
@@ -175,6 +175,9 @@ const field_t	inode_v3_flds[] = {
 	{ "crtime", FLDT_TIMESTAMP, OI(COFF(crtime)), C1, 0, TYP_NONE },
 	{ "inumber", FLDT_INO, OI(COFF(ino)), C1, 0, TYP_NONE },
 	{ "uuid", FLDT_UUID, OI(COFF(uuid)), C1, 0, TYP_NONE },
+	{ "reflink", FLDT_UINT1,
+	  OI(COFF(flags2) + bitsz(__uint64_t) - XFS_DIFLAG2_REFLINK_BIT-1), C1,
+	  0, TYP_NONE },
 	{ NULL }
 };
 
diff --git a/db/sb.c b/db/sb.c
index 79a3c1d..8e7722c 100644
--- a/db/sb.c
+++ b/db/sb.c
@@ -694,6 +694,8 @@ version_string(
 		strcat(s, ",SPARSE_INODES");
 	if (xfs_sb_version_hasmetauuid(sbp))
 		strcat(s, ",META_UUID");
+	if (xfs_sb_version_hasreflink(sbp))
+		strcat(s, ",REFLINK");
 	return s;
 }
 
diff --git a/db/type.c b/db/type.c
index 337243a..10fa54e 100644
--- a/db/type.c
+++ b/db/type.c
@@ -61,6 +61,7 @@ static const typ_t	__typtab[] = {
 	{ TYP_BNOBT, "bnobt", handle_struct, bnobt_hfld, NULL, TYP_F_NO_CRC_OFF },
 	{ TYP_CNTBT, "cntbt", handle_struct, cntbt_hfld, NULL, TYP_F_NO_CRC_OFF },
 	{ TYP_RMAPBT, NULL },
+	{ TYP_REFCBT, NULL },
 	{ TYP_DATA, "data", handle_block, NULL, NULL, TYP_F_NO_CRC_OFF },
 	{ TYP_DIR2, "dir2", handle_struct, dir2_hfld, NULL, TYP_F_NO_CRC_OFF },
 	{ TYP_DQBLK, "dqblk", handle_struct, dqblk_hfld, NULL, TYP_F_NO_CRC_OFF },
@@ -97,6 +98,8 @@ static const typ_t	__typtab_crc[] = {
 		&xfs_allocbt_buf_ops, XFS_BTREE_SBLOCK_CRC_OFF },
 	{ TYP_RMAPBT, "rmapbt", handle_struct, rmapbt_crc_hfld,
 		&xfs_rmapbt_buf_ops, XFS_BTREE_SBLOCK_CRC_OFF },
+	{ TYP_REFCBT, "refcntbt", handle_struct, refcbt_crc_hfld,
+		&xfs_refcountbt_buf_ops, XFS_BTREE_SBLOCK_CRC_OFF },
 	{ TYP_DATA, "data", handle_block, NULL, NULL, TYP_F_NO_CRC_OFF },
 	{ TYP_DIR2, "dir3", handle_struct, dir3_hfld,
 		&xfs_dir3_db_buf_ops, TYP_F_NO_CRC_OFF },
@@ -139,6 +142,8 @@ static const typ_t	__typtab_spcrc[] = {
 		&xfs_allocbt_buf_ops, XFS_BTREE_SBLOCK_CRC_OFF },
 	{ TYP_RMAPBT, "rmapbt", handle_struct, rmapbt_crc_hfld,
 		&xfs_rmapbt_buf_ops, XFS_BTREE_SBLOCK_CRC_OFF },
+	{ TYP_REFCBT, "refcntbt", handle_struct, refcbt_crc_hfld,
+		&xfs_refcountbt_buf_ops, XFS_BTREE_SBLOCK_CRC_OFF },
 	{ TYP_DATA, "data", handle_block, NULL, NULL, TYP_F_NO_CRC_OFF },
 	{ TYP_DIR2, "dir3", handle_struct, dir3_hfld,
 		&xfs_dir3_db_buf_ops, TYP_F_NO_CRC_OFF },
diff --git a/db/type.h b/db/type.h
index b5a21a7..87ff107 100644
--- a/db/type.h
+++ b/db/type.h
@@ -24,7 +24,7 @@ struct field;
 typedef enum typnm
 {
 	TYP_AGF, TYP_AGFL, TYP_AGI, TYP_ATTR, TYP_BMAPBTA,
-	TYP_BMAPBTD, TYP_BNOBT, TYP_CNTBT, TYP_RMAPBT, TYP_DATA,
+	TYP_BMAPBTD, TYP_BNOBT, TYP_CNTBT, TYP_RMAPBT, TYP_REFCBT, TYP_DATA,
 	TYP_DIR2, TYP_DQBLK, TYP_INOBT, TYP_INODATA, TYP_INODE,
 	TYP_LOG, TYP_RTBITMAP, TYP_RTSUMMARY, TYP_SB, TYP_SYMLINK,
 	TYP_TEXT, TYP_FINOBT, TYP_NONE
diff --git a/man/man8/xfs_db.8 b/man/man8/xfs_db.8
index e177913..514e3aa 100644
--- a/man/man8/xfs_db.8
+++ b/man/man8/xfs_db.8
@@ -682,8 +682,8 @@ If no argument is given, show the current data type.
 The possible data types are:
 .BR agf ", " agfl ", " agi ", " attr ", " bmapbta ", " bmapbtd ,
 .BR bnobt ", " cntbt ", " data ", " dir ", " dir2 ", " dqblk ,
-.BR inobt ", " inode ", " log ", " rmapbt ", " rtbitmap ", " rtsummary ,
-.BR sb ", " symlink " and " text .
+.BR inobt ", " inode ", " log ", " refcntbt ", " rmapbt ", " rtbitmap ,
+.BR rtsummary ", " sb ", " symlink " and " text .
 See the TYPES section below for more information on these data types.
 .TP
 .BI "uuid [" uuid " | " generate " | " rewrite " | " restore ]
@@ -1667,6 +1667,49 @@ use
 .BR xfs_logprint (8)
 instead.
 .TP
+.B refcntbt
+There is one set of filesystem blocks forming the reference count Btree for
+each allocation group. The root block of this Btree is designated by the
+.B refcntroot
+field in the corresponding AGF block.  The blocks are linked to sibling left
+and right blocks at each level, as well as by pointers from parent to child
+blocks.  Each block has the following fields:
+.RS 1.4i
+.PD 0
+.TP 1.2i
+.B magic
+REFC block magic number, 0x52334643 ('R3FC').
+.TP
+.B level
+level number of this block, 0 is a leaf.
+.TP
+.B numrecs
+number of data entries in the block.
+.TP
+.B leftsib
+left (logically lower) sibling block, 0 if none.
+.TP
+.B rightsib
+right (logically higher) sibling block, 0 if none.
+.TP
+.B recs
+[leaf blocks only] array of reference count records. Each record contains
+.BR startblock ,
+.BR blockcount ,
+and
+.BR refcount .
+.TP
+.B keys
+[non-leaf blocks only] array of key records. These are the first value
+of each block in the level below this one. Each record contains
+.BR startblock .
+.TP
+.B ptrs
+[non-leaf blocks only] array of child block pointers. Each pointer is a
+block number within the allocation group to the next level in the Btree.
+.PD
+.RE
+.TP
 .B rmapbt
 There is one set of filesystem blocks forming the reverse mapping Btree for
 each allocation group. The root block of this Btree is designated by the

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 45/71] xfs_db: add support for checking the refcount btree

[Darrick J. Wong]

Do some basic checks of the refcount btree.  xfs_repair will have to
check that the reference counts match the various bmbt mappings.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 db/check.c |  136 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++----
 1 file changed, 128 insertions(+), 8 deletions(-)


diff --git a/db/check.c b/db/check.c
index a6a8372..5b90182 100644
--- a/db/check.c
+++ b/db/check.c
@@ -44,7 +44,8 @@ typedef enum {
 	DBM_FREE1,	DBM_FREE2,	DBM_FREELIST,	DBM_INODE,
 	DBM_LOG,	DBM_MISSING,	DBM_QUOTA,	DBM_RTBITMAP,
 	DBM_RTDATA,	DBM_RTFREE,	DBM_RTSUM,	DBM_SB,
-	DBM_SYMLINK,	DBM_BTFINO,	DBM_BTRMAP,
+	DBM_SYMLINK,	DBM_BTFINO,	DBM_BTRMAP,	DBM_BTREFC,
+	DBM_RLDATA,
 	DBM_NDBM
 } dbm_t;
 
@@ -52,7 +53,8 @@ typedef struct inodata {
 	struct inodata	*next;
 	nlink_t		link_set;
 	nlink_t		link_add;
-	char		isdir;
+	char		isdir:1;
+	char		isreflink:1;
 	char		security;
 	char		ilist;
 	xfs_ino_t	ino;
@@ -172,6 +174,8 @@ static const char	*typename[] = {
 	"symlink",
 	"btfino",
 	"btrmap",
+	"btrefcnt",
+	"rldata",
 	NULL
 };
 static int		verbose;
@@ -229,7 +233,8 @@ static int		blocktrash_f(int argc, char **argv);
 static int		blockuse_f(int argc, char **argv);
 static int		check_blist(xfs_fsblock_t bno);
 static void		check_dbmap(xfs_agnumber_t agno, xfs_agblock_t agbno,
-				    xfs_extlen_t len, dbm_t type);
+				    xfs_extlen_t len, dbm_t type,
+				    int ignore_reflink);
 static int		check_inomap(xfs_agnumber_t agno, xfs_agblock_t agbno,
 				     xfs_extlen_t len, xfs_ino_t c_ino);
 static void		check_linkcounts(xfs_agnumber_t agno);
@@ -353,6 +358,9 @@ static void		scanfunc_fino(struct xfs_btree_block *block, int level,
 static void		scanfunc_rmap(struct xfs_btree_block *block, int level,
 				     struct xfs_agf *agf, xfs_agblock_t bno,
 				     int isroot);
+static void		scanfunc_refcnt(struct xfs_btree_block *block, int level,
+				     struct xfs_agf *agf, xfs_agblock_t bno,
+				     int isroot);
 static void		set_dbmap(xfs_agnumber_t agno, xfs_agblock_t agbno,
 				  xfs_extlen_t len, dbm_t type,
 				  xfs_agnumber_t c_agno, xfs_agblock_t c_agbno);
@@ -1055,6 +1063,7 @@ blocktrash_f(
 		   (1 << DBM_SYMLINK) |
 		   (1 << DBM_BTFINO) |
 		   (1 << DBM_BTRMAP) |
+		   (1 << DBM_BTREFC) |
 		   (1 << DBM_SB);
 	while ((c = getopt(argc, argv, "0123n:o:s:t:x:y:z")) != EOF) {
 		switch (c) {
@@ -1291,18 +1300,25 @@ check_dbmap(
 	xfs_agnumber_t	agno,
 	xfs_agblock_t	agbno,
 	xfs_extlen_t	len,
-	dbm_t		type)
+	dbm_t		type,
+	int		ignore_reflink)
 {
 	xfs_extlen_t	i;
 	char		*p;
+	dbm_t		d;
 
 	for (i = 0, p = &dbmap[agno][agbno]; i < len; i++, p++) {
+		d = (dbm_t)*p;
+		if (ignore_reflink && (d == DBM_UNKNOWN || d == DBM_DATA ||
+				       d == DBM_RLDATA))
+			continue;
 		if ((dbm_t)*p != type) {
-			if (!sflag || CHECK_BLISTA(agno, agbno + i))
+			if (!sflag || CHECK_BLISTA(agno, agbno + i)) {
 				dbprintf(_("block %u/%u expected type %s got "
 					 "%s\n"),
 					agno, agbno + i, typename[type],
 					typename[(dbm_t)*p]);
+			}
 			error++;
 		}
 	}
@@ -1336,7 +1352,7 @@ check_inomap(
 		return 0;
 	}
 	for (i = 0, rval = 1, idp = &inomap[agno][agbno]; i < len; i++, idp++) {
-		if (*idp) {
+		if (*idp && !(*idp)->isreflink) {
 			if (!sflag || (*idp)->ilist ||
 			    CHECK_BLISTA(agno, agbno + i))
 				dbprintf(_("block %u/%u claimed by inode %lld, "
@@ -1542,6 +1558,26 @@ check_rrange(
 	return 1;
 }
 
+/*
+ * We don't check the accuracy of reference counts -- all we do is ensure
+ * that a data block never crosses with non-data blocks.  repair can check
+ * those kinds of things.
+ *
+ * So with that in mind, if we're setting a block to be data or rldata,
+ * don't complain so long as the block is currently unknown, data, or rldata.
+ * Don't let blocks downgrade from rldata -> data.
+ */
+static bool
+is_reflink(
+	dbm_t		type2)
+{
+	if (!xfs_sb_version_hasreflink(&mp->m_sb))
+		return false;
+	if (type2 == DBM_DATA || type2 == DBM_RLDATA)
+		return true;
+	return false;
+}
+
 static void
 check_set_dbmap(
 	xfs_agnumber_t	agno,
@@ -1561,10 +1597,15 @@ check_set_dbmap(
 			agbno, agbno + len - 1, c_agno, c_agbno);
 		return;
 	}
-	check_dbmap(agno, agbno, len, type1);
+	check_dbmap(agno, agbno, len, type1, is_reflink(type2));
 	mayprint = verbose | blist_size;
 	for (i = 0, p = &dbmap[agno][agbno]; i < len; i++, p++) {
-		*p = (char)type2;
+		if (*p == DBM_RLDATA && type2 == DBM_DATA)
+			;	/* do nothing */
+		if (*p == DBM_DATA && type2 == DBM_DATA)
+			*p = (char)DBM_RLDATA;
+		else
+			*p = (char)type2;
 		if (mayprint && (verbose || CHECK_BLISTA(agno, agbno + i)))
 			dbprintf(_("setting block %u/%u to %s\n"), agno, agbno + i,
 				typename[type2]);
@@ -2804,6 +2845,7 @@ process_inode(
 		break;
 	}
 
+	id->isreflink = !!(xino.i_d.di_flags2 & XFS_DIFLAG2_REFLINK);
 	setlink_inode(id, VFS_I(&xino)->i_nlink, type == DBM_DIR, security);
 
 	switch (xino.i_d.di_format) {
@@ -3910,6 +3952,12 @@ scan_ag(
 			be32_to_cpu(agf->agf_levels[XFS_BTNUM_RMAP]),
 			1, scanfunc_rmap, TYP_RMAPBT);
 	}
+	if (agf->agf_refcount_root) {
+		scan_sbtree(agf,
+			be32_to_cpu(agf->agf_refcount_root),
+			be32_to_cpu(agf->agf_refcount_level),
+			1, scanfunc_refcnt, TYP_REFCBT);
+	}
 	scan_sbtree(agf,
 		be32_to_cpu(agi->agi_root),
 		be32_to_cpu(agi->agi_level),
@@ -4643,6 +4691,78 @@ scanfunc_rmap(
 }
 
 static void
+scanfunc_refcnt(
+	struct xfs_btree_block	*block,
+	int			level,
+	struct xfs_agf		*agf,
+	xfs_agblock_t		bno,
+	int			isroot)
+{
+	xfs_agnumber_t		seqno = be32_to_cpu(agf->agf_seqno);
+	int			i;
+	xfs_refcount_ptr_t	*pp;
+	struct xfs_refcount_rec	*rp;
+	xfs_agblock_t		lastblock;
+
+	if (be32_to_cpu(block->bb_magic) != XFS_REFC_CRC_MAGIC) {
+		dbprintf(_("bad magic # %#x in refcntbt block %u/%u\n"),
+			be32_to_cpu(block->bb_magic), seqno, bno);
+		serious_error++;
+		return;
+	}
+	if (be16_to_cpu(block->bb_level) != level) {
+		if (!sflag)
+			dbprintf(_("expected level %d got %d in refcntbt block "
+				 "%u/%u\n"),
+				level, be16_to_cpu(block->bb_level), seqno, bno);
+		error++;
+	}
+	set_dbmap(seqno, bno, 1, DBM_BTREFC, seqno, bno);
+	if (level == 0) {
+		if (be16_to_cpu(block->bb_numrecs) > mp->m_refc_mxr[0] ||
+		    (isroot == 0 && be16_to_cpu(block->bb_numrecs) < mp->m_refc_mnr[0])) {
+			dbprintf(_("bad btree nrecs (%u, min=%u, max=%u) in "
+				 "refcntbt block %u/%u\n"),
+				be16_to_cpu(block->bb_numrecs), mp->m_refc_mnr[0],
+				mp->m_refc_mxr[0], seqno, bno);
+			serious_error++;
+			return;
+		}
+		rp = XFS_REFCOUNT_REC_ADDR(block, 1);
+		lastblock = 0;
+		for (i = 0; i < be16_to_cpu(block->bb_numrecs); i++) {
+			set_dbmap(seqno, be32_to_cpu(rp[i].rc_startblock),
+				be32_to_cpu(rp[i].rc_blockcount), DBM_RLDATA,
+				seqno, bno);
+			if (be32_to_cpu(rp[i].rc_startblock) < lastblock) {
+				dbprintf(_(
+		"out-of-order refcnt btree record %d (%u %u) block %u/%u\n"),
+					 i, be32_to_cpu(rp[i].rc_startblock),
+					 be32_to_cpu(rp[i].rc_startblock),
+					 be32_to_cpu(agf->agf_seqno), bno);
+			} else {
+				lastblock = be32_to_cpu(rp[i].rc_startblock) +
+					    be32_to_cpu(rp[i].rc_blockcount);
+			}
+		}
+		return;
+	}
+	if (be16_to_cpu(block->bb_numrecs) > mp->m_refc_mxr[1] ||
+	    (isroot == 0 && be16_to_cpu(block->bb_numrecs) < mp->m_refc_mnr[1])) {
+		dbprintf(_("bad btree nrecs (%u, min=%u, max=%u) in refcntbt "
+			 "block %u/%u\n"),
+			be16_to_cpu(block->bb_numrecs), mp->m_refc_mnr[1],
+			mp->m_refc_mxr[1], seqno, bno);
+		serious_error++;
+		return;
+	}
+	pp = XFS_REFCOUNT_PTR_ADDR(block, 1, mp->m_refc_mxr[1]);
+	for (i = 0; i < be16_to_cpu(block->bb_numrecs); i++)
+		scan_sbtree(agf, be32_to_cpu(pp[i]), level, 0, scanfunc_refcnt,
+				TYP_REFCBT);
+}
+
+static void
 set_dbmap(
 	xfs_agnumber_t	agno,
 	xfs_agblock_t	agbno,

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 46/71] xfs_db: metadump should copy the refcount btree too

[Darrick J. Wong]

Teach metadump to copy the refcount btree.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 db/metadump.c |   74 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 74 insertions(+)


diff --git a/db/metadump.c b/db/metadump.c
index 44359e1..ab64fbd 100644
--- a/db/metadump.c
+++ b/db/metadump.c
@@ -615,6 +615,78 @@ copy_rmap_btree(
 	return scan_btree(agno, root, levels, TYP_RMAPBT, agf, scanfunc_rmapbt);
 }
 
+static int
+scanfunc_refcntbt(
+	struct xfs_btree_block	*block,
+	xfs_agnumber_t		agno,
+	xfs_agblock_t		agbno,
+	int			level,
+	typnm_t			btype,
+	void			*arg)
+{
+	xfs_refcount_ptr_t	*pp;
+	int			i;
+	int			numrecs;
+
+	if (level == 0)
+		return 1;
+
+	numrecs = be16_to_cpu(block->bb_numrecs);
+	if (numrecs > mp->m_refc_mxr[1]) {
+		if (show_warnings)
+			print_warning("invalid numrecs (%u) in %s block %u/%u",
+				numrecs, typtab[btype].name, agno, agbno);
+		return 1;
+	}
+
+	pp = XFS_REFCOUNT_PTR_ADDR(block, 1, mp->m_refc_mxr[1]);
+	for (i = 0; i < numrecs; i++) {
+		if (!valid_bno(agno, be32_to_cpu(pp[i]))) {
+			if (show_warnings)
+				print_warning("invalid block number (%u/%u) "
+					"in %s block %u/%u",
+					agno, be32_to_cpu(pp[i]),
+					typtab[btype].name, agno, agbno);
+			continue;
+		}
+		if (!scan_btree(agno, be32_to_cpu(pp[i]), level, btype, arg,
+				scanfunc_refcntbt))
+			return 0;
+	}
+	return 1;
+}
+
+static int
+copy_refcount_btree(
+	xfs_agnumber_t	agno,
+	struct xfs_agf	*agf)
+{
+	xfs_agblock_t	root;
+	int		levels;
+
+	if (!xfs_sb_version_hasreflink(&mp->m_sb))
+		return 1;
+
+	root = be32_to_cpu(agf->agf_refcount_root);
+	levels = be32_to_cpu(agf->agf_refcount_level);
+
+	/* validate root and levels before processing the tree */
+	if (root == 0 || root > mp->m_sb.sb_agblocks) {
+		if (show_warnings)
+			print_warning("invalid block number (%u) in refcntbt "
+					"root in agf %u", root, agno);
+		return 1;
+	}
+	if (levels >= XFS_BTREE_MAXLEVELS) {
+		if (show_warnings)
+			print_warning("invalid level (%u) in refcntbt root "
+					"in agf %u", levels, agno);
+		return 1;
+	}
+
+	return scan_btree(agno, root, levels, TYP_REFCBT, agf, scanfunc_refcntbt);
+}
+
 /* filename and extended attribute obfuscation routines */
 
 struct name_ent {
@@ -2525,6 +2597,8 @@ scan_ag(
 			goto pop_out;
 		if (!copy_rmap_btree(agno, agf))
 			goto pop_out;
+		if (!copy_refcount_btree(agno, agf))
+			goto pop_out;
 	}
 
 	/* copy inode btrees and the inodes and their associated metadata */

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 47/71] xfs_db: deal with the CoW extent size hint

[Darrick J. Wong]

Display the CoW extent hint size when dumping inodes.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 db/inode.c |    4 ++++
 1 file changed, 4 insertions(+)


diff --git a/db/inode.c b/db/inode.c
index 702cdf8..cac19fc 100644
--- a/db/inode.c
+++ b/db/inode.c
@@ -172,12 +172,16 @@ const field_t	inode_v3_flds[] = {
 	{ "change_count", FLDT_UINT64D, OI(COFF(changecount)), C1, 0, TYP_NONE },
 	{ "lsn", FLDT_UINT64X, OI(COFF(lsn)), C1, 0, TYP_NONE },
 	{ "flags2", FLDT_UINT64X, OI(COFF(flags2)), C1, 0, TYP_NONE },
+	{ "cowextsize", FLDT_EXTLEN, OI(COFF(cowextsize)), C1, 0, TYP_NONE },
 	{ "crtime", FLDT_TIMESTAMP, OI(COFF(crtime)), C1, 0, TYP_NONE },
 	{ "inumber", FLDT_INO, OI(COFF(ino)), C1, 0, TYP_NONE },
 	{ "uuid", FLDT_UUID, OI(COFF(uuid)), C1, 0, TYP_NONE },
 	{ "reflink", FLDT_UINT1,
 	  OI(COFF(flags2) + bitsz(__uint64_t) - XFS_DIFLAG2_REFLINK_BIT-1), C1,
 	  0, TYP_NONE },
+	{ "cowextsz", FLDT_UINT1,
+	  OI(COFF(flags2) + bitsz(__uint64_t) - XFS_DIFLAG2_COWEXTSIZE_BIT-1), C1,
+	  0, TYP_NONE },
 	{ NULL }
 };
 

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 48/71] xfs_db: print one array element per line

[Darrick J. Wong]

Print one array element per line so that the debugger output isn't
a gigantic pile of screen snow.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 db/print.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)


diff --git a/db/print.c b/db/print.c
index 998daf4..e31372f 100644
--- a/db/print.c
+++ b/db/print.c
@@ -197,7 +197,7 @@ print_sarray(
 	     i < count && !seenint();
 	     i++, bitoff += size) {
 		if (array)
-			dbprintf("%d:", i + base);
+			dbprintf("\n%d:", i + base);
 		for (f = flds, first = 1; f->name; f++) {
 			if (f->flags & FLD_SKIPALL)
 				continue;

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 49/71] xfs_growfs: report the presence of the reflink feature

[Darrick J. Wong]

Report the presence of the reflink feature in xfs_info.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 growfs/xfs_growfs.c |   12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)


diff --git a/growfs/xfs_growfs.c b/growfs/xfs_growfs.c
index 2b46480..a294e14 100644
--- a/growfs/xfs_growfs.c
+++ b/growfs/xfs_growfs.c
@@ -59,12 +59,14 @@ report_info(
 	int		ftype_enabled,
 	int		finobt_enabled,
 	int		spinodes,
-	int		rmapbt_enabled)
+	int		rmapbt_enabled,
+	int		reflink_enabled)
 {
 	printf(_(
 	    "meta-data=%-22s isize=%-6u agcount=%u, agsize=%u blks\n"
 	    "         =%-22s sectsz=%-5u attr=%u, projid32bit=%u\n"
 	    "         =%-22s crc=%-8u finobt=%u spinodes=%u rmapbt=%u\n"
+	    "         =%-22s reflink=%u\n"
 	    "data     =%-22s bsize=%-6u blocks=%llu, imaxpct=%u\n"
 	    "         =%-22s sunit=%-6u swidth=%u blks\n"
 	    "naming   =version %-14u bsize=%-6u ascii-ci=%d ftype=%d\n"
@@ -75,6 +77,7 @@ report_info(
 		mntpoint, geo.inodesize, geo.agcount, geo.agblocks,
 		"", geo.sectsize, attrversion, projid32bit,
 		"", crcs_enabled, finobt_enabled, spinodes, rmapbt_enabled,
+		"", reflink_enabled,
 		"", geo.blocksize, (unsigned long long)geo.datablocks,
 			geo.imaxpct,
 		"", geo.sunit, geo.swidth,
@@ -129,6 +132,7 @@ main(int argc, char **argv)
 	int			finobt_enabled;	/* free inode btree */
 	int			spinodes;
 	int			rmapbt_enabled;
+	int			reflink_enabled;
 
 	progname = basename(argv[0]);
 	setlocale(LC_ALL, "");
@@ -253,12 +257,13 @@ main(int argc, char **argv)
 	finobt_enabled = geo.flags & XFS_FSOP_GEOM_FLAGS_FINOBT ? 1 : 0;
 	spinodes = geo.flags & XFS_FSOP_GEOM_FLAGS_SPINODES ? 1 : 0;
 	rmapbt_enabled = geo.flags & XFS_FSOP_GEOM_FLAGS_RMAPBT ? 1 : 0;
+	reflink_enabled = geo.flags & XFS_FSOP_GEOM_FLAGS_REFLINK ? 1 : 0;
 	if (nflag) {
 		report_info(geo, datadev, isint, logdev, rtdev,
 				lazycount, dirversion, logversion,
 				attrversion, projid32bit, crcs_enabled, ci,
 				ftype_enabled, finobt_enabled, spinodes,
-				rmapbt_enabled);
+				rmapbt_enabled, reflink_enabled);
 		exit(0);
 	}
 
@@ -296,7 +301,8 @@ main(int argc, char **argv)
 	report_info(geo, datadev, isint, logdev, rtdev,
 			lazycount, dirversion, logversion,
 			attrversion, projid32bit, crcs_enabled, ci, ftype_enabled,
-			finobt_enabled, spinodes, rmapbt_enabled);
+			finobt_enabled, spinodes, rmapbt_enabled,
+			reflink_enabled);
 
 	ddsize = xi.dsize;
 	dlsize = ( xi.logBBsize? xi.logBBsize :

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 50/71] xfs_io: bmap should support querying CoW fork, shared blocks

[Darrick J. Wong]

Teach the bmap command to report shared and delayed allocation
extents, and to be able to query the CoW fork.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 io/bmap.c           |   43 ++++++++++++++++++++++++++++++++++---------
 man/man8/xfs_bmap.8 |   14 ++++++++++++++
 man/man8/xfs_io.8   |    2 +-
 3 files changed, 49 insertions(+), 10 deletions(-)


diff --git a/io/bmap.c b/io/bmap.c
index b2e48da..2333244 100644
--- a/io/bmap.c
+++ b/io/bmap.c
@@ -41,7 +41,9 @@ bmap_help(void)
 " Holes are marked by replacing the startblock..endblock with 'hole'.\n"
 " All the file offsets and disk blocks are in units of 512-byte blocks.\n"
 " -a -- prints the attribute fork map instead of the data fork.\n"
+" -c -- prints the copy-on-write fork map instead of the data fork.\n"
 " -d -- suppresses a DMAPI read event, offline portions shown as holes.\n"
+" -e -- print delayed allocation extents.\n"
 " -l -- also displays the length of each extent in 512-byte blocks.\n"
 " -n -- query n extents.\n"
 " -p -- obtain all unwritten extents as well (w/ -v show which are unwritten.)\n"
@@ -75,6 +77,7 @@ bmap_f(
 	int			loop = 0;
 	int			flg = 0;
 	int			aflag = 0;
+	int			cflag = 0;
 	int			lflag = 0;
 	int			nflag = 0;
 	int			pflag = 0;
@@ -85,12 +88,19 @@ bmap_f(
 	int			c;
 	int			egcnt;
 
-	while ((c = getopt(argc, argv, "adln:pv")) != EOF) {
+	while ((c = getopt(argc, argv, "acdeln:pv")) != EOF) {
 		switch (c) {
 		case 'a':	/* Attribute fork. */
 			bmv_iflags |= BMV_IF_ATTRFORK;
 			aflag = 1;
 			break;
+		case 'c':	/* CoW fork. */
+			bmv_iflags |= BMV_IF_COWFORK | BMV_IF_DELALLOC;
+			cflag = 1;
+			break;
+		case 'e':
+			bmv_iflags |= BMV_IF_DELALLOC;
+			break;
 		case 'l':	/* list number of blocks with each extent */
 			lflag = 1;
 			break;
@@ -113,7 +123,7 @@ bmap_f(
 			return command_usage(&bmap_cmd);
 		}
 	}
-	if (aflag)
+	if (aflag || cflag)
 		bmv_iflags &= ~(BMV_IF_PREALLOC|BMV_IF_NO_DMAPI_READ);
 
 	if (vflag) {
@@ -273,13 +283,14 @@ bmap_f(
 #define MINRANGE_WIDTH	16
 #define MINAG_WIDTH	2
 #define MINTOT_WIDTH	5
-#define NFLG		5	/* count of flags */
-#define	FLG_NULL	000000	/* Null flag */
-#define	FLG_PRE		010000	/* Unwritten extent */
-#define	FLG_BSU		001000	/* Not on begin of stripe unit  */
-#define	FLG_ESU		000100	/* Not on end   of stripe unit  */
-#define	FLG_BSW		000010	/* Not on begin of stripe width */
-#define	FLG_ESW		000001	/* Not on end   of stripe width */
+#define NFLG		6	/* count of flags */
+#define	FLG_NULL	0000000	/* Null flag */
+#define	FLG_SHARED	0100000	/* shared extent */
+#define	FLG_PRE		0010000	/* Unwritten extent */
+#define	FLG_BSU		0001000	/* Not on begin of stripe unit  */
+#define	FLG_ESU		0000100	/* Not on end   of stripe unit  */
+#define	FLG_BSW		0000010	/* Not on begin of stripe width */
+#define	FLG_ESW		0000001	/* Not on end   of stripe width */
 		int	agno;
 		off64_t agoff, bbperag;
 		int	foff_w, boff_w, aoff_w, tot_w, agno_w;
@@ -350,6 +361,10 @@ bmap_f(
 			if (map[i + 1].bmv_oflags & BMV_OF_PREALLOC) {
 				flg |= FLG_PRE;
 			}
+			if (map[i + 1].bmv_oflags & BMV_OF_SHARED)
+				flg |= FLG_SHARED;
+			if (map[i + 1].bmv_oflags & BMV_OF_DELALLOC)
+				map[i + 1].bmv_block = -2;
 			/*
 			 * If striping enabled, determine if extent starts/ends
 			 * on a stripe unit boundary.
@@ -382,6 +397,14 @@ bmap_f(
 					agno_w, "",
 					aoff_w, "",
 					tot_w, (long long)map[i+1].bmv_length);
+			} else if (map[i + 1].bmv_block == -2) {
+				printf("%4d: %-*s %-*s %*s %-*s %*lld\n",
+					i,
+					foff_w, rbuf,
+					boff_w, _("delalloc"),
+					agno_w, "",
+					aoff_w, "",
+					tot_w, (long long)map[i+1].bmv_length);
 			} else {
 				snprintf(bbuf, sizeof(bbuf), "%lld..%lld",
 					(long long) map[i + 1].bmv_block,
@@ -413,6 +436,8 @@ bmap_f(
 		}
 		if ((flg || pflag) && vflag > 1) {
 			printf(_(" FLAG Values:\n"));
+			printf(_("    %*.*o Shared extent\n"),
+				NFLG+1, NFLG+1, FLG_SHARED);
 			printf(_("    %*.*o Unwritten preallocated extent\n"),
 				NFLG+1, NFLG+1, FLG_PRE);
 			printf(_("    %*.*o Doesn't begin on stripe unit\n"),
diff --git a/man/man8/xfs_bmap.8 b/man/man8/xfs_bmap.8
index e196559..098cfae 100644
--- a/man/man8/xfs_bmap.8
+++ b/man/man8/xfs_bmap.8
@@ -36,6 +36,10 @@ no matter what the filesystem's block size is.
 If this option is specified, information about the file's
 attribute fork is printed instead of the default data fork.
 .TP
+.B \-c
+If this option is specified, information about the file's
+copy on write fork is printed instead of the default data fork.
+.TP
 .B \-d
 If portions of the file have been migrated offline by
 a DMAPI application, a DMAPI read event will be generated to
@@ -45,6 +49,16 @@ printed.  However if the
 option is used, no DMAPI read event will be generated for a
 DMAPI file and offline portions will be reported as holes.
 .TP
+.B \-e
+If this option is used,
+.B xfs_bmap
+obtains all delayed allocation extents, and does not flush dirty pages
+to disk before querying extent data. With the
+.B \-v
+option, the
+.I flags
+column will show which extents have not yet been allocated.
+.TP
 .B \-l
 If this option is used, then
 .IP
diff --git a/man/man8/xfs_io.8 b/man/man8/xfs_io.8
index 2c56f09..d089524 100644
--- a/man/man8/xfs_io.8
+++ b/man/man8/xfs_io.8
@@ -256,7 +256,7 @@ See the
 .B pwrite
 command.
 .TP
-.BI "bmap [ \-adlpv ] [ \-n " nx " ]"
+.BI "bmap [ \-acdelpv ] [ \-n " nx " ]"
 Prints the block mapping for the current open file. Refer to the
 .BR xfs_bmap (8)
 manual page for complete documentation.

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 51/71] libxfs: add configure option to override system header fsxattr

[Darrick J. Wong]

By default, libxfs will use the kernel/system headers to define struct
fsxattr.  Unfortunately, this creates a problem for developers who are
writing new features but building xfsprogs on a stable system, because
the stable kernel's headers don't reflect the new feature.  In this
case, we want to be able to use the internal fsxattr definition while
the kernel headers catch up, so provide some configure magic to allow
developers to force the use of the internal definition.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 configure.ac         |    5 +++++
 include/builddefs.in |    4 ++++
 include/linux.h      |   10 +++++++++-
 io/fiemap.c          |    1 -
 4 files changed, 18 insertions(+), 2 deletions(-)


diff --git a/configure.ac b/configure.ac
index 1bb5fef..66a562f 100644
--- a/configure.ac
+++ b/configure.ac
@@ -61,6 +61,11 @@ AC_ARG_ENABLE(librt,
 	enable_librt=yes)
 AC_SUBST(enable_librt)
 
+AC_ARG_ENABLE(internal-fsxattr,
+[ --enable-internal-fsxattr=[yes/no] Override system definition of struct fsxattr [default=no]],,
+	enable_internal_fsxattr=no)
+AC_SUBST(enable_internal_fsxattr)
+
 #
 # If the user specified a libdir ending in lib64 do not append another
 # 64 to the library names.
diff --git a/include/builddefs.in b/include/builddefs.in
index 7153d7a..fd7eb74 100644
--- a/include/builddefs.in
+++ b/include/builddefs.in
@@ -109,6 +109,7 @@ HAVE_MNTENT = @have_mntent@
 HAVE_FLS = @have_fls@
 HAVE_FSETXATTR = @have_fsetxattr@
 HAVE_MREMAP = @have_mremap@
+ENABLE_INTERNAL_FSXATTR = @enable_internal_fsxattr@
 
 GCCFLAGS = -funsigned-char -fno-strict-aliasing -Wall
 #	   -Wbitwise -Wno-transparent-union -Wno-old-initializer -Wno-decl
@@ -148,6 +149,9 @@ endif
 ifeq ($(ENABLE_BLKID),yes)
 PCFLAGS+= -DENABLE_BLKID
 endif
+ifeq ($(ENABLE_INTERNAL_FSXATTR),yes)
+PCFLAGS+= -DOVERRIDE_SYSTEM_FSXATTR
+endif
 
 
 GCFLAGS = $(OPTIMIZER) $(DEBUG) \
diff --git a/include/linux.h b/include/linux.h
index bc7af9f..cd0de71 100644
--- a/include/linux.h
+++ b/include/linux.h
@@ -32,7 +32,13 @@
 #include <stdio.h>
 #include <asm/types.h>
 #include <mntent.h>
+#ifdef OVERRIDE_SYSTEM_FSXATTR
+# define fsxattr sys_fsxattr
+#endif
 #include <linux/fs.h> /* fsxattr defintion for new kernels */
+#ifdef OVERRIDE_SYSTEM_FSXATTR
+# undef fsxattr
+#endif
 
 static __inline__ int xfsctl(const char *path, int fd, int cmd, void *p)
 {
@@ -175,7 +181,7 @@ static inline void platform_mntent_close(struct mntent_cursor * cursor)
  * are a copy of the definitions moved to linux/uapi/fs.h in the 4.5 kernel,
  * so this is purely for supporting builds against old kernel headers.
  */
-#ifndef FS_IOC_FSGETXATTR
+#if !defined FS_IOC_FSGETXATTR || defined OVERRIDE_SYSTEM_FSXATTR
 struct fsxattr {
 	__u32		fsx_xflags;	/* xflags field value (get/set) */
 	__u32		fsx_extsize;	/* extsize field value (get/set)*/
@@ -184,7 +190,9 @@ struct fsxattr {
 	__u32		fsx_cowextsize;	/* cow extsize field value (get/set) */
 	unsigned char	fsx_pad[8];
 };
+#endif
 
+#ifndef FS_IOC_FSGETXATTR
 /*
  * Flags for the fsx_xflags field
  */
diff --git a/io/fiemap.c b/io/fiemap.c
index f89da06..bcbae49 100644
--- a/io/fiemap.c
+++ b/io/fiemap.c
@@ -19,7 +19,6 @@
 #include "platform_defs.h"
 #include "command.h"
 #include <linux/fiemap.h>
-#include <linux/fs.h>
 #include "init.h"
 #include "io.h"
 

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 52/71] xfs_io: get and set the CoW extent size hint

[Darrick J. Wong]

Enable administrators to get or set the CoW extent size hint.
Report the hint when we run stat.  This also requires some
autoconf magic to detect whether or not fsx_cowextsize exists.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 configure.ac          |    1 
 include/builddefs.in  |    4 +
 io/Makefile           |    5 +
 io/attr.c             |    8 ++
 io/cowextsize.c       |  202 +++++++++++++++++++++++++++++++++++++++++++++++++
 io/init.c             |    1 
 io/io.h               |    6 +
 io/open.c             |    3 +
 m4/package_libcdev.m4 |   26 ++++++
 man/man8/xfs_io.8     |   16 ++++
 10 files changed, 271 insertions(+), 1 deletion(-)
 create mode 100644 io/cowextsize.c


diff --git a/configure.ac b/configure.ac
index 66a562f..875d4bb 100644
--- a/configure.ac
+++ b/configure.ac
@@ -134,6 +134,7 @@ AC_HAVE_FLS
 AC_HAVE_READDIR
 AC_HAVE_FSETXATTR
 AC_HAVE_MREMAP
+AC_HAVE_FSXATTR_COWEXTSIZE
 
 if test "$enable_blkid" = yes; then
 AC_HAVE_BLKID_TOPO
diff --git a/include/builddefs.in b/include/builddefs.in
index fd7eb74..165fa78 100644
--- a/include/builddefs.in
+++ b/include/builddefs.in
@@ -109,6 +109,7 @@ HAVE_MNTENT = @have_mntent@
 HAVE_FLS = @have_fls@
 HAVE_FSETXATTR = @have_fsetxattr@
 HAVE_MREMAP = @have_mremap@
+HAVE_FSXATTR_COWEXTSIZE = @have_fsxattr_cowextsize@
 ENABLE_INTERNAL_FSXATTR = @enable_internal_fsxattr@
 
 GCCFLAGS = -funsigned-char -fno-strict-aliasing -Wall
@@ -149,6 +150,9 @@ endif
 ifeq ($(ENABLE_BLKID),yes)
 PCFLAGS+= -DENABLE_BLKID
 endif
+ifeq ($(HAVE_FSXATTR_COWEXTSIZE),yes)
+PCFLAGS+= -DHAVE_FSXATTR_COWEXTSIZE
+endif
 ifeq ($(ENABLE_INTERNAL_FSXATTR),yes)
 PCFLAGS+= -DOVERRIDE_SYSTEM_FSXATTR
 endif
diff --git a/io/Makefile b/io/Makefile
index 62bc03b..1997ca9 100644
--- a/io/Makefile
+++ b/io/Makefile
@@ -99,6 +99,11 @@ ifeq ($(HAVE_MREMAP),yes)
 LCFLAGS += -DHAVE_MREMAP
 endif
 
+ifeq ($(HAVE_FSXATTR_COWEXTSIZE),yes)
+CFILES += cowextsize.c
+# -DHAVE_FSXATTR_COWEXTSIZE already set in PCFLAGS
+endif
+
 default: depend $(LTCOMMAND)
 
 include $(BUILDRULES)
diff --git a/io/attr.c b/io/attr.c
index 0186b1d..13bec73 100644
--- a/io/attr.c
+++ b/io/attr.c
@@ -48,9 +48,11 @@ static struct xflags {
 	{ FS_XFLAG_NODEFRAG,		"f", "no-defrag"	},
 	{ FS_XFLAG_FILESTREAM,		"S", "filestream"	},
 	{ FS_XFLAG_DAX,			"x", "dax"		},
+	{ FS_XFLAG_REFLINK,		"R", "reflink"		},
+	{ FS_XFLAG_COWEXTSIZE,		"C", "cowextsize"	},
 	{ 0, NULL, NULL }
 };
-#define CHATTR_XFLAG_LIST	"r"/*p*/"iasAdtPneEfSx"
+#define CHATTR_XFLAG_LIST	"r"/*p*/"iasAdtPneEfSxRC"
 
 static void
 lsattr_help(void)
@@ -75,6 +77,8 @@ lsattr_help(void)
 " f -- do not include this file when defragmenting the filesystem\n"
 " S -- enable filestreams allocator for this directory\n"
 " x -- Use direct access (DAX) for data in this file\n"
+" R -- file data blocks may be shared with another file\n"
+" C -- for files with shared blocks, observe the inode CoW extent size value\n"
 "\n"
 " Options:\n"
 " -R -- recursively descend (useful when current file is a directory)\n"
@@ -111,6 +115,8 @@ chattr_help(void)
 " +/-f -- set/clear the no-defrag flag\n"
 " +/-S -- set/clear the filestreams allocator flag\n"
 " +/-x -- set/clear the direct access (DAX) flag\n"
+" +/-R -- set/clear the reflink flag\n"
+" +/-C -- set/clear the CoW extent-size flag\n"
 " Note1: user must have certain capabilities to modify immutable/append-only.\n"
 " Note2: immutable/append-only files cannot be deleted; removing these files\n"
 "        requires the immutable/append-only flag to be cleared first.\n"
diff --git a/io/cowextsize.c b/io/cowextsize.c
new file mode 100644
index 0000000..b4a1c2e
--- /dev/null
+++ b/io/cowextsize.c
@@ -0,0 +1,202 @@
+/*
+ * Copyright (C) 2016 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+/*
+ * If configure didn't find a struct fsxattr with fsx_cowextsize,
+ * disable the only other source (so far) of struct fsxattr.  Thus,
+ * build with the internal definition of struct fsxattr, which has
+ * fsx_cowextsize.
+ */
+#include "platform_defs.h"
+#include "command.h"
+#include "init.h"
+#include "io.h"
+#include "input.h"
+#include "path.h"
+
+static cmdinfo_t cowextsize_cmd;
+static long cowextsize;
+
+static void
+cowextsize_help(void)
+{
+	printf(_(
+"\n"
+" report or modify preferred CoW extent size (in bytes) for the current path\n"
+"\n"
+" -R -- recursively descend (useful when current path is a directory)\n"
+" -D -- recursively descend, only modifying cowextsize on directories\n"
+"\n"));
+}
+
+static int
+get_cowextsize(const char *path, int fd)
+{
+	struct fsxattr	fsx;
+
+	if ((xfsctl(path, fd, XFS_IOC_FSGETXATTR, &fsx)) < 0) {
+		printf("%s: XFS_IOC_FSGETXATTR %s: %s\n",
+			progname, path, strerror(errno));
+		return 0;
+	}
+	printf("[%u] %s\n", fsx.fsx_cowextsize, path);
+	return 0;
+}
+
+static int
+set_cowextsize(const char *path, int fd, long extsz)
+{
+	struct fsxattr	fsx;
+	struct stat64	stat;
+
+	if (fstat64(fd, &stat) < 0) {
+		perror("fstat64");
+		return 0;
+	}
+	if ((xfsctl(path, fd, XFS_IOC_FSGETXATTR, &fsx)) < 0) {
+		printf("%s: XFS_IOC_FSGETXATTR %s: %s\n",
+			progname, path, strerror(errno));
+		return 0;
+	}
+
+	if (S_ISREG(stat.st_mode) || S_ISDIR(stat.st_mode)) {
+		fsx.fsx_xflags |= FS_XFLAG_COWEXTSIZE;
+	} else {
+		printf(_("invalid target file type - file %s\n"), path);
+		return 0;
+	}
+	fsx.fsx_cowextsize = extsz;
+
+	if ((xfsctl(path, fd, XFS_IOC_FSSETXATTR, &fsx)) < 0) {
+		printf("%s: XFS_IOC_FSSETXATTR %s: %s\n",
+			progname, path, strerror(errno));
+		return 0;
+	}
+
+	return 0;
+}
+
+static int
+get_cowextsize_callback(
+	const char		*path,
+	const struct stat	*stat,
+	int			status,
+	struct FTW		*data)
+{
+	int			fd;
+
+	if (recurse_dir && !S_ISDIR(stat->st_mode))
+		return 0;
+
+	fd = open(path, O_RDONLY);
+	if (fd < 0) {
+		fprintf(stderr, _("%s: cannot open %s: %s\n"),
+			progname, path, strerror(errno));
+	} else {
+		get_cowextsize(path, fd);
+		close(fd);
+	}
+	return 0;
+}
+
+static int
+set_cowextsize_callback(
+	const char		*path,
+	const struct stat	*stat,
+	int			status,
+	struct FTW		*data)
+{
+	int			fd;
+
+	if (recurse_dir && !S_ISDIR(stat->st_mode))
+		return 0;
+
+	fd = open(path, O_RDONLY);
+	if (fd < 0) {
+		fprintf(stderr, _("%s: cannot open %s: %s\n"),
+			progname, path, strerror(errno));
+	} else {
+		set_cowextsize(path, fd, cowextsize);
+		close(fd);
+	}
+	return 0;
+}
+
+static int
+cowextsize_f(
+	int		argc,
+	char		**argv)
+{
+	size_t			blocksize, sectsize;
+	int			c;
+
+	recurse_all = recurse_dir = 0;
+	init_cvtnum(&blocksize, &sectsize);
+	while ((c = getopt(argc, argv, "DR")) != EOF) {
+		switch (c) {
+		case 'D':
+			recurse_all = 0;
+			recurse_dir = 1;
+			break;
+		case 'R':
+			recurse_all = 1;
+			recurse_dir = 0;
+			break;
+		default:
+			return command_usage(&cowextsize_cmd);
+		}
+	}
+
+	if (optind < argc) {
+		cowextsize = (long)cvtnum(blocksize, sectsize, argv[optind]);
+		if (cowextsize < 0) {
+			printf(_("non-numeric cowextsize argument -- %s\n"),
+				argv[optind]);
+			return 0;
+		}
+	} else {
+		cowextsize = -1;
+	}
+
+	if (recurse_all || recurse_dir)
+		nftw(file->name, (cowextsize >= 0) ?
+			set_cowextsize_callback : get_cowextsize_callback,
+			100, FTW_PHYS | FTW_MOUNT | FTW_DEPTH);
+	else if (cowextsize >= 0)
+		set_cowextsize(file->name, file->fd, cowextsize);
+	else
+		get_cowextsize(file->name, file->fd);
+	return 0;
+}
+
+void
+cowextsize_init(void)
+{
+	cowextsize_cmd.name = "cowextsize";
+	cowextsize_cmd.cfunc = cowextsize_f;
+	cowextsize_cmd.args = _("[-D | -R] [cowextsize]");
+	cowextsize_cmd.argmin = 0;
+	cowextsize_cmd.argmax = -1;
+	cowextsize_cmd.flags = CMD_NOMAP_OK;
+	cowextsize_cmd.oneline =
+		_("get/set preferred CoW extent size (in bytes) for the open file");
+	cowextsize_cmd.help = cowextsize_help;
+
+	add_command(&cowextsize_cmd);
+}
diff --git a/io/init.c b/io/init.c
index efe7390..6b88cc6 100644
--- a/io/init.c
+++ b/io/init.c
@@ -85,6 +85,7 @@ init_commands(void)
 	sync_range_init();
 	truncate_init();
 	reflink_init();
+	cowextsize_init();
 }
 
 static int
diff --git a/io/io.h b/io/io.h
index 2bc7ac4..4264e4d 100644
--- a/io/io.h
+++ b/io/io.h
@@ -169,3 +169,9 @@ extern void		readdir_init(void);
 #endif
 
 extern void		reflink_init(void);
+
+#ifdef HAVE_FSXATTR_COWEXTSIZE
+extern void		cowextsize_init(void);
+#else
+#define cowextsize_init()	do { } while (0)
+#endif
diff --git a/io/open.c b/io/open.c
index a5d465a..9a3563c 100644
--- a/io/open.c
+++ b/io/open.c
@@ -125,6 +125,9 @@ stat_f(
 		printxattr(fsx.fsx_xflags, verbose, 0, file->name, 1, 1);
 		printf(_("fsxattr.projid = %u\n"), fsx.fsx_projid);
 		printf(_("fsxattr.extsize = %u\n"), fsx.fsx_extsize);
+#if defined HAVE_FSXATTR_COWEXTSIZE
+		printf(_("fsxattr.cowextsize = %u\n"), fsx.fsx_cowextsize);
+#endif
 		printf(_("fsxattr.nextents = %u\n"), fsx.fsx_nextents);
 		printf(_("fsxattr.naextents = %u\n"), fsxa.fsx_nextents);
 	}
diff --git a/m4/package_libcdev.m4 b/m4/package_libcdev.m4
index 7a847e9..45954c2 100644
--- a/m4/package_libcdev.m4
+++ b/m4/package_libcdev.m4
@@ -265,3 +265,29 @@ AC_DEFUN([AC_HAVE_MREMAP],
        )
     AC_SUBST(have_mremap)
   ])
+
+#
+# Check if we have a struct fsxattr with a fsx_cowextsize field.
+# If linux/fs.h has a struct with that field, then we're ok.
+# If we can't find fsxattr in linux/fs.h at all, the internal
+# definitions provide it, and we're ok.
+#
+# The only way we won't have this is if the kernel headers don't
+# have the field.
+#
+AC_DEFUN([AC_HAVE_FSXATTR_COWEXTSIZE],
+  [ AM_CONDITIONAL([INTERNAL_FSXATTR], [test "x$enable_internal_fsxattr" = xyes])
+    AM_COND_IF([INTERNAL_FSXATTR],
+    [have_fsxattr_cowextsize=yes],
+    [ AC_CHECK_TYPE(struct fsxattr,
+	  [AC_CHECK_MEMBER(struct fsxattr.fsx_cowextsize,
+		  have_fsxattr_cowextsize=yes,
+		  have_fsxattr_cowextsize=no,
+		  [#include <linux/fs.h>]
+          )],
+	  have_fsxattr_cowextsize=yes,
+	  [#include <linux/fs.h>]
+      )
+    ])
+    AC_SUBST(have_fsxattr_cowextsize)
+  ])
diff --git a/man/man8/xfs_io.8 b/man/man8/xfs_io.8
index d089524..2365550 100644
--- a/man/man8/xfs_io.8
+++ b/man/man8/xfs_io.8
@@ -283,6 +283,22 @@ The
 should be specified in bytes, or using one of the usual units suffixes
 (k, m, g, b, etc). The extent size is always reported in units of bytes.
 .TP
+.BI "cowextsize [ \-R | \-D ] [ " value " ]"
+Display and/or modify the preferred copy-on-write extent size used
+when allocating space for the currently open file. If the
+.B \-R
+option is specified, a recursive descent is performed
+for all directory entries below the currently open file
+.RB ( \-D
+can be used to restrict the output to directories only).
+If the target file is a directory, then the inherited CoW extent size
+is set for that directory (new files created in that directory
+inherit that CoW extent size).
+The
+.I value
+should be specified in bytes, or using one of the usual units suffixes
+(k, m, g, b, etc). The extent size is always reported in units of bytes.
+.TP
 .BI "allocsp " size " 0"
 Sets the size of the file to
 .I size

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 53/71] xfs_io: add refcount+bmap error injection types

[Darrick J. Wong]

Add refcount and bmap deferred finish to the types of errors we can
inject.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 io/inject.c |    8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)


diff --git a/io/inject.c b/io/inject.c
index 16ac925..56642b8 100644
--- a/io/inject.c
+++ b/io/inject.c
@@ -78,7 +78,13 @@ error_tag(char *name)
 		{ XFS_ERRTAG_FREE_EXTENT,		"free_extent" },
 #define XFS_ERRTAG_RMAP_FINISH_ONE			23
 		{ XFS_ERRTAG_RMAP_FINISH_ONE,		"rmap_finish_one" },
-#define XFS_ERRTAG_MAX                                  24
+#define XFS_ERRTAG_REFCOUNT_CONTINUE_UPDATE		24
+		{ XFS_ERRTAG_REFCOUNT_CONTINUE_UPDATE,	"refcount_continue_update" },
+#define XFS_ERRTAG_REFCOUNT_FINISH_ONE			25
+		{ XFS_ERRTAG_REFCOUNT_FINISH_ONE,	"refcount_finish_one" },
+#define XFS_ERRTAG_BMAP_FINISH_ONE			26
+		{ XFS_ERRTAG_BMAP_FINISH_ONE,		"bmap_finish_one" },
+#define XFS_ERRTAG_MAX                                  27
 		{ XFS_ERRTAG_MAX,			NULL }
 	};
 	int	count;

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 54/71] xfs_logprint: support cowextsize reporting in log contents

[Darrick J. Wong]

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 logprint/log_misc.c      |    4 ++++
 logprint/log_print_all.c |    4 ++++
 2 files changed, 8 insertions(+)


diff --git a/logprint/log_misc.c b/logprint/log_misc.c
index 479fc14..f6488d9 100644
--- a/logprint/log_misc.c
+++ b/logprint/log_misc.c
@@ -513,6 +513,10 @@ xlog_print_trans_inode_core(
 	   ip->di_dmstate);
     printf(_("flags 0x%x gen 0x%x\n"),
 	   ip->di_flags, ip->di_gen);
+    if (ip->di_version == 3) {
+        printf(_("flags2 0x%llx cowextsize 0x%x\n"),
+            (unsigned long long)ip->di_flags2, ip->di_cowextsize);
+    }
 }
 
 void
diff --git a/logprint/log_print_all.c b/logprint/log_print_all.c
index 0fe354b..46952c4 100644
--- a/logprint/log_print_all.c
+++ b/logprint/log_print_all.c
@@ -272,6 +272,10 @@ xlog_recover_print_inode_core(
 	     "gen:%d\n"),
 	       (int)di->di_forkoff, di->di_dmevmask, (int)di->di_dmstate,
 	       (int)di->di_flags, di->di_gen);
+	if (di->di_version == 3) {
+		printf(_("flags2 0x%llx cowextsize 0x%x\n"),
+			(unsigned long long)di->di_flags2, di->di_cowextsize);
+	}
 }
 
 STATIC void

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 55/71] xfs_logprint: support refcount redo items

[Darrick J. Wong]

Print reference count update redo items.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 logprint/log_misc.c      |   11 +++
 logprint/log_print_all.c |   12 ++++
 logprint/log_redo.c      |  150 ++++++++++++++++++++++++++++++++++++++++++++++
 logprint/logprint.h      |    5 ++
 4 files changed, 178 insertions(+)


diff --git a/logprint/log_misc.c b/logprint/log_misc.c
index f6488d9..5389b72 100644
--- a/logprint/log_misc.c
+++ b/logprint/log_misc.c
@@ -1008,6 +1008,17 @@ xlog_print_record(
 					be32_to_cpu(op_head->oh_len));
 			break;
 		    }
+		    case XFS_LI_CUI: {
+			skip = xlog_print_trans_cui(&ptr,
+					be32_to_cpu(op_head->oh_len),
+					continued);
+			break;
+		    }
+		    case XFS_LI_CUD: {
+			skip = xlog_print_trans_cud(&ptr,
+					be32_to_cpu(op_head->oh_len));
+			break;
+		    }
 		    case XFS_LI_QUOTAOFF: {
 			skip = xlog_print_trans_qoff(&ptr,
 					be32_to_cpu(op_head->oh_len));
diff --git a/logprint/log_print_all.c b/logprint/log_print_all.c
index 46952c4..eb3e326 100644
--- a/logprint/log_print_all.c
+++ b/logprint/log_print_all.c
@@ -418,6 +418,12 @@ xlog_recover_print_logitem(
 	case XFS_LI_RUI:
 		xlog_recover_print_rui(item);
 		break;
+	case XFS_LI_CUD:
+		xlog_recover_print_cud(item);
+		break;
+	case XFS_LI_CUI:
+		xlog_recover_print_cui(item);
+		break;
 	case XFS_LI_DQUOT:
 		xlog_recover_print_dquot(item);
 		break;
@@ -458,6 +464,12 @@ xlog_recover_print_item(
 	case XFS_LI_RUI:
 		printf("RUI");
 		break;
+	case XFS_LI_CUD:
+		printf("CUD");
+		break;
+	case XFS_LI_CUI:
+		printf("CUI");
+		break;
 	case XFS_LI_DQUOT:
 		printf("DQ ");
 		break;
diff --git a/logprint/log_redo.c b/logprint/log_redo.c
index add0764..6a2e30a 100644
--- a/logprint/log_redo.c
+++ b/logprint/log_redo.c
@@ -380,3 +380,153 @@ xlog_recover_print_rud(
 	f = item->ri_buf[0].i_addr;
 	xlog_print_trans_rud(&f, sizeof(struct xfs_rud_log_format));
 }
+
+/* Reference Count Update Items */
+
+static int
+xfs_cui_copy_format(
+	char			  *buf,
+	uint			  len,
+	struct xfs_cui_log_format *dst_fmt,
+	int			  continued)
+{
+	uint nextents = ((struct xfs_cui_log_format *)buf)->cui_nextents;
+	uint dst_len = sizeof(struct xfs_cui_log_format) +
+			(nextents - 1) * sizeof(struct xfs_phys_extent);
+
+	if (len == dst_len || continued) {
+		memcpy((char *)dst_fmt, buf, len);
+		return 0;
+	}
+	fprintf(stderr, _("%s: bad size of CUI format: %u; expected %u; nextents = %u\n"),
+		progname, len, dst_len, nextents);
+	return 1;
+}
+
+int
+xlog_print_trans_cui(
+	char			**ptr,
+	uint			src_len,
+	int			continued)
+{
+	struct xfs_cui_log_format	*src_f, *f = NULL;
+	uint			dst_len;
+	uint			nextents;
+	struct xfs_phys_extent	*ex;
+	int			i;
+	int			error = 0;
+	int			core_size;
+
+	core_size = offsetof(struct xfs_cui_log_format, cui_extents);
+
+	/*
+	 * memmove to ensure 8-byte alignment for the long longs in
+	 * struct xfs_cui_log_format structure
+	 */
+	src_f = malloc(src_len);
+	if (src_f == NULL) {
+		fprintf(stderr, _("%s: %s: malloc failed\n"),
+			progname, __func__);
+		exit(1);
+	}
+	memmove((char*)src_f, *ptr, src_len);
+	*ptr += src_len;
+
+	/* convert to native format */
+	nextents = src_f->cui_nextents;
+	dst_len = sizeof(struct xfs_cui_log_format) +
+			(nextents - 1) * sizeof(struct xfs_phys_extent);
+
+	if (continued && src_len < core_size) {
+		printf(_("CUI: Not enough data to decode further\n"));
+		error = 1;
+		goto error;
+	}
+
+	f = malloc(dst_len);
+	if (f == NULL) {
+		fprintf(stderr, _("%s: %s: malloc failed\n"),
+			progname, __func__);
+		exit(1);
+	}
+	if (xfs_cui_copy_format((char *)src_f, src_len, f, continued)) {
+		error = 1;
+		goto error;
+	}
+
+	printf(_("CUI:  #regs: %d	num_extents: %d  id: 0x%llx\n"),
+		f->cui_size, f->cui_nextents, (unsigned long long)f->cui_id);
+
+	if (continued) {
+		printf(_("CUI extent data skipped (CONTINUE set, no space)\n"));
+		goto error;
+	}
+
+	ex = f->cui_extents;
+	for (i=0; i < f->cui_nextents; i++) {
+		printf("(s: 0x%llx, l: %d, f: 0x%x) ",
+			(unsigned long long)ex->pe_startblock, ex->pe_len,
+			ex->pe_flags);
+		printf("\n");
+		ex++;
+	}
+error:
+	free(src_f);
+	free(f);
+	return error;
+}
+
+void
+xlog_recover_print_cui(
+	struct xlog_recover_item	*item)
+{
+	char				*src_f;
+	uint				src_len;
+
+	src_f = item->ri_buf[0].i_addr;
+	src_len = item->ri_buf[0].i_len;
+
+	xlog_print_trans_cui(&src_f, src_len, 0);
+}
+
+int
+xlog_print_trans_cud(
+	char				**ptr,
+	uint				len)
+{
+	struct xfs_cud_log_format	*f;
+	struct xfs_cud_log_format	lbuf;
+
+	/* size without extents at end */
+	uint core_size = sizeof(struct xfs_cud_log_format);
+
+	/*
+	 * memmove to ensure 8-byte alignment for the long longs in
+	 * xfs_efd_log_format_t structure
+	 */
+	memmove(&lbuf, *ptr, MIN(core_size, len));
+	f = &lbuf;
+	*ptr += len;
+	if (len >= core_size) {
+		printf(_("CUD:  #regs: %d	                 id: 0x%llx\n"),
+			f->cud_size,
+			(unsigned long long)f->cud_cui_id);
+
+		/* don't print extents as they are not used */
+
+		return 0;
+	} else {
+		printf(_("CUD: Not enough data to decode further\n"));
+		return 1;
+	}
+}
+
+void
+xlog_recover_print_cud(
+	struct xlog_recover_item	*item)
+{
+	char				*f;
+
+	f = item->ri_buf[0].i_addr;
+	xlog_print_trans_cud(&f, sizeof(struct xfs_cud_log_format));
+}
diff --git a/logprint/logprint.h b/logprint/logprint.h
index 0c03c08..a1115e2 100644
--- a/logprint/logprint.h
+++ b/logprint/logprint.h
@@ -56,4 +56,9 @@ extern void xlog_recover_print_rui(struct xlog_recover_item *item);
 extern int xlog_print_trans_rud(char **ptr, uint len);
 extern void xlog_recover_print_rud(struct xlog_recover_item *item);
 
+extern int xlog_print_trans_cui(char **ptr, uint src_len, int continued);
+extern void xlog_recover_print_cui(struct xlog_recover_item *item);
+extern int xlog_print_trans_cud(char **ptr, uint len);
+extern void xlog_recover_print_cud(struct xlog_recover_item *item);
+
 #endif	/* LOGPRINT_H */

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 56/71] xfs_logprint: support bmap redo items

[Darrick J. Wong]

Print block mapping update redo items.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 logprint/log_misc.c      |   11 +++
 logprint/log_print_all.c |   12 ++++
 logprint/log_redo.c      |  151 ++++++++++++++++++++++++++++++++++++++++++++++
 logprint/logprint.h      |    5 ++
 4 files changed, 179 insertions(+)


diff --git a/logprint/log_misc.c b/logprint/log_misc.c
index 5389b72..c10a1d1 100644
--- a/logprint/log_misc.c
+++ b/logprint/log_misc.c
@@ -1019,6 +1019,17 @@ xlog_print_record(
 					be32_to_cpu(op_head->oh_len));
 			break;
 		    }
+		    case XFS_LI_BUI: {
+			skip = xlog_print_trans_bui(&ptr,
+					be32_to_cpu(op_head->oh_len),
+					continued);
+			break;
+		    }
+		    case XFS_LI_BUD: {
+			skip = xlog_print_trans_bud(&ptr,
+					be32_to_cpu(op_head->oh_len));
+			break;
+		    }
 		    case XFS_LI_QUOTAOFF: {
 			skip = xlog_print_trans_qoff(&ptr,
 					be32_to_cpu(op_head->oh_len));
diff --git a/logprint/log_print_all.c b/logprint/log_print_all.c
index eb3e326..f49316e 100644
--- a/logprint/log_print_all.c
+++ b/logprint/log_print_all.c
@@ -424,6 +424,12 @@ xlog_recover_print_logitem(
 	case XFS_LI_CUI:
 		xlog_recover_print_cui(item);
 		break;
+	case XFS_LI_BUD:
+		xlog_recover_print_bud(item);
+		break;
+	case XFS_LI_BUI:
+		xlog_recover_print_bui(item);
+		break;
 	case XFS_LI_DQUOT:
 		xlog_recover_print_dquot(item);
 		break;
@@ -470,6 +476,12 @@ xlog_recover_print_item(
 	case XFS_LI_CUI:
 		printf("CUI");
 		break;
+	case XFS_LI_BUD:
+		printf("BUD");
+		break;
+	case XFS_LI_BUI:
+		printf("BUI");
+		break;
 	case XFS_LI_DQUOT:
 		printf("DQ ");
 		break;
diff --git a/logprint/log_redo.c b/logprint/log_redo.c
index 6a2e30a..dcca427 100644
--- a/logprint/log_redo.c
+++ b/logprint/log_redo.c
@@ -530,3 +530,154 @@ xlog_recover_print_cud(
 	f = item->ri_buf[0].i_addr;
 	xlog_print_trans_cud(&f, sizeof(struct xfs_cud_log_format));
 }
+
+/* Block Mapping Update Items */
+
+static int
+xfs_bui_copy_format(
+	char			  *buf,
+	uint			  len,
+	struct xfs_bui_log_format *dst_fmt,
+	int			  continued)
+{
+	uint nextents = ((struct xfs_bui_log_format *)buf)->bui_nextents;
+	uint dst_len = sizeof(struct xfs_bui_log_format) +
+			(nextents - 1) * sizeof(struct xfs_map_extent);
+
+	if (len == dst_len || continued) {
+		memcpy((char *)dst_fmt, buf, len);
+		return 0;
+	}
+	fprintf(stderr, _("%s: bad size of BUI format: %u; expected %u; nextents = %u\n"),
+		progname, len, dst_len, nextents);
+	return 1;
+}
+
+int
+xlog_print_trans_bui(
+	char			**ptr,
+	uint			src_len,
+	int			continued)
+{
+	struct xfs_bui_log_format	*src_f, *f = NULL;
+	uint			dst_len;
+	uint			nextents;
+	struct xfs_map_extent	*ex;
+	int			i;
+	int			error = 0;
+	int			core_size;
+
+	core_size = offsetof(struct xfs_bui_log_format, bui_extents);
+
+	/*
+	 * memmove to ensure 8-byte alignment for the long longs in
+	 * struct xfs_bui_log_format structure
+	 */
+	src_f = malloc(src_len);
+	if (src_f == NULL) {
+		fprintf(stderr, _("%s: %s: malloc failed\n"),
+			progname, __func__);
+		exit(1);
+	}
+	memmove((char*)src_f, *ptr, src_len);
+	*ptr += src_len;
+
+	/* convert to native format */
+	nextents = src_f->bui_nextents;
+	dst_len = sizeof(struct xfs_bui_log_format) +
+			(nextents - 1) * sizeof(struct xfs_map_extent);
+
+	if (continued && src_len < core_size) {
+		printf(_("BUI: Not enough data to decode further\n"));
+		error = 1;
+		goto error;
+	}
+
+	f = malloc(dst_len);
+	if (f == NULL) {
+		fprintf(stderr, _("%s: %s: malloc failed\n"),
+			progname, __func__);
+		exit(1);
+	}
+	if (xfs_bui_copy_format((char *)src_f, src_len, f, continued)) {
+		error = 1;
+		goto error;
+	}
+
+	printf(_("BUI:  #regs: %d	num_extents: %d  id: 0x%llx\n"),
+		f->bui_size, f->bui_nextents, (unsigned long long)f->bui_id);
+
+	if (continued) {
+		printf(_("BUI extent data skipped (CONTINUE set, no space)\n"));
+		goto error;
+	}
+
+	ex = f->bui_extents;
+	for (i=0; i < f->bui_nextents; i++) {
+		printf("(s: 0x%llx, l: %d, own: %lld, off: %llu, f: 0x%x) ",
+			(unsigned long long)ex->me_startblock, ex->me_len,
+			(long long)ex->me_owner,
+			(unsigned long long)ex->me_startoff, ex->me_flags);
+		printf("\n");
+		ex++;
+	}
+error:
+	free(src_f);
+	free(f);
+	return error;
+}
+
+void
+xlog_recover_print_bui(
+	struct xlog_recover_item	*item)
+{
+	char				*src_f;
+	uint				src_len;
+
+	src_f = item->ri_buf[0].i_addr;
+	src_len = item->ri_buf[0].i_len;
+
+	xlog_print_trans_bui(&src_f, src_len, 0);
+}
+
+int
+xlog_print_trans_bud(
+	char				**ptr,
+	uint				len)
+{
+	struct xfs_bud_log_format	*f;
+	struct xfs_bud_log_format	lbuf;
+
+	/* size without extents at end */
+	uint core_size = sizeof(struct xfs_bud_log_format);
+
+	/*
+	 * memmove to ensure 8-byte alignment for the long longs in
+	 * xfs_efd_log_format_t structure
+	 */
+	memmove(&lbuf, *ptr, MIN(core_size, len));
+	f = &lbuf;
+	*ptr += len;
+	if (len >= core_size) {
+		printf(_("BUD:  #regs: %d	                 id: 0x%llx\n"),
+			f->bud_size,
+			(unsigned long long)f->bud_bui_id);
+
+		/* don't print extents as they are not used */
+
+		return 0;
+	} else {
+		printf(_("BUD: Not enough data to decode further\n"));
+		return 1;
+	}
+}
+
+void
+xlog_recover_print_bud(
+	struct xlog_recover_item	*item)
+{
+	char				*f;
+
+	f = item->ri_buf[0].i_addr;
+	xlog_print_trans_bud(&f, sizeof(struct xfs_bud_log_format));
+}
diff --git a/logprint/logprint.h b/logprint/logprint.h
index a1115e2..81feff3 100644
--- a/logprint/logprint.h
+++ b/logprint/logprint.h
@@ -61,4 +61,9 @@ extern void xlog_recover_print_cui(struct xlog_recover_item *item);
 extern int xlog_print_trans_cud(char **ptr, uint len);
 extern void xlog_recover_print_cud(struct xlog_recover_item *item);
 
+extern int xlog_print_trans_bui(char **ptr, uint src_len, int continued);
+extern void xlog_recover_print_bui(struct xlog_recover_item *item);
+extern int xlog_print_trans_bud(char **ptr, uint len);
+extern void xlog_recover_print_bud(struct xlog_recover_item *item);
+
 #endif	/* LOGPRINT_H */

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 57/71] man: document the reflink inode flag in fsxattr

[Darrick J. Wong]

Document the new inode flag in struct fsxattr for reflink.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 man/man3/xfsctl.3 |   10 ++++++++++
 1 file changed, 10 insertions(+)


diff --git a/man/man3/xfsctl.3 b/man/man3/xfsctl.3
index 9e7f138..ef6c992 100644
--- a/man/man3/xfsctl.3
+++ b/man/man3/xfsctl.3
@@ -230,6 +230,16 @@ If the filesystem lives on directly accessible persistent memory, reads and
 writes to this file will go straight to the persistent memory, bypassing the
 page cache.
 .TP
+.SM "Bit 16 (0x10000) \- XFS_XFLAG_REFLINK"
+This file is sharing or has shared blocks with another file.
+This flag can be set by reflinking or deduping blocks with another file
+and cleared by fallocating the entire file to pre-copy all shared extents.
+A file cannot have
+.BR XFS_XFLAG_REFLINK
+and
+.BR XFS_XFLAG_DAX
+set at the same time, that is to say that DAX files cannot share blocks.
+.TP
 .SM "Bit 31 (0x80000000) \- XFS_XFLAG_HASATTR"
 The file has extended attributes associated with it.
 .RE

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 58/71] man: document the inode cowextsize flags & fields

[Darrick J. Wong]

Document the new copy-on-write extent size fields and inode flags
available in struct fsxattr.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 man/man3/xfsctl.3 |   24 +++++++++++++++++++++++-
 1 file changed, 23 insertions(+), 1 deletion(-)


diff --git a/man/man3/xfsctl.3 b/man/man3/xfsctl.3
index ef6c992..77506aa 100644
--- a/man/man3/xfsctl.3
+++ b/man/man3/xfsctl.3
@@ -150,6 +150,15 @@ value returned indicates that a preferred extent size was previously
 set on the file, a
 .B fsx_extsize
 of zero indicates that the defaults for that filesystem will be used.
+A
+.B fsx_cowextsize
+value returned indicates that a preferred copy on write extent size was
+previously set on the file, whereas a
+.B fsx_cowextsize
+of zero indicates that the defaults for that filesystem will be used.
+The current default for
+.B fsx_cowextsize
+is 128 blocks.
 Currently the meaningful bits for the
 .B fsx_xflags
 field are:
@@ -240,6 +249,15 @@ and
 .BR XFS_XFLAG_DAX
 set at the same time, that is to say that DAX files cannot share blocks.
 .TP
+.SM "Bit 17 (0x20000) \- XFS_XFLAG_COWEXTSIZE"
+Copy on Write Extent size bit - if a CoW extent size value is set on the file,
+the allocator will allocate extents for staging a copy on write operation
+in multiples of the set size for this file (see
+.B XFS_IOC_FSSETXATTR
+below).
+If the CoW extent size is set on a directory, then new file and directories
+created in the directory will inherit the parent's CoW extent size value.
+.TP
 .SM "Bit 31 (0x80000000) \- XFS_XFLAG_HASATTR"
 The file has extended attributes associated with it.
 .RE
@@ -261,7 +279,8 @@ The final argument points to a variable of type
 .BR "struct fsxattr" ,
 but only the following fields are used in this call:
 .BR fsx_xflags ,
-.B fsx_extsize
+.BR fsx_extsize ,
+.BR fsx_cowextsize ,
 and
 .BR fsx_projid .
 The
@@ -271,6 +290,9 @@ when the file is empty, except in the case of a directory where
 the extent size can be set at any time (this value is only used
 for regular file allocations, so should only be set on a directory
 in conjunction with the XFS_XFLAG_EXTSZINHERIT flag).
+The copy on write extent size,
+.BR fsx_cowextsize ,
+can be set at any time.
 
 .TP
 .B XFS_IOC_GETBMAP

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 59/71] xfs_repair: fix get_agino_buf to avoid corrupting inodes

[Darrick J. Wong]

The inode buffering code tries to read inodes in units of chunks,
which are the larger of 8K or 1 FSB.  Each chunk gets its own xfs_buf,
which means that get_agino_buf must calculate the disk address of the
chunk and feed that to libxfs_readbuf in order to find the inode data
correctly.  The current code simply grabs the chunk for the start
inode and indexes from that, which corrupts memory because the start
inode and the target inode could be in different inode chunks.  That
causes the assert in rmap.c to blow when we clear the reflink flag.

(Also fix some minor errors in the debugging printfs.)

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/rdwr.c   |    8 +++---
 repair/dinode.c |   73 +++++++++++++++++++++++++++++++++----------------------
 repair/dinode.h |   12 +++++----
 3 files changed, 54 insertions(+), 39 deletions(-)


diff --git a/libxfs/rdwr.c b/libxfs/rdwr.c
index 533a064..9fcc319 100644
--- a/libxfs/rdwr.c
+++ b/libxfs/rdwr.c
@@ -1038,9 +1038,9 @@ libxfs_readbufr_map(struct xfs_buftarg *btp, struct xfs_buf *bp, int flags)
 	if (!error)
 		bp->b_flags |= LIBXFS_B_UPTODATE;
 #ifdef IO_DEBUG
-	printf("%lx: %s: read %u bytes, error %d, blkno=0x%llx(0x%llx), %p\n",
-		pthread_self(), __FUNCTION__, , error,
-		(long long)LIBXFS_BBTOOFF64(blkno), (long long)blkno, bp);
+	printf("%lx: %s: read %lu bytes, error %d, blkno=%llu(%llu), %p\n",
+		pthread_self(), __FUNCTION__, buf - (char *)bp->b_addr, error,
+		(long long)LIBXFS_BBTOOFF64(bp->b_bn), (long long)bp->b_bn, bp);
 #endif
 	return error;
 }
@@ -1070,7 +1070,7 @@ libxfs_readbuf_map(struct xfs_buftarg *btp, struct xfs_buf_map *map, int nmaps,
 	if (!error)
 		libxfs_readbuf_verify(bp, ops);
 
-#ifdef IO_DEBUG
+#ifdef IO_DEBUGX
 	printf("%lx: %s: read %lu bytes, error %d, blkno=%llu(%llu), %p\n",
 		pthread_self(), __FUNCTION__, buf - (char *)bp->b_addr, error,
 		(long long)LIBXFS_BBTOOFF64(bp->b_bn), (long long)bp->b_bn, bp);
diff --git a/repair/dinode.c b/repair/dinode.c
index 512a668..16e0a06 100644
--- a/repair/dinode.c
+++ b/repair/dinode.c
@@ -847,43 +847,58 @@ scan_bmbt_reclist(
 }
 
 /*
- * these two are meant for routines that read and work with inodes
- * one at a time where the inodes may be in any order (like walking
- * the unlinked lists to look for inodes).  the caller is responsible
- * for writing/releasing the buffer.
+ * Grab the buffer backing an inode.  This is meant for routines that
+ * work with inodes one at a time in any order (like walking the
+ * unlinked lists to look for inodes).  The caller is responsible for
+ * writing/releasing the buffer.
  */
-xfs_buf_t *
-get_agino_buf(xfs_mount_t	 *mp,
-		xfs_agnumber_t	agno,
-		xfs_agino_t	agino,
-		xfs_dinode_t	**dipp)
+struct xfs_buf *
+get_agino_buf(
+	struct xfs_mount	*mp,
+	xfs_agnumber_t		agno,
+	xfs_agino_t		agino,
+	struct xfs_dinode	**dipp)
 {
-	ino_tree_node_t *irec;
-	xfs_buf_t *bp;
-	int size;
-
-	if ((irec = find_inode_rec(mp, agno, agino)) == NULL)
-		return(NULL);
+	struct xfs_buf		*bp;
+	int			cluster_size;
+	int			ino_per_cluster;
+	xfs_agino_t		cluster_agino;
+	xfs_daddr_t		cluster_daddr;
+	xfs_daddr_t		cluster_blks;
 
-	size = MAX(1, XFS_FSB_TO_BB(mp,
+	/*
+	 * Inode buffers have been read into memory in inode_cluster_size
+	 * chunks (or one FSB).  To find the correct buffer for an inode,
+	 * we must find the buffer for its cluster, add the appropriate
+	 * offset, and return that.
+	 */
+	cluster_size = MAX(mp->m_inode_cluster_size, mp->m_sb.sb_blocksize);
+	ino_per_cluster = cluster_size / mp->m_sb.sb_inodesize;
+	cluster_agino = agino & ~(ino_per_cluster - 1);
+	cluster_blks = XFS_FSB_TO_DADDR(mp, MAX(1,
 			mp->m_inode_cluster_size >> mp->m_sb.sb_blocklog));
-	bp = libxfs_readbuf(mp->m_dev, XFS_AGB_TO_DADDR(mp, agno,
-		XFS_AGINO_TO_AGBNO(mp, irec->ino_startnum)), size, 0,
-		&xfs_inode_buf_ops);
+	cluster_daddr = XFS_AGB_TO_DADDR(mp, agno,
+			XFS_AGINO_TO_AGBNO(mp, cluster_agino));
+
+#ifdef XR_INODE_TRACE
+	printf("cluster_size %d ipc %d clusagino %d daddr %lld sectors %lld\n",
+		cluster_size, ino_per_cluster, cluster_agino, cluster_daddr,
+		cluster_blks);
+#endif
+
+	bp = libxfs_readbuf(mp->m_dev, cluster_daddr, cluster_blks,
+			0, &xfs_inode_buf_ops);
 	if (!bp) {
 		do_warn(_("cannot read inode (%u/%u), disk block %" PRIu64 "\n"),
-			agno, irec->ino_startnum,
-			XFS_AGB_TO_DADDR(mp, agno,
-				XFS_AGINO_TO_AGBNO(mp, irec->ino_startnum)));
-		return(NULL);
+			agno, cluster_agino, cluster_daddr);
+		return NULL;
 	}
 
-	*dipp = xfs_make_iptr(mp, bp, agino -
-		XFS_OFFBNO_TO_AGINO(mp, XFS_AGINO_TO_AGBNO(mp,
-						irec->ino_startnum),
-		0));
-
-	return(bp);
+	*dipp = xfs_make_iptr(mp, bp, agino - cluster_agino);
+	ASSERT(!xfs_sb_version_hascrc(&mp->m_sb) ||
+			XFS_AGINO_TO_INO(mp, agno, agino) ==
+			be64_to_cpu((*dipp)->di_ino));
+	return bp;
 }
 
 /*
diff --git a/repair/dinode.h b/repair/dinode.h
index 5aebf5b..61d0736 100644
--- a/repair/dinode.h
+++ b/repair/dinode.h
@@ -113,12 +113,12 @@ void
 check_uncertain_aginodes(xfs_mount_t	*mp,
 			xfs_agnumber_t	agno);
 
-xfs_buf_t *
-get_agino_buf(xfs_mount_t	*mp,
-		xfs_agnumber_t	agno,
-		xfs_agino_t	agino,
-		xfs_dinode_t	**dipp);
-
+struct xfs_buf *
+get_agino_buf(
+	struct xfs_mount	*mp,
+	xfs_agnumber_t		agno,
+	xfs_agino_t		agino,
+	struct xfs_dinode	**dipp);
 
 void dinode_bmbt_translation_init(void);
 char * get_forkname(int whichfork);

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 60/71] xfs_repair: check the existing refcount btree

[Darrick J. Wong]

Spot-check the refcount btree for obvious errors, and mark the
refcount btree blocks as such.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 repair/incore.h     |    3 +
 repair/scan.c       |  185 +++++++++++++++++++++++++++++++++++++++++++++++++++
 repair/xfs_repair.c |    2 +
 3 files changed, 189 insertions(+), 1 deletion(-)


diff --git a/repair/incore.h b/repair/incore.h
index bc0810b..b6c4b4f 100644
--- a/repair/incore.h
+++ b/repair/incore.h
@@ -106,7 +106,8 @@ typedef struct rt_extent_tree_node  {
 #define XR_E_INUSE_FS1	9	/* used by fs ag header or log (rmap btree) */
 #define XR_E_INO1	10	/* used by inodes (marked by rmap btree) */
 #define XR_E_FS_MAP1	11	/* used by fs space/inode maps (rmap btree) */
-#define XR_E_BAD_STATE	12
+#define XR_E_REFC	12	/* used by fs ag reference count btree */
+#define XR_E_BAD_STATE	13
 
 /* separate state bit, OR'ed into high (4th) bit of ex_state field */
 
diff --git a/repair/scan.c b/repair/scan.c
index 35a974a..c27f969 100644
--- a/repair/scan.c
+++ b/repair/scan.c
@@ -810,6 +810,9 @@ process_rmap_rec(
 		case XFS_RMAP_OWN_INODES:
 			set_bmap_ext(agno, b, blen, XR_E_INO1);
 			break;
+		case XFS_RMAP_OWN_REFC:
+			set_bmap_ext(agno, b, blen, XR_E_REFC);
+			break;
 		case XFS_RMAP_OWN_NULL:
 			/* still unknown */
 			break;
@@ -845,6 +848,14 @@ _("inode block (%d,%d-%d) mismatch in %s tree, state - %d,%" PRIx64 "\n"),
 			agno, b, b + blen - 1,
 			name, state, owner);
 		break;
+	case XR_E_REFC:
+		if (owner == XFS_RMAP_OWN_REFC)
+			break;
+		do_warn(
+_("AG refcount block (%d,%d-%d) mismatch in %s tree, state - %d,%" PRIx64 "\n"),
+			agno, b, b + blen - 1,
+			name, state, owner);
+		break;
 	case XR_E_INUSE:
 		if (owner >= 0 &&
 		    owner < mp->m_sb.sb_dblocks)
@@ -1161,6 +1172,167 @@ out:
 		rmap_avoid_check();
 }
 
+static void
+scan_refcbt(
+	struct xfs_btree_block	*block,
+	int			level,
+	xfs_agblock_t		bno,
+	xfs_agnumber_t		agno,
+	int			suspect,
+	int			isroot,
+	__uint32_t		magic,
+	void			*priv)
+{
+	const char		*name = "refcount";
+	int			i;
+	xfs_refcount_ptr_t	*pp;
+	struct xfs_refcount_rec	*rp;
+	int			hdr_errors = 0;
+	int			numrecs;
+	int			state;
+	xfs_agblock_t		lastblock = 0;
+
+	if (magic != XFS_REFC_CRC_MAGIC) {
+		name = "(unknown)";
+		hdr_errors++;
+		suspect++;
+		goto out;
+	}
+
+	if (be32_to_cpu(block->bb_magic) != magic) {
+		do_warn(_("bad magic # %#x in %s btree block %d/%d\n"),
+			be32_to_cpu(block->bb_magic), name, agno, bno);
+		hdr_errors++;
+		if (suspect)
+			goto out;
+	}
+
+	if (be16_to_cpu(block->bb_level) != level) {
+		do_warn(_("expected level %d got %d in %s btree block %d/%d\n"),
+			level, be16_to_cpu(block->bb_level), name, agno, bno);
+		hdr_errors++;
+		if (suspect)
+			goto out;
+	}
+
+	/* check for btree blocks multiply claimed */
+	state = get_bmap(agno, bno);
+	if (!(state == XR_E_UNKNOWN || state == XR_E_REFC))  {
+		set_bmap(agno, bno, XR_E_MULT);
+		do_warn(
+_("%s btree block claimed (state %d), agno %d, bno %d, suspect %d\n"),
+				name, state, agno, bno, suspect);
+		goto out;
+	}
+	set_bmap(agno, bno, XR_E_FS_MAP);
+
+	numrecs = be16_to_cpu(block->bb_numrecs);
+	if (level == 0) {
+		if (numrecs > mp->m_refc_mxr[0])  {
+			numrecs = mp->m_refc_mxr[0];
+			hdr_errors++;
+		}
+		if (isroot == 0 && numrecs < mp->m_refc_mnr[0])  {
+			numrecs = mp->m_refc_mnr[0];
+			hdr_errors++;
+		}
+
+		if (hdr_errors) {
+			do_warn(
+	_("bad btree nrecs (%u, min=%u, max=%u) in %s btree block %u/%u\n"),
+				be16_to_cpu(block->bb_numrecs),
+				mp->m_refc_mnr[0], mp->m_refc_mxr[0],
+				name, agno, bno);
+			suspect++;
+		}
+
+		rp = XFS_REFCOUNT_REC_ADDR(block, 1);
+		for (i = 0; i < numrecs; i++) {
+			xfs_agblock_t		b, end;
+			xfs_extlen_t		len;
+			xfs_nlink_t		nr;
+
+			b = be32_to_cpu(rp[i].rc_startblock);
+			len = be32_to_cpu(rp[i].rc_blockcount);
+			nr = be32_to_cpu(rp[i].rc_refcount);
+			end = b + len;
+
+			if (!verify_agbno(mp, agno, b)) {
+				do_warn(
+	_("invalid start block %u in record %u of %s btree block %u/%u\n"),
+					b, i, name, agno, bno);
+				continue;
+			}
+			if (len == 0 || !verify_agbno(mp, agno, end - 1)) {
+				do_warn(
+	_("invalid length %u in record %u of %s btree block %u/%u\n"),
+					len, i, name, agno, bno);
+				continue;
+			}
+
+			if (nr < 2 || nr > MAXREFCOUNT) {
+				do_warn(
+	_("invalid reference count %u in record %u of %s btree block %u/%u\n"),
+					nr, i, name, agno, bno);
+				continue;
+			}
+
+			if (b && b <= lastblock) {
+				do_warn(_(
+	"out-of-order %s btree record %d (%u %u) block %u/%u\n"),
+					name, i, b, len, agno, bno);
+			} else {
+				lastblock = b;
+			}
+
+			/* XXX: probably want to mark the reflinked areas? */
+		}
+		goto out;
+	}
+
+	/*
+	 * interior record
+	 */
+	pp = XFS_REFCOUNT_PTR_ADDR(block, 1, mp->m_refc_mxr[1]);
+
+	if (numrecs > mp->m_refc_mxr[1])  {
+		numrecs = mp->m_refc_mxr[1];
+		hdr_errors++;
+	}
+	if (isroot == 0 && numrecs < mp->m_refc_mnr[1])  {
+		numrecs = mp->m_refc_mnr[1];
+		hdr_errors++;
+	}
+
+	/*
+	 * don't pass bogus tree flag down further if this block
+	 * looked ok.  bail out if two levels in a row look bad.
+	 */
+	if (hdr_errors)  {
+		do_warn(
+	_("bad btree nrecs (%u, min=%u, max=%u) in %s btree block %u/%u\n"),
+			be16_to_cpu(block->bb_numrecs),
+			mp->m_refc_mnr[1], mp->m_refc_mxr[1],
+			name, agno, bno);
+		if (suspect)
+			goto out;
+		suspect++;
+	} else if (suspect) {
+		suspect = 0;
+	}
+
+	for (i = 0; i < numrecs; i++)  {
+		xfs_agblock_t		bno = be32_to_cpu(pp[i]);
+
+		if (bno != 0 && verify_agbno(mp, agno, bno)) {
+			scan_sbtree(bno, level, agno, suspect, scan_refcbt, 0,
+				    magic, priv, &xfs_refcountbt_buf_ops);
+		}
+	}
+out:
+	return;
+}
+
 /*
  * The following helpers are to help process and validate individual on-disk
  * inode btree records. We have two possible inode btrees with slightly
@@ -1951,6 +2123,19 @@ validate_agf(
 		}
 	}
 
+	if (xfs_sb_version_hasreflink(&mp->m_sb)) {
+		bno = be32_to_cpu(agf->agf_refcount_root);
+		if (bno != 0 && verify_agbno(mp, agno, bno)) {
+			scan_sbtree(bno,
+				    be32_to_cpu(agf->agf_refcount_level),
+				    agno, 0, scan_refcbt, 1, XFS_REFC_CRC_MAGIC,
+				    agcnts, &xfs_refcountbt_buf_ops);
+		} else  {
+			do_warn(_("bad agbno %u for refcntbt root, agno %d\n"),
+				bno, agno);
+		}
+	}
+
 	if (be32_to_cpu(agf->agf_freeblks) != agcnts->agffreeblks) {
 		do_warn(_("agf_freeblks %u, counted %u in ag %u\n"),
 			be32_to_cpu(agf->agf_freeblks), agcnts->agffreeblks, agno);
diff --git a/repair/xfs_repair.c b/repair/xfs_repair.c
index dc38ece..4d92b90 100644
--- a/repair/xfs_repair.c
+++ b/repair/xfs_repair.c
@@ -423,6 +423,8 @@ calc_mkfs(xfs_mount_t *mp)
 		fino_bno += min(2, mp->m_rmap_maxlevels); /* agfl blocks */
 		fino_bno++;
 	}
+	if (xfs_sb_version_hasreflink(&mp->m_sb))
+		fino_bno++;
 
 	/*
 	 * If the log is allocated in the first allocation group we need to

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 61/71] xfs_repair: handle multiple owners of data blocks

[Darrick J. Wong]

If reflink is enabled, don't freak out if there are multiple owners of
a given block; that's just a sign that each of those owners are
reflink files.

v2: owner and offset are unsigned types, so use those for inorder
comparison.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 repair/dinode.c |   66 +++++++++++++++++++++++++++++++++++++++++++++++++++++++
 repair/scan.c   |   42 ++++++++++++++++++++++++++++++++++-
 2 files changed, 107 insertions(+), 1 deletion(-)


diff --git a/repair/dinode.c b/repair/dinode.c
index 16e0a06..98afdc9 100644
--- a/repair/dinode.c
+++ b/repair/dinode.c
@@ -722,6 +722,9 @@ _("Fatal error: inode %" PRIu64 " - blkmap_set_ext(): %s\n"
 			 * checking each entry without setting the
 			 * block bitmap
 			 */
+			if (type == XR_INO_DATA &&
+			    xfs_sb_version_hasreflink(&mp->m_sb))
+				goto skip_dup;
 			if (search_dup_extent(agno, agbno, ebno)) {
 				do_warn(
 _("%s fork in ino %" PRIu64 " claims dup extent, "
@@ -731,6 +734,7 @@ _("%s fork in ino %" PRIu64 " claims dup extent, "
 					irec.br_blockcount);
 				goto done;
 			}
+skip_dup:
 			*tot += irec.br_blockcount;
 			continue;
 		}
@@ -770,6 +774,9 @@ _("%s fork in inode %" PRIu64 " claims metadata block %" PRIu64 "\n"),
 			case XR_E_INUSE:
 			case XR_E_MULT:
 				set_bmap_ext(agno, agbno, blen, XR_E_MULT);
+				if (type == XR_INO_DATA &&
+				    xfs_sb_version_hasreflink(&mp->m_sb))
+					break;
 				do_warn(
 _("%s fork in %s inode %" PRIu64 " claims used block %" PRIu64 "\n"),
 					forkname, ftype, ino, b);
@@ -2475,6 +2482,65 @@ _("bad (negative) size %" PRId64 " on inode %" PRIu64 "\n"),
 		}
 	}
 
+	/*
+	 * check that we only have valid flags2 set, and those that are set make
+	 * sense.
+	 */
+	if (dino->di_version >= 3) {
+		uint16_t flags = be16_to_cpu(dino->di_flags);
+		uint64_t flags2 = be64_to_cpu(dino->di_flags2);
+
+		if (flags2 & ~XFS_DIFLAG2_ANY) {
+			if (!uncertain) {
+				do_warn(
+	_("Bad flags2 set in inode %" PRIu64 "\n"),
+					lino);
+			}
+			flags2 &= XFS_DIFLAG2_ANY;
+		}
+
+		if ((flags2 & XFS_DIFLAG2_REFLINK) &&
+		    !xfs_sb_version_hasreflink(&mp->m_sb)) {
+			if (!uncertain) {
+				do_warn(
+	_("inode %" PRIu64 " is marked reflinked but file system does not support reflink\n"),
+					lino);
+			}
+			goto clear_bad_out;
+		}
+
+		if (flags2 & XFS_DIFLAG2_REFLINK) {
+			/* must be a file */
+			if (di_mode && !S_ISREG(di_mode)) {
+				if (!uncertain) {
+					do_warn(
+	_("reflink flag set on non-file inode %" PRIu64 "\n"),
+						lino);
+				}
+				goto clear_bad_out;
+			}
+		}
+
+		if ((flags2 & XFS_DIFLAG2_REFLINK) &&
+		    (flags & (XFS_DIFLAG_REALTIME | XFS_DIFLAG_RTINHERIT))) {
+			if (!uncertain) {
+				do_warn(
+	_("Cannot have a reflinked realtime inode %" PRIu64 "\n"),
+					lino);
+			}
+			goto clear_bad_out;
+		}
+
+		if (!verify_mode && flags2 != be64_to_cpu(dino->di_flags2)) {
+			if (!no_modify) {
+				do_warn(_("fixing bad flags2.\n"));
+				dino->di_flags2 = cpu_to_be64(flags2);
+				*dirty = 1;
+			} else
+				do_warn(_("would fix bad flags2.\n"));
+		}
+	}
+
 	if (verify_mode)
 		return retval;
 
diff --git a/repair/scan.c b/repair/scan.c
index c27f969..d3a1a82 100644
--- a/repair/scan.c
+++ b/repair/scan.c
@@ -872,6 +872,15 @@ _("in use block (%d,%d-%d) mismatch in %s tree, state - %d,%" PRIx64 "\n"),
 		 * be caught later.
 		 */
 		break;
+	case XR_E_INUSE1:
+		/*
+		 * multiple inode owners are ok with
+		 * reflink enabled
+		 */
+		if (xfs_sb_version_hasreflink(&mp->m_sb) &&
+		    !XFS_RMAP_NON_INODE_OWNER(owner))
+			break;
+		/* fall through */
 	default:
 		do_warn(
 _("unknown block (%d,%d-%d) mismatch on %s tree, state - %d,%" PRIx64 "\n"),
@@ -888,6 +897,28 @@ struct rmap_priv {
 	xfs_agblock_t		nr_blocks;
 };
 
+static bool
+rmap_in_order(
+	xfs_agblock_t	b,
+	xfs_agblock_t	lastblock,
+	uint64_t	owner,
+	uint64_t	lastowner,
+	uint64_t	offset,
+	uint64_t	lastoffset)
+{
+	if (b > lastblock)
+		return true;
+	else if (b < lastblock)
+		return false;
+
+	if (owner > lastowner)
+		return true;
+	else if (owner < lastowner)
+		return false;
+
+	return offset > lastoffset;
+}
+
 static void
 scan_rmapbt(
 	struct xfs_btree_block	*block,
@@ -908,6 +939,8 @@ scan_rmapbt(
 	int			numrecs;
 	int			state;
 	xfs_agblock_t		lastblock = 0;
+	uint64_t		lastowner = 0;
+	uint64_t		lastoffset = 0;
 	struct xfs_rmap_key	*kp;
 	struct xfs_rmap_irec	key = {0};
 
@@ -1038,10 +1071,17 @@ _("%s rmap btree block claimed (state %d), agno %d, bno %d, suspect %d\n"),
 			if (i == 0) {
 advance:
 				lastblock = b;
+				lastowner = owner;
+				lastoffset = offset;
 			} else {
 				bool bad;
 
-				bad = b <= lastblock;
+				if (xfs_sb_version_hasreflink(&mp->m_sb))
+					bad = !rmap_in_order(b, lastblock,
+							owner, lastowner,
+							offset, lastoffset);
+				else
+					bad = b <= lastblock;
 				if (bad)
 					do_warn(
 	_("out-of-order rmap btree record %d (%u %"PRId64" %"PRIx64" %u) block %u/%u\n"),

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 62/71] xfs_repair: process reverse-mapping data into refcount data

[Darrick J. Wong]

Take all the reverse-mapping data we've acquired and use it to generate
reference count data.  This data is used in phase 5 to rebuild the
refcount btree.

v2: Update to reflect separation of rmap_irec flags.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 repair/phase4.c |   27 ++++++
 repair/rmap.c   |  232 +++++++++++++++++++++++++++++++++++++++++++++++++++++++
 repair/rmap.h   |    2 
 3 files changed, 259 insertions(+), 2 deletions(-)


diff --git a/repair/phase4.c b/repair/phase4.c
index 9da1bb1..86992c9 100644
--- a/repair/phase4.c
+++ b/repair/phase4.c
@@ -193,6 +193,21 @@ _("%s while checking reverse-mappings"),
 }
 
 static void
+compute_ag_refcounts(
+	work_queue_t	*wq,
+	xfs_agnumber_t	agno,
+	void		*arg)
+{
+	int		error;
+
+	error = compute_refcounts(wq->mp, agno);
+	if (error)
+		do_error(
+_("%s while computing reference count records.\n"),
+			 strerror(-error));
+}
+
+static void
 process_rmap_data(
 	struct xfs_mount	*mp)
 {
@@ -206,6 +221,14 @@ process_rmap_data(
 	for (i = 0; i < mp->m_sb.sb_agcount; i++)
 		queue_work(&wq, check_rmap_btrees, i, NULL);
 	destroy_work_queue(&wq);
+
+	if (!xfs_sb_version_hasreflink(&mp->m_sb))
+		return;
+
+	create_work_queue(&wq, mp, libxfs_nproc());
+	for (i = 0; i < mp->m_sb.sb_agcount; i++)
+		queue_work(&wq, compute_ag_refcounts, i, NULL);
+	destroy_work_queue(&wq);
 }
 
 void
@@ -359,7 +382,9 @@ phase4(xfs_mount_t *mp)
 
 	/*
 	 * Process all the reverse-mapping data that we collected.  This
-	 * involves checking the rmap data against the btree.
+	 * involves checking the rmap data against the btree, computing
+	 * reference counts based on the rmap data, and checking the counts
+	 * against the refcount btree.
 	 */
 	process_rmap_data(mp);
 
diff --git a/repair/rmap.c b/repair/rmap.c
index 0baf4eb..0753448 100644
--- a/repair/rmap.c
+++ b/repair/rmap.c
@@ -42,6 +42,7 @@ struct xfs_ag_rmap {
 	int		ar_flcount;		/* agfl entries from leftover */
 						/* agbt allocations */
 	struct xfs_rmap_irec	ar_last_rmap;	/* last rmap seen */
+	struct xfs_slab	*ar_refcount_items;	/* refcount items, p4-5 */
 };
 
 static struct xfs_ag_rmap *ag_rmaps;
@@ -88,7 +89,8 @@ bool
 rmap_needs_work(
 	struct xfs_mount	*mp)
 {
-	return xfs_sb_version_hasrmapbt(&mp->m_sb);
+	return xfs_sb_version_hasreflink(&mp->m_sb) ||
+	       xfs_sb_version_hasrmapbt(&mp->m_sb);
 }
 
 /*
@@ -120,6 +122,11 @@ _("Insufficient memory while allocating reverse mapping slabs."));
 			do_error(
 _("Insufficient memory while allocating raw metadata reverse mapping slabs."));
 		ag_rmaps[i].ar_last_rmap.rm_owner = XFS_RMAP_OWN_UNKNOWN;
+		error = init_slab(&ag_rmaps[i].ar_refcount_items,
+				  sizeof(struct xfs_refcount_irec));
+		if (error)
+			do_error(
+_("Insufficient memory while allocating refcount item slabs."));
 	}
 }
 
@@ -138,6 +145,7 @@ rmaps_free(
 	for (i = 0; i < mp->m_sb.sb_agcount; i++) {
 		free_slab(&ag_rmaps[i].ar_rmaps);
 		free_slab(&ag_rmaps[i].ar_raw_rmaps);
+		free_slab(&ag_rmaps[i].ar_refcount_items);
 	}
 	free(ag_rmaps);
 	ag_rmaps = NULL;
@@ -591,6 +599,228 @@ rmap_dump(
 #endif
 
 /*
+ * Rebuilding the Reference Count & Reverse Mapping Btrees
+ *
+ * The reference count (refcnt) and reverse mapping (rmap) btrees are rebuilt
+ * during phase 5, like all other AG btrees.  Therefore, reverse mappings must
+ * be processed into reference counts at the end of phase 4, and the rmaps must
+ * be recorded during phase 4.  There is a need to access the rmaps in physical
+ * block order, but no particular need for random access, so the slab.c code
+ * provides a big logical array (consisting of smaller slabs) and some inorder
+ * iterator functions.
+ *
+ * Once we've recorded all the reverse mappings, we're ready to translate the
+ * rmaps into refcount entries.  Imagine the rmap entries as rectangles
+ * representing extents of physical blocks, and that the rectangles can be laid
+ * down to allow them to overlap each other; then we know that we must emit
+ * a refcnt btree entry wherever the amount of overlap changes, i.e. the
+ * emission stimulus is level-triggered:
+ *
+ *                 -    ---
+ *       --      ----- ----   ---        ------
+ * --   ----     ----------- ----     ---------
+ * -------------------------------- -----------
+ * ^ ^  ^^ ^^    ^ ^^ ^^^  ^^^^  ^ ^^ ^  ^     ^
+ * 2 1  23 21    3 43 234  2123  1 01 2  3     0
+ *
+ * For our purposes, a rmap is a tuple (startblock, len, fileoff, owner).
+ *
+ * Note that in the actual refcnt btree we don't store the refcount < 2 cases
+ * because the bnobt tells us which blocks are free; single-use blocks aren't
+ * recorded in the bnobt or the refcntbt.  If the rmapbt supports storing
+ * multiple entries covering a given block we could theoretically dispense with
+ * the refcntbt and simply count rmaps, but that's inefficient in the (hot)
+ * write path, so we'll take the cost of the extra tree to save time.  Also
+ * there's no guarantee that rmap will be enabled.
+ *
+ * Given an array of rmaps sorted by physical block number, a starting physical
+ * block (sp), a bag to hold rmaps that cover sp, and the next physical
+ * block where the level changes (np), we can reconstruct the refcount
+ * btree as follows:
+ *
+ * While there are still unprocessed rmaps in the array,
+ *  - Set sp to the physical block (pblk) of the next unprocessed rmap.
+ *  - Add to the bag all rmaps in the array where startblock == sp.
+ *  - Set np to the physical block where the bag size will change.
+ *    This is the minimum of (the pblk of the next unprocessed rmap) and
+ *    (startblock + len of each rmap in the bag).
+ *  - Record the bag size as old_bag_size.
+ *
+ *  - While the bag isn't empty,
+ *     - Remove from the bag all rmaps where startblock + len == np.
+ *     - Add to the bag all rmaps in the array where startblock == np.
+ *     - If the bag size isn't old_bag_size, store the refcount entry
+ *       (sp, np - sp, bag_size) in the refcnt btree.
+ *     - If the bag is empty, break out of the inner loop.
+ *     - Set old_bag_size to the bag size
+ *     - Set sp = np.
+ *     - Set np to the physical block where the bag size will change.
+ *       This is the minimum of (the pblk of the next unprocessed rmap) and
+ *       (startblock + len of each rmap in the bag).
+ *
+ * An implementation detail is that because this processing happens during
+ * phase 4, the refcount entries are stored in an array so that phase 5 can
+ * load them into the refcount btree.  The rmaps can be loaded directly into
+ * the rmap btree during phase 5 as well.
+ */
+
+/*
+ * Emit a refcount object for refcntbt reconstruction during phase 5.
+ */
+#define REFCOUNT_CLAMP(nr)	((nr) > MAXREFCOUNT ? MAXREFCOUNT : (nr))
+static void
+refcount_emit(
+	struct xfs_mount		*mp,
+	xfs_agnumber_t		agno,
+	xfs_agblock_t		agbno,
+	xfs_extlen_t		len,
+	size_t			nr_rmaps)
+{
+	struct xfs_refcount_irec	rlrec;
+	int			error;
+	struct xfs_slab		*rlslab;
+
+	rlslab = ag_rmaps[agno].ar_refcount_items;
+	ASSERT(nr_rmaps > 0);
+
+	dbg_printf("REFL: agno=%u pblk=%u, len=%u -> refcount=%zu\n",
+		agno, agbno, len, nr_rmaps);
+	rlrec.rc_startblock = agbno;
+	rlrec.rc_blockcount = len;
+	rlrec.rc_refcount = REFCOUNT_CLAMP(nr_rmaps);
+	error = slab_add(rlslab, &rlrec);
+	if (error)
+		do_error(
+_("Insufficient memory while recreating refcount tree."));
+}
+#undef REFCOUNT_CLAMP
+
+/*
+ * Transform a pile of physical block mapping observations into refcount data
+ * for eventual rebuilding of the btrees.
+ */
+#define RMAP_END(r)	((r)->rm_startblock + (r)->rm_blockcount)
+int
+compute_refcounts(
+	struct xfs_mount		*mp,
+	xfs_agnumber_t		agno)
+{
+	struct xfs_bag		*stack_top = NULL;
+	struct xfs_slab		*rmaps;
+	struct xfs_slab_cursor	*rmaps_cur;
+	struct xfs_rmap_irec	*array_cur;
+	struct xfs_rmap_irec	*rmap;
+	xfs_agblock_t		sbno;	/* first bno of this rmap set */
+	xfs_agblock_t		cbno;	/* first bno of this refcount set */
+	xfs_agblock_t		nbno;	/* next bno where rmap set changes */
+	size_t			n, idx;
+	size_t			old_stack_nr;
+	int			error;
+
+	if (!xfs_sb_version_hasreflink(&mp->m_sb))
+		return 0;
+
+	rmaps = ag_rmaps[agno].ar_rmaps;
+
+	error = init_slab_cursor(rmaps, rmap_compare, &rmaps_cur);
+	if (error)
+		return error;
+
+	error = init_bag(&stack_top);
+	if (error)
+		goto err;
+
+	/* While there are rmaps to be processed... */
+	n = 0;
+	while (n < slab_count(rmaps)) {
+		array_cur = peek_slab_cursor(rmaps_cur);
+		sbno = cbno = array_cur->rm_startblock;
+		/* Push all rmaps with pblk == sbno onto the stack */
+		for (;
+		     array_cur && array_cur->rm_startblock == sbno;
+		     array_cur = peek_slab_cursor(rmaps_cur)) {
+			advance_slab_cursor(rmaps_cur); n++;
+			rmap_dump("push0", agno, array_cur);
+			error = bag_add(stack_top, array_cur);
+			if (error)
+				goto err;
+		}
+
+		/* Set nbno to the bno of the next refcount change */
+		if (n < slab_count(rmaps))
+			nbno = array_cur->rm_startblock;
+		else
+			nbno = NULLAGBLOCK;
+		foreach_bag_ptr(stack_top, idx, rmap) {
+			nbno = min(nbno, RMAP_END(rmap));
+		}
+
+		/* Emit reverse mappings, if needed */
+		ASSERT(nbno > sbno);
+		old_stack_nr = bag_count(stack_top);
+
+		/* While stack isn't empty... */
+		while (bag_count(stack_top)) {
+			/* Pop all rmaps that end at nbno */
+			foreach_bag_ptr_reverse(stack_top, idx, rmap) {
+				if (RMAP_END(rmap) != nbno)
+					continue;
+				rmap_dump("pop", agno, rmap);
+				error = bag_remove(stack_top, idx);
+				if (error)
+					goto err;
+			}
+
+			/* Push array items that start at nbno */
+			for (;
+			     array_cur && array_cur->rm_startblock == nbno;
+			     array_cur = peek_slab_cursor(rmaps_cur)) {
+				advance_slab_cursor(rmaps_cur); n++;
+				rmap_dump("push1", agno, array_cur);
+				error = bag_add(stack_top, array_cur);
+				if (error)
+					goto err;
+			}
+
+			/* Emit refcount if necessary */
+			ASSERT(nbno > cbno);
+			if (bag_count(stack_top) != old_stack_nr) {
+				if (old_stack_nr > 1) {
+					refcount_emit(mp, agno, cbno,
+						      nbno - cbno,
+						      old_stack_nr);
+				}
+				cbno = nbno;
+			}
+
+			/* Stack empty, go find the next rmap */
+			if (bag_count(stack_top) == 0)
+				break;
+			old_stack_nr = bag_count(stack_top);
+			sbno = nbno;
+
+			/* Set nbno to the bno of the next refcount change */
+			if (n < slab_count(rmaps))
+				nbno = array_cur->rm_startblock;
+			else
+				nbno = NULLAGBLOCK;
+			foreach_bag_ptr(stack_top, idx, rmap) {
+				nbno = min(nbno, RMAP_END(rmap));
+			}
+
+			/* Emit reverse mappings, if needed */
+			ASSERT(nbno > sbno);
+		}
+	}
+err:
+	free_bag(&stack_top);
+	free_slab_cursor(&rmaps_cur);
+
+	return error;
+}
+#undef RMAP_END
+
+/*
  * Return the number of rmap objects for an AG.
  */
 size_t
diff --git a/repair/rmap.h b/repair/rmap.h
index 7106dfc..01dec9f 100644
--- a/repair/rmap.h
+++ b/repair/rmap.h
@@ -49,6 +49,8 @@ extern __int64_t rmap_diffkeys(struct xfs_rmap_irec *kp1,
 extern void rmap_high_key_from_rec(struct xfs_rmap_irec *rec,
 		struct xfs_rmap_irec *key);
 
+extern int compute_refcounts(struct xfs_mount *, xfs_agnumber_t);
+
 extern void fix_freelist(struct xfs_mount *, xfs_agnumber_t, bool);
 extern void rmap_store_agflcount(struct xfs_mount *, xfs_agnumber_t, int);
 

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 63/71] xfs_repair: record reflink inode state

[Darrick J. Wong]

Record the state of the per-inode reflink flag, so that we can
compare against the rmap data and update the flags accordingly.
Clear the (reflink) state if we clear the inode.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 repair/dino_chunks.c |    1 +
 repair/dinode.c      |    6 ++++++
 repair/incore.h      |   38 ++++++++++++++++++++++++++++++++++++++
 repair/incore_ino.c  |    2 ++
 repair/rmap.c        |   26 ++++++++++++++++++++++++++
 repair/rmap.h        |    2 ++
 6 files changed, 75 insertions(+)


diff --git a/repair/dino_chunks.c b/repair/dino_chunks.c
index 7dbaca6..4db9512 100644
--- a/repair/dino_chunks.c
+++ b/repair/dino_chunks.c
@@ -931,6 +931,7 @@ next_readbuf:
 				do_warn(_("would have cleared inode %" PRIu64 "\n"),
 					ino);
 			}
+			clear_inode_was_rl(ino_rec, irec_offset);
 		}
 
 process_next:
diff --git a/repair/dinode.c b/repair/dinode.c
index 98afdc9..64fc983 100644
--- a/repair/dinode.c
+++ b/repair/dinode.c
@@ -2636,6 +2636,12 @@ _("bad non-zero extent size %u for non-realtime/extsize inode %" PRIu64 ", "),
 		goto clear_bad_out;
 
 	/*
+	 * record the state of the reflink flag
+	 */
+	if (collect_rmaps)
+		record_inode_reflink_flag(mp, dino, agno, ino, lino);
+
+	/*
 	 * check data fork -- if it's bad, clear the inode
 	 */
 	if (process_inode_data_fork(mp, agno, ino, dino, type, dirty,
diff --git a/repair/incore.h b/repair/incore.h
index b6c4b4f..bcd2f4b 100644
--- a/repair/incore.h
+++ b/repair/incore.h
@@ -283,6 +283,8 @@ typedef struct ino_tree_node  {
 	__uint64_t		ir_sparse;	/* sparse inode bitmask */
 	__uint64_t		ino_confirmed;	/* confirmed bitmask */
 	__uint64_t		ino_isa_dir;	/* bit == 1 if a directory */
+	__uint64_t		ino_was_rl;	/* bit == 1 if reflink flag set */
+	__uint64_t		ino_is_rl;	/* bit == 1 if reflink flag should be set */
 	__uint8_t		nlink_size;
 	union ino_nlink		disk_nlinks;	/* on-disk nlinks, set in P3 */
 	union  {
@@ -494,6 +496,42 @@ static inline bool is_inode_sparse(struct ino_tree_node *irec, int offset)
 }
 
 /*
+ * set/clear/test was inode marked as reflinked
+ */
+static inline void set_inode_was_rl(struct ino_tree_node *irec, int offset)
+{
+	irec->ino_was_rl |= IREC_MASK(offset);
+}
+
+static inline void clear_inode_was_rl(struct ino_tree_node *irec, int offset)
+{
+	irec->ino_was_rl &= ~IREC_MASK(offset);
+}
+
+static inline int inode_was_rl(struct ino_tree_node *irec, int offset)
+{
+	return (irec->ino_was_rl & IREC_MASK(offset)) != 0;
+}
+
+/*
+ * set/clear/test should inode be marked as reflinked
+ */
+static inline void set_inode_is_rl(struct ino_tree_node *irec, int offset)
+{
+	irec->ino_is_rl |= IREC_MASK(offset);
+}
+
+static inline void clear_inode_is_rl(struct ino_tree_node *irec, int offset)
+{
+	irec->ino_is_rl &= ~IREC_MASK(offset);
+}
+
+static inline int inode_is_rl(struct ino_tree_node *irec, int offset)
+{
+	return (irec->ino_is_rl & IREC_MASK(offset)) != 0;
+}
+
+/*
  * add_inode_reached() is set on inode I only if I has been reached
  * by an inode P claiming to be the parent and if I is a directory,
  * the .. link in the I says that P is I's parent.
diff --git a/repair/incore_ino.c b/repair/incore_ino.c
index 1898257..2ec1765 100644
--- a/repair/incore_ino.c
+++ b/repair/incore_ino.c
@@ -257,6 +257,8 @@ alloc_ino_node(
 	irec->ino_startnum = starting_ino;
 	irec->ino_confirmed = 0;
 	irec->ino_isa_dir = 0;
+	irec->ino_was_rl = 0;
+	irec->ino_is_rl = 0;
 	irec->ir_free = (xfs_inofree_t) - 1;
 	irec->ir_sparse = 0;
 	irec->ino_un.ex_data = NULL;
diff --git a/repair/rmap.c b/repair/rmap.c
index 0753448..4507420 100644
--- a/repair/rmap.c
+++ b/repair/rmap.c
@@ -1076,6 +1076,32 @@ rmap_high_key_from_rec(
 }
 
 /*
+ * Record that an inode had the reflink flag set when repair started.  The
+ * inode reflink flag will be adjusted as necessary.
+ */
+void
+record_inode_reflink_flag(
+	struct xfs_mount	*mp,
+	struct xfs_dinode	*dino,
+	xfs_agnumber_t		agno,
+	xfs_agino_t		ino,
+	xfs_ino_t		lino)
+{
+	struct ino_tree_node	*irec;
+	int			off;
+
+	ASSERT(XFS_AGINO_TO_INO(mp, agno, ino) == be64_to_cpu(dino->di_ino));
+	if (!(be64_to_cpu(dino->di_flags2) & XFS_DIFLAG2_REFLINK))
+		return;
+	irec = find_inode_rec(mp, agno, ino);
+	off = get_inode_offset(mp, lino, irec);
+	ASSERT(!inode_was_rl(irec, off));
+	set_inode_was_rl(irec, off);
+	dbg_printf("set was_rl lino=%llu was=0x%llx\n",
+		(unsigned long long)lino, (unsigned long long)irec->ino_was_rl);
+}
+
+/*
  * Regenerate the AGFL so that we don't run out of it while rebuilding the
  * rmap btree.  If skip_rmapbt is true, don't update the rmapbt (most probably
  * because we're updating the rmapbt).
diff --git a/repair/rmap.h b/repair/rmap.h
index 01dec9f..ab6f434 100644
--- a/repair/rmap.h
+++ b/repair/rmap.h
@@ -50,6 +50,8 @@ extern void rmap_high_key_from_rec(struct xfs_rmap_irec *rec,
 		struct xfs_rmap_irec *key);
 
 extern int compute_refcounts(struct xfs_mount *, xfs_agnumber_t);
+extern void record_inode_reflink_flag(struct xfs_mount *, struct xfs_dinode *,
+	xfs_agnumber_t, xfs_agino_t, xfs_ino_t);
 
 extern void fix_freelist(struct xfs_mount *, xfs_agnumber_t, bool);
 extern void rmap_store_agflcount(struct xfs_mount *, xfs_agnumber_t, int);

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 64/71] xfs_repair: fix inode reflink flags

[Darrick J. Wong]

While we're computing reference counts, record which inodes actually
share blocks with other files and fix the flags as necessary.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 repair/phase4.c |   20 ++++++++
 repair/rmap.c   |  133 +++++++++++++++++++++++++++++++++++++++++++++++++++++++
 repair/rmap.h   |    1 
 3 files changed, 154 insertions(+)


diff --git a/repair/phase4.c b/repair/phase4.c
index 86992c9..2c2d611 100644
--- a/repair/phase4.c
+++ b/repair/phase4.c
@@ -208,6 +208,21 @@ _("%s while computing reference count records.\n"),
 }
 
 static void
+process_inode_reflink_flags(
+	struct work_queue	*wq,
+	xfs_agnumber_t		agno,
+	void			*arg)
+{
+	int			error;
+
+	error = fix_inode_reflink_flags(wq->mp, agno);
+	if (error)
+		do_error(
+_("%s while fixing inode reflink flags.\n"),
+			 strerror(-error));
+}
+
+static void
 process_rmap_data(
 	struct xfs_mount	*mp)
 {
@@ -229,6 +244,11 @@ process_rmap_data(
 	for (i = 0; i < mp->m_sb.sb_agcount; i++)
 		queue_work(&wq, compute_ag_refcounts, i, NULL);
 	destroy_work_queue(&wq);
+
+	create_work_queue(&wq, mp, libxfs_nproc());
+	for (i = 0; i < mp->m_sb.sb_agcount; i++)
+		queue_work(&wq, process_inode_reflink_flags, i, NULL);
+	destroy_work_queue(&wq);
 }
 
 void
diff --git a/repair/rmap.c b/repair/rmap.c
index 4507420..5206206 100644
--- a/repair/rmap.c
+++ b/repair/rmap.c
@@ -665,6 +665,39 @@ rmap_dump(
  */
 
 /*
+ * Mark all inodes in the reverse-mapping observation stack as requiring the
+ * reflink inode flag, if the stack depth is greater than 1.
+ */
+static void
+mark_inode_rl(
+	struct xfs_mount		*mp,
+	struct xfs_bag		*rmaps)
+{
+	xfs_agnumber_t		iagno;
+	struct xfs_rmap_irec	*rmap;
+	struct ino_tree_node	*irec;
+	int			off;
+	size_t			idx;
+	xfs_agino_t		ino;
+
+	if (bag_count(rmaps) < 2)
+		return;
+
+	/* Reflink flag accounting */
+	foreach_bag_ptr(rmaps, idx, rmap) {
+		ASSERT(!XFS_RMAP_NON_INODE_OWNER(rmap->rm_owner));
+		iagno = XFS_INO_TO_AGNO(mp, rmap->rm_owner);
+		ino = XFS_INO_TO_AGINO(mp, rmap->rm_owner);
+		pthread_mutex_lock(&ag_locks[iagno].lock);
+		irec = find_inode_rec(mp, iagno, ino);
+		off = get_inode_offset(mp, rmap->rm_owner, irec);
+		/* lock here because we might go outside this ag */
+		set_inode_is_rl(irec, off);
+		pthread_mutex_unlock(&ag_locks[iagno].lock);
+	}
+}
+
+/*
  * Emit a refcount object for refcntbt reconstruction during phase 5.
  */
 #define REFCOUNT_CLAMP(nr)	((nr) > MAXREFCOUNT ? MAXREFCOUNT : (nr))
@@ -745,6 +778,7 @@ compute_refcounts(
 			if (error)
 				goto err;
 		}
+		mark_inode_rl(mp, stack_top);
 
 		/* Set nbno to the bno of the next refcount change */
 		if (n < slab_count(rmaps))
@@ -781,6 +815,7 @@ compute_refcounts(
 				if (error)
 					goto err;
 			}
+			mark_inode_rl(mp, stack_top);
 
 			/* Emit refcount if necessary */
 			ASSERT(nbno > cbno);
@@ -1102,6 +1137,104 @@ record_inode_reflink_flag(
 }
 
 /*
+ * Fix an inode's reflink flag.
+ */
+static int
+fix_inode_reflink_flag(
+	struct xfs_mount	*mp,
+	xfs_agnumber_t		agno,
+	xfs_agino_t		agino,
+	bool			set)
+{
+	struct xfs_dinode	*dino;
+	struct xfs_buf		*buf;
+
+	if (set)
+		do_warn(
+_("setting reflink flag on inode %"PRIu64"\n"),
+			XFS_AGINO_TO_INO(mp, agno, agino));
+	else if (!no_modify) /* && !set */
+		do_warn(
+_("clearing reflink flag on inode %"PRIu64"\n"),
+			XFS_AGINO_TO_INO(mp, agno, agino));
+	if (no_modify)
+		return 0;
+
+	buf = get_agino_buf(mp, agno, agino, &dino);
+	if (!buf)
+		return 1;
+	ASSERT(XFS_AGINO_TO_INO(mp, agno, agino) == be64_to_cpu(dino->di_ino));
+	if (set)
+		dino->di_flags2 |= cpu_to_be64(XFS_DIFLAG2_REFLINK);
+	else
+		dino->di_flags2 &= cpu_to_be64(~XFS_DIFLAG2_REFLINK);
+	libxfs_dinode_calc_crc(mp, dino);
+	libxfs_writebuf(buf, 0);
+
+	return 0;
+}
+
+/*
+ * Fix discrepancies between the state of the inode reflink flag and our
+ * observations as to whether or not the inode really needs it.
+ */
+int
+fix_inode_reflink_flags(
+	struct xfs_mount	*mp,
+	xfs_agnumber_t		agno)
+{
+	struct ino_tree_node	*irec;
+	int			bit;
+	__uint64_t		was;
+	__uint64_t		is;
+	__uint64_t		diff;
+	__uint64_t		mask;
+	int			error = 0;
+	xfs_agino_t		agino;
+
+	/*
+	 * Update the reflink flag for any inode where there's a discrepancy
+	 * between the inode flag and whether or not we found any reflinked
+	 * extents.
+	 */
+	for (irec = findfirst_inode_rec(agno);
+	     irec != NULL;
+	     irec = next_ino_rec(irec)) {
+		ASSERT((irec->ino_was_rl & irec->ir_free) == 0);
+		ASSERT((irec->ino_is_rl & irec->ir_free) == 0);
+		was = irec->ino_was_rl;
+		is = irec->ino_is_rl;
+		if (was == is)
+			continue;
+		diff = was ^ is;
+		dbg_printf("mismatch ino=%llu was=0x%lx is=0x%lx dif=0x%lx\n",
+			(unsigned long long)XFS_AGINO_TO_INO(mp, agno,
+						irec->ino_startnum),
+			was, is, diff);
+
+		for (bit = 0, mask = 1; bit < 64; bit++, mask <<= 1) {
+			agino = bit + irec->ino_startnum;
+			if (!(diff & mask))
+				continue;
+			else if (was & mask)
+				error = fix_inode_reflink_flag(mp, agno, agino,
+						false);
+			else if (is & mask)
+				error = fix_inode_reflink_flag(mp, agno, agino,
+						true);
+			else
+				ASSERT(0);
+			if (error)
+				do_error(
+_("Unable to fix reflink flag on inode %"PRIu64".\n"),
+					XFS_AGINO_TO_INO(mp, agno, agino));
+		}
+	}
+
+	return error;
+}
+
+/*
  * Regenerate the AGFL so that we don't run out of it while rebuilding the
  * rmap btree.  If skip_rmapbt is true, don't update the rmapbt (most probably
  * because we're updating the rmapbt).
diff --git a/repair/rmap.h b/repair/rmap.h
index ab6f434..266448f 100644
--- a/repair/rmap.h
+++ b/repair/rmap.h
@@ -52,6 +52,7 @@ extern void rmap_high_key_from_rec(struct xfs_rmap_irec *rec,
 extern int compute_refcounts(struct xfs_mount *, xfs_agnumber_t);
 extern void record_inode_reflink_flag(struct xfs_mount *, struct xfs_dinode *,
 	xfs_agnumber_t, xfs_agino_t, xfs_ino_t);
+extern int fix_inode_reflink_flags(struct xfs_mount *, xfs_agnumber_t);
 
 extern void fix_freelist(struct xfs_mount *, xfs_agnumber_t, bool);
 extern void rmap_store_agflcount(struct xfs_mount *, xfs_agnumber_t, int);

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 65/71] xfs_repair: check the refcount btree against our observed reference counts when -n

[Darrick J. Wong]

Check the observed reference counts against whatever's in the refcount
btree for discrepancies.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/libxfs_api_defs.h |    4 +
 repair/phase4.c          |   20 +++++++
 repair/rmap.c            |  126 ++++++++++++++++++++++++++++++++++++++++++++++
 repair/rmap.h            |    5 ++
 repair/scan.c            |    3 +
 5 files changed, 158 insertions(+)


diff --git a/libxfs/libxfs_api_defs.h b/libxfs/libxfs_api_defs.h
index b9eb2cf..28bac11 100644
--- a/libxfs/libxfs_api_defs.h
+++ b/libxfs/libxfs_api_defs.h
@@ -137,4 +137,8 @@
 #define xfs_prealloc_blocks		libxfs_prealloc_blocks
 #define xfs_dinode_good_version		libxfs_dinode_good_version
 
+#define xfs_refcountbt_init_cursor	libxfs_refcountbt_init_cursor
+#define xfs_refcount_lookup_le		libxfs_refcount_lookup_le
+#define xfs_refcount_get_rec		libxfs_refcount_get_rec
+
 #endif /* __LIBXFS_API_DEFS_H__ */
diff --git a/repair/phase4.c b/repair/phase4.c
index 2c2d611..395d373 100644
--- a/repair/phase4.c
+++ b/repair/phase4.c
@@ -223,6 +223,21 @@ _("%s while fixing inode reflink flags.\n"),
 }
 
 static void
+check_refcount_btrees(
+	work_queue_t	*wq,
+	xfs_agnumber_t	agno,
+	void		*arg)
+{
+	int		error;
+
+	error = check_refcounts(wq->mp, agno);
+	if (error)
+		do_error(
+_("%s while checking reference counts"),
+			 strerror(-error));
+}
+
+static void
 process_rmap_data(
 	struct xfs_mount	*mp)
 {
@@ -249,6 +264,11 @@ process_rmap_data(
 	for (i = 0; i < mp->m_sb.sb_agcount; i++)
 		queue_work(&wq, process_inode_reflink_flags, i, NULL);
 	destroy_work_queue(&wq);
+
+	create_work_queue(&wq, mp, libxfs_nproc());
+	for (i = 0; i < mp->m_sb.sb_agcount; i++)
+		queue_work(&wq, check_refcount_btrees, i, NULL);
+	destroy_work_queue(&wq);
 }
 
 void
diff --git a/repair/rmap.c b/repair/rmap.c
index 5206206..c53d6b7 100644
--- a/repair/rmap.c
+++ b/repair/rmap.c
@@ -47,6 +47,7 @@ struct xfs_ag_rmap {
 
 static struct xfs_ag_rmap *ag_rmaps;
 static bool rmapbt_suspect;
+static bool refcbt_suspect;
 
 /*
  * Compare rmap observations for array sorting.
@@ -1235,6 +1236,131 @@ _("Unable to fix reflink flag on inode %"PRIu64".\n"),
 }
 
 /*
+ * Return the number of refcount objects for an AG.
+ */
+size_t
+refcount_record_count(
+	struct xfs_mount		*mp,
+	xfs_agnumber_t		agno)
+{
+	return slab_count(ag_rmaps[agno].ar_refcount_items);
+}
+
+/*
+ * Return a slab cursor that will return refcount objects in order.
+ */
+int
+init_refcount_cursor(
+	xfs_agnumber_t		agno,
+	struct xfs_slab_cursor	**cur)
+{
+	return init_slab_cursor(ag_rmaps[agno].ar_refcount_items, NULL, cur);
+}
+
+/*
+ * Disable the refcount btree check.
+ */
+void
+refcount_avoid_check(void)
+{
+	refcbt_suspect = true;
+}
+
+/*
+ * Compare the observed reference counts against what's in the ag btree.
+ */
+int
+check_refcounts(
+	struct xfs_mount	*mp,
+	xfs_agnumber_t		agno)
+{
+	struct xfs_slab_cursor	*rl_cur;
+	struct xfs_btree_cur	*bt_cur = NULL;
+	int			error;
+	int			have;
+	int			i;
+	struct xfs_buf		*agbp = NULL;
+	struct xfs_refcount_irec	*rl_rec;
+	struct xfs_refcount_irec	tmp;
+	struct xfs_perag	*pag;		/* per allocation group data */
+
+	if (!xfs_sb_version_hasreflink(&mp->m_sb))
+		return 0;
+	if (refcbt_suspect) {
+		if (no_modify && agno == 0)
+			do_warn(_("would rebuild corrupt refcount btrees.\n"));
+		return 0;
+	}
+
+	/* Create cursors to refcount structures */
+	error = init_refcount_cursor(agno, &rl_cur);
+	if (error)
+		return error;
+
+	error = -libxfs_alloc_read_agf(mp, NULL, agno, 0, &agbp);
+	if (error)
+		goto err;
+
+	/* Leave the per-ag data "uninitialized" since we rewrite it later */
+	pag = libxfs_perag_get(mp, agno);
+	pag->pagf_init = 0;
+	libxfs_perag_put(pag);
+
+	bt_cur = libxfs_refcountbt_init_cursor(mp, NULL, agbp, agno, NULL);
+	if (!bt_cur) {
+		error = -ENOMEM;
+		goto err;
+	}
+
+	rl_rec = pop_slab_cursor(rl_cur);
+	while (rl_rec) {
+		/* Look for a refcount record in the btree */
+		error = -libxfs_refcount_lookup_le(bt_cur,
+				rl_rec->rc_startblock, &have);
+		if (error)
+			goto err;
+		if (!have) {
+			do_warn(
+_("Missing reference count record for (%u/%u) len %u count %u\n"),
+				agno, rl_rec->rc_startblock,
+				rl_rec->rc_blockcount, rl_rec->rc_refcount);
+			goto next_loop;
+		}
+
+		error = -libxfs_refcount_get_rec(bt_cur, &tmp, &i);
+		if (error)
+			goto err;
+		if (!i) {
+			do_warn(
+_("Missing reference count record for (%u/%u) len %u count %u\n"),
+				agno, rl_rec->rc_startblock,
+				rl_rec->rc_blockcount, rl_rec->rc_refcount);
+			goto next_loop;
+		}
+
+		/* Compare each refcount observation against the btree's */
+		if (tmp.rc_startblock != rl_rec->rc_startblock ||
+		    tmp.rc_blockcount < rl_rec->rc_blockcount ||
+		    tmp.rc_refcount < rl_rec->rc_refcount)
+			do_warn(
+_("Incorrect reference count: saw (%u/%u) len %u nlinks %u; should be (%u/%u) len %u nlinks %u\n"),
+				agno, tmp.rc_startblock, tmp.rc_blockcount,
+				tmp.rc_refcount, agno, rl_rec->rc_startblock,
+				rl_rec->rc_blockcount, rl_rec->rc_refcount);
+next_loop:
+		rl_rec = pop_slab_cursor(rl_cur);
+	}
+
+err:
+	if (bt_cur)
+		libxfs_btree_del_cursor(bt_cur, XFS_BTREE_NOERROR);
+	if (agbp)
+		libxfs_putbuf(agbp);
+	free_slab_cursor(&rl_cur);
+	return 0;
+}
+
+/*
  * Regenerate the AGFL so that we don't run out of it while rebuilding the
  * rmap btree.  If skip_rmapbt is true, don't update the rmapbt (most probably
  * because we're updating the rmapbt).
diff --git a/repair/rmap.h b/repair/rmap.h
index 266448f..752ece8 100644
--- a/repair/rmap.h
+++ b/repair/rmap.h
@@ -50,6 +50,11 @@ extern void rmap_high_key_from_rec(struct xfs_rmap_irec *rec,
 		struct xfs_rmap_irec *key);
 
 extern int compute_refcounts(struct xfs_mount *, xfs_agnumber_t);
+extern size_t refcount_record_count(struct xfs_mount *, xfs_agnumber_t);
+extern int init_refcount_cursor(xfs_agnumber_t, struct xfs_slab_cursor **);
+extern void refcount_avoid_check(void);
+extern int check_refcounts(struct xfs_mount *, xfs_agnumber_t);
+
 extern void record_inode_reflink_flag(struct xfs_mount *, struct xfs_dinode *,
 	xfs_agnumber_t, xfs_agino_t, xfs_ino_t);
 extern int fix_inode_reflink_flags(struct xfs_mount *, xfs_agnumber_t);
diff --git a/repair/scan.c b/repair/scan.c
index d3a1a82..1c60784 100644
--- a/repair/scan.c
+++ b/repair/scan.c
@@ -1370,6 +1370,8 @@ _("%s btree block claimed (state %d), agno %d, bno %d, suspect %d\n"),
 		}
 	}
 out:
+	if (suspect)
+		refcount_avoid_check();
 	return;
 }
 
@@ -2173,6 +2175,7 @@ validate_agf(
 		} else  {
 			do_warn(_("bad agbno %u for refcntbt root, agno %d\n"),
 				bno, agno);
+			refcount_avoid_check();
 		}
 	}
 

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 66/71] xfs_repair: rebuild the refcount btree

[Darrick J. Wong]

Rebuild the refcount btree with the reference count data we assembled
during phase 4.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 repair/phase5.c |  318 +++++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 316 insertions(+), 2 deletions(-)


diff --git a/repair/phase5.c b/repair/phase5.c
index b464b56..1b29043 100644
--- a/repair/phase5.c
+++ b/repair/phase5.c
@@ -1692,6 +1692,297 @@ _("Insufficient memory to construct reverse-map cursor."));
 	free_slab_cursor(&rmap_cur);
 }
 
+/* rebuild the refcount tree */
+
+/*
+ * we don't have to worry here about how chewing up free extents
+ * may perturb things because reflink tree building happens before
+ * freespace tree building.
+ */
+static void
+init_refc_cursor(
+	struct xfs_mount	*mp,
+	xfs_agnumber_t		agno,
+	struct bt_status	*btree_curs)
+{
+	size_t			num_recs;
+	int			level;
+	struct bt_stat_level	*lptr;
+	struct bt_stat_level	*p_lptr;
+	xfs_extlen_t		blocks_allocated;
+
+	if (!xfs_sb_version_hasreflink(&mp->m_sb)) {
+		memset(btree_curs, 0, sizeof(struct bt_status));
+		return;
+	}
+
+	lptr = &btree_curs->level[0];
+	btree_curs->init = 1;
+	btree_curs->owner = XFS_RMAP_OWN_REFC;
+
+	/*
+	 * build up statistics
+	 */
+	num_recs = refcount_record_count(mp, agno);
+	if (num_recs == 0) {
+		/*
+		 * easy corner-case -- no refcount records
+		 */
+		lptr->num_blocks = 1;
+		lptr->modulo = 0;
+		lptr->num_recs_pb = 0;
+		lptr->num_recs_tot = 0;
+
+		btree_curs->num_levels = 1;
+		btree_curs->num_tot_blocks = btree_curs->num_free_blocks = 1;
+
+		setup_cursor(mp, agno, btree_curs);
+
+		return;
+	}
+
+	blocks_allocated = lptr->num_blocks = howmany(num_recs,
+					mp->m_refc_mxr[0]);
+
+	lptr->modulo = num_recs % lptr->num_blocks;
+	lptr->num_recs_pb = num_recs / lptr->num_blocks;
+	lptr->num_recs_tot = num_recs;
+	level = 1;
+
+	if (lptr->num_blocks > 1)  {
+		for (; btree_curs->level[level-1].num_blocks > 1
+				&& level < XFS_BTREE_MAXLEVELS;
+				level++)  {
+			lptr = &btree_curs->level[level];
+			p_lptr = &btree_curs->level[level - 1];
+			lptr->num_blocks = howmany(p_lptr->num_blocks,
+					mp->m_refc_mxr[1]);
+			lptr->modulo = p_lptr->num_blocks % lptr->num_blocks;
+			lptr->num_recs_pb = p_lptr->num_blocks
+					/ lptr->num_blocks;
+			lptr->num_recs_tot = p_lptr->num_blocks;
+
+			blocks_allocated += lptr->num_blocks;
+		}
+	}
+	ASSERT(lptr->num_blocks == 1);
+	btree_curs->num_levels = level;
+
+	btree_curs->num_tot_blocks = btree_curs->num_free_blocks
+			= blocks_allocated;
+
+	setup_cursor(mp, agno, btree_curs);
+}
+
+static void
+prop_refc_cursor(
+	struct xfs_mount	*mp,
+	xfs_agnumber_t		agno,
+	struct bt_status	*btree_curs,
+	xfs_agblock_t		startbno,
+	int			level)
+{
+	struct xfs_btree_block	*bt_hdr;
+	struct xfs_refcount_key	*bt_key;
+	xfs_refcount_ptr_t	*bt_ptr;
+	xfs_agblock_t		agbno;
+	struct bt_stat_level	*lptr;
+
+	level++;
+
+	if (level >= btree_curs->num_levels)
+		return;
+
+	lptr = &btree_curs->level[level];
+	bt_hdr = XFS_BUF_TO_BLOCK(lptr->buf_p);
+
+	if (be16_to_cpu(bt_hdr->bb_numrecs) == 0)  {
+		/*
+		 * this only happens once to initialize the
+		 * first path up the left side of the tree
+		 * where the agbno's are already set up
+		 */
+		prop_refc_cursor(mp, agno, btree_curs, startbno, level);
+	}
+
+	if (be16_to_cpu(bt_hdr->bb_numrecs) ==
+				lptr->num_recs_pb + (lptr->modulo > 0))  {
+		/*
+		 * write out current prev block, grab us a new block,
+		 * and set the rightsib pointer of current block
+		 */
+#ifdef XR_BLD_INO_TRACE
+		fprintf(stderr, " ino prop agbno %d ", lptr->prev_agbno);
+#endif
+		if (lptr->prev_agbno != NULLAGBLOCK)  {
+			ASSERT(lptr->prev_buf_p != NULL);
+			libxfs_writebuf(lptr->prev_buf_p, 0);
+		}
+		lptr->prev_agbno = lptr->agbno;
+		lptr->prev_buf_p = lptr->buf_p;
+		agbno = get_next_blockaddr(agno, level, btree_curs);
+
+		bt_hdr->bb_u.s.bb_rightsib = cpu_to_be32(agbno);
+
+		lptr->buf_p = libxfs_getbuf(mp->m_dev,
+					XFS_AGB_TO_DADDR(mp, agno, agbno),
+					XFS_FSB_TO_BB(mp, 1));
+		lptr->agbno = agbno;
+
+		if (lptr->modulo)
+			lptr->modulo--;
+
+		/*
+		 * initialize block header
+		 */
+		lptr->buf_p->b_ops = &xfs_refcountbt_buf_ops;
+		bt_hdr = XFS_BUF_TO_BLOCK(lptr->buf_p);
+		memset(bt_hdr, 0, mp->m_sb.sb_blocksize);
+		libxfs_btree_init_block(mp, lptr->buf_p, XFS_REFC_CRC_MAGIC,
+					level, 0, agno,
+					XFS_BTREE_CRC_BLOCKS);
+
+		bt_hdr->bb_u.s.bb_leftsib = cpu_to_be32(lptr->prev_agbno);
+
+		/*
+		 * propagate extent record for first extent in new block up
+		 */
+		prop_refc_cursor(mp, agno, btree_curs, startbno, level);
+	}
+	/*
+	 * add inode info to current block
+	 */
+	be16_add_cpu(&bt_hdr->bb_numrecs, 1);
+
+	bt_key = XFS_REFCOUNT_KEY_ADDR(bt_hdr,
+				    be16_to_cpu(bt_hdr->bb_numrecs));
+	bt_ptr = XFS_REFCOUNT_PTR_ADDR(bt_hdr,
+				    be16_to_cpu(bt_hdr->bb_numrecs),
+				    mp->m_refc_mxr[1]);
+
+	bt_key->rc_startblock = cpu_to_be32(startbno);
+	*bt_ptr = cpu_to_be32(btree_curs->level[level-1].agbno);
+}
+
+/*
+ * rebuilds a refcount btree given a cursor.
+ */
+static void
+build_refcount_tree(
+	struct xfs_mount	*mp,
+	xfs_agnumber_t		agno,
+	struct bt_status	*btree_curs)
+{
+	xfs_agnumber_t		i;
+	xfs_agblock_t		j;
+	xfs_agblock_t		agbno;
+	struct xfs_btree_block	*bt_hdr;
+	struct xfs_refcount_irec	*refc_rec;
+	struct xfs_slab_cursor	*refc_cur;
+	struct xfs_refcount_rec	*bt_rec;
+	struct bt_stat_level	*lptr;
+	int			level = btree_curs->num_levels;
+	int			error;
+
+	for (i = 0; i < level; i++)  {
+		lptr = &btree_curs->level[i];
+
+		agbno = get_next_blockaddr(agno, i, btree_curs);
+		lptr->buf_p = libxfs_getbuf(mp->m_dev,
+					XFS_AGB_TO_DADDR(mp, agno, agbno),
+					XFS_FSB_TO_BB(mp, 1));
+
+		if (i == btree_curs->num_levels - 1)
+			btree_curs->root = agbno;
+
+		lptr->agbno = agbno;
+		lptr->prev_agbno = NULLAGBLOCK;
+		lptr->prev_buf_p = NULL;
+		/*
+		 * initialize block header
+		 */
+
+		lptr->buf_p->b_ops = &xfs_refcountbt_buf_ops;
+		bt_hdr = XFS_BUF_TO_BLOCK(lptr->buf_p);
+		memset(bt_hdr, 0, mp->m_sb.sb_blocksize);
+		libxfs_btree_init_block(mp, lptr->buf_p, XFS_REFC_CRC_MAGIC,
+					i, 0, agno,
+					XFS_BTREE_CRC_BLOCKS);
+	}
+
+	/*
+	 * run along leaf, setting up records.  as we have to switch
+	 * blocks, call the prop_refc_cursor routine to set up the new
+	 * pointers for the parent.  that can recurse up to the root
+	 * if required.  set the sibling pointers for leaf level here.
+	 */
+	error = init_refcount_cursor(agno, &refc_cur);
+	if (error)
+		do_error(
+_("Insufficient memory to construct refcount cursor."));
+	refc_rec = pop_slab_cursor(refc_cur);
+	lptr = &btree_curs->level[0];
+
+	for (i = 0; i < lptr->num_blocks; i++)  {
+		/*
+		 * block initialization, lay in block header
+		 */
+		lptr->buf_p->b_ops = &xfs_refcountbt_buf_ops;
+		bt_hdr = XFS_BUF_TO_BLOCK(lptr->buf_p);
+		memset(bt_hdr, 0, mp->m_sb.sb_blocksize);
+		libxfs_btree_init_block(mp, lptr->buf_p, XFS_REFC_CRC_MAGIC,
+					0, 0, agno,
+					XFS_BTREE_CRC_BLOCKS);
+
+		bt_hdr->bb_u.s.bb_leftsib = cpu_to_be32(lptr->prev_agbno);
+		bt_hdr->bb_numrecs = cpu_to_be16(lptr->num_recs_pb +
+							(lptr->modulo > 0));
+
+		if (lptr->modulo > 0)
+			lptr->modulo--;
+
+		if (lptr->num_recs_pb > 0)
+			prop_refc_cursor(mp, agno, btree_curs,
+					refc_rec->rc_startblock, 0);
+
+		bt_rec = (struct xfs_refcount_rec *)
+			  ((char *)bt_hdr + XFS_REFCOUNT_BLOCK_LEN);
+		for (j = 0; j < be16_to_cpu(bt_hdr->bb_numrecs); j++) {
+			ASSERT(refc_rec != NULL);
+			bt_rec[j].rc_startblock =
+					cpu_to_be32(refc_rec->rc_startblock);
+			bt_rec[j].rc_blockcount =
+					cpu_to_be32(refc_rec->rc_blockcount);
+			bt_rec[j].rc_refcount = cpu_to_be32(refc_rec->rc_refcount);
+
+			refc_rec = pop_slab_cursor(refc_cur);
+		}
+
+		if (refc_rec != NULL)  {
+			/*
+			 * get next leaf level block
+			 */
+			if (lptr->prev_buf_p != NULL)  {
+#ifdef XR_BLD_RL_TRACE
+				fprintf(stderr, "writing refcntbt agbno %u\n",
+					lptr->prev_agbno);
+#endif
+				ASSERT(lptr->prev_agbno != NULLAGBLOCK);
+				libxfs_writebuf(lptr->prev_buf_p, 0);
+			}
+			lptr->prev_buf_p = lptr->buf_p;
+			lptr->prev_agbno = lptr->agbno;
+			lptr->agbno = get_next_blockaddr(agno, 0, btree_curs);
+			bt_hdr->bb_u.s.bb_rightsib = cpu_to_be32(lptr->agbno);
+
+			lptr->buf_p = libxfs_getbuf(mp->m_dev,
+					XFS_AGB_TO_DADDR(mp, agno, lptr->agbno),
+					XFS_FSB_TO_BB(mp, 1));
+		}
+	}
+	free_slab_cursor(&refc_cur);
+}
+
 /*
  * build both the agf and the agfl for an agno given both
  * btree cursors.
@@ -1706,7 +1997,8 @@ build_agf_agfl(
 	struct bt_status	*bcnt_bt,
 	xfs_extlen_t		freeblks,	/* # free blocks in tree */
 	int			lostblocks,	/* # blocks that will be lost */
-	struct bt_status	*rmap_bt)
+	struct bt_status	*rmap_bt,
+	struct bt_status	*refcnt_bt)
 {
 	struct extent_tree_node	*ext_ptr;
 	struct xfs_buf		*agf_buf, *agfl_buf;
@@ -1750,6 +2042,10 @@ build_agf_agfl(
 	agf->agf_freeblks = cpu_to_be32(freeblks);
 	agf->agf_rmap_blocks = cpu_to_be32(rmap_bt->num_tot_blocks -
 			rmap_bt->num_free_blocks);
+	agf->agf_refcount_root = cpu_to_be32(refcnt_bt->root);
+	agf->agf_refcount_level = cpu_to_be32(refcnt_bt->num_levels);
+	agf->agf_refcount_blocks = cpu_to_be32(refcnt_bt->num_tot_blocks -
+			refcnt_bt->num_free_blocks);
 
 	/*
 	 * Count and record the number of btree blocks consumed if required.
@@ -1867,6 +2163,10 @@ build_agf_agfl(
 
 	ASSERT(be32_to_cpu(agf->agf_roots[XFS_BTNUM_BNOi]) !=
 		be32_to_cpu(agf->agf_roots[XFS_BTNUM_CNTi]));
+	ASSERT(be32_to_cpu(agf->agf_refcount_root) !=
+		be32_to_cpu(agf->agf_roots[XFS_BTNUM_BNOi]));
+	ASSERT(be32_to_cpu(agf->agf_refcount_root) !=
+		be32_to_cpu(agf->agf_roots[XFS_BTNUM_CNTi]));
 
 	libxfs_writebuf(agf_buf, 0);
 
@@ -1936,6 +2236,7 @@ phase5_func(
 	bt_status_t	ino_btree_curs;
 	bt_status_t	fino_btree_curs;
 	bt_status_t	rmap_btree_curs;
+	bt_status_t	refcnt_btree_curs;
 	int		extra_blocks = 0;
 	uint		num_freeblocks;
 	xfs_extlen_t	freeblks1;
@@ -1998,6 +2299,12 @@ phase5_func(
 		 */
 		init_rmapbt_cursor(mp, agno, &rmap_btree_curs);
 
+		/*
+		 * Set up the btree cursors for the on-disk refcount btrees,
+		 * which includes pre-allocating all required blocks.
+		 */
+		init_refc_cursor(mp, agno, &refcnt_btree_curs);
+
 		num_extents = count_bno_extents_blocks(agno, &num_freeblocks);
 		/*
 		 * lose two blocks per AG -- the space tree roots
@@ -2091,12 +2398,17 @@ phase5_func(
 					rmap_btree_curs.num_free_blocks) - 1;
 		}
 
+		if (xfs_sb_version_hasreflink(&mp->m_sb)) {
+			build_refcount_tree(mp, agno, &refcnt_btree_curs);
+			write_cursor(&refcnt_btree_curs);
+		}
+
 		/*
 		 * set up agf and agfl
 		 */
 		build_agf_agfl(mp, agno, &bno_btree_curs,
 				&bcnt_btree_curs, freeblks1, extra_blocks,
-				&rmap_btree_curs);
+				&rmap_btree_curs, &refcnt_btree_curs);
 		/*
 		 * build inode allocation tree.
 		 */
@@ -2127,6 +2439,8 @@ phase5_func(
 		finish_cursor(&ino_btree_curs);
 		if (xfs_sb_version_hasrmapbt(&mp->m_sb))
 			finish_cursor(&rmap_btree_curs);
+		if (xfs_sb_version_hasreflink(&mp->m_sb))
+			finish_cursor(&refcnt_btree_curs);
 		if (xfs_sb_version_hasfinobt(&mp->m_sb))
 			finish_cursor(&fino_btree_curs);
 		finish_cursor(&bcnt_btree_curs);

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 67/71] xfs_repair: complain about copy-on-write leftovers

[Darrick J. Wong]

Complain about leftover CoW allocations that are hanging off the
refcount btree.  These are cleaned out at mount time, but we could be
louder about flagging down evidence of trouble.

Since these extents aren't "owned" by anything, we'll free them up by
reconstructing the free space btrees.

v2: When we're processing rmap records, we inadvertently forgot to
handle the CoW owner, so the leftover CoW staging blocks got marked as
file data.  These blocks will just get freed later, so mark them
"CoW".  When we process the refcountbt, complain about leftovers if
the type is unknown or "CoW".

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 db/check.c      |   21 +++++++++++++++++----
 repair/incore.h |    3 ++-
 repair/scan.c   |   26 +++++++++++++++++++++++++-
 3 files changed, 44 insertions(+), 6 deletions(-)


diff --git a/db/check.c b/db/check.c
index 5b90182..7392852 100644
--- a/db/check.c
+++ b/db/check.c
@@ -45,7 +45,7 @@ typedef enum {
 	DBM_LOG,	DBM_MISSING,	DBM_QUOTA,	DBM_RTBITMAP,
 	DBM_RTDATA,	DBM_RTFREE,	DBM_RTSUM,	DBM_SB,
 	DBM_SYMLINK,	DBM_BTFINO,	DBM_BTRMAP,	DBM_BTREFC,
-	DBM_RLDATA,
+	DBM_RLDATA,	DBM_COWDATA,
 	DBM_NDBM
 } dbm_t;
 
@@ -4731,9 +4731,22 @@ scanfunc_refcnt(
 		rp = XFS_REFCOUNT_REC_ADDR(block, 1);
 		lastblock = 0;
 		for (i = 0; i < be16_to_cpu(block->bb_numrecs); i++) {
-			set_dbmap(seqno, be32_to_cpu(rp[i].rc_startblock),
-				be32_to_cpu(rp[i].rc_blockcount), DBM_RLDATA,
-				seqno, bno);
+			if (be32_to_cpu(rp[i].rc_refcount) == 1) {
+				dbprintf(_(
+		"leftover CoW extent (%u/%u) len %u\n"),
+					seqno,
+					be32_to_cpu(rp[i].rc_startblock),
+					be32_to_cpu(rp[i].rc_blockcount));
+				set_dbmap(seqno,
+					be32_to_cpu(rp[i].rc_startblock),
+					be32_to_cpu(rp[i].rc_blockcount),
+					DBM_COWDATA, seqno, bno);
+			} else {
+				set_dbmap(seqno,
+					be32_to_cpu(rp[i].rc_startblock),
+					be32_to_cpu(rp[i].rc_blockcount),
+					DBM_RLDATA, seqno, bno);
+			}
 			if (be32_to_cpu(rp[i].rc_startblock) < lastblock) {
 				dbprintf(_(
 		"out-of-order refcnt btree record %d (%u %u) block %u/%u\n"),
diff --git a/repair/incore.h b/repair/incore.h
index bcd2f4b..c23a3a3 100644
--- a/repair/incore.h
+++ b/repair/incore.h
@@ -107,7 +107,8 @@ typedef struct rt_extent_tree_node  {
 #define XR_E_INO1	10	/* used by inodes (marked by rmap btree) */
 #define XR_E_FS_MAP1	11	/* used by fs space/inode maps (rmap btree) */
 #define XR_E_REFC	12	/* used by fs ag reference count btree */
-#define XR_E_BAD_STATE	13
+#define XR_E_COW	13	/* leftover cow extent */
+#define XR_E_BAD_STATE	14
 
 /* separate state bit, OR'ed into high (4th) bit of ex_state field */
 
diff --git a/repair/scan.c b/repair/scan.c
index 1c60784..499253e 100644
--- a/repair/scan.c
+++ b/repair/scan.c
@@ -813,6 +813,9 @@ process_rmap_rec(
 		case XFS_RMAP_OWN_REFC:
 			set_bmap_ext(agno, b, blen, XR_E_REFC);
 			break;
+		case XFS_RMAP_OWN_COW:
+			set_bmap_ext(agno, b, blen, XR_E_COW);
+			break;
 		case XFS_RMAP_OWN_NULL:
 			/* still unknown */
 			break;
@@ -1310,7 +1313,28 @@ _("%s btree block claimed (state %d), agno %d, bno %d, suspect %d\n"),
 				continue;
 			}
 
-			if (nr < 2 || nr > MAXREFCOUNT) {
+			if (nr == 1) {
+				xfs_agblock_t	c;
+				xfs_extlen_t	cnr;
+
+				for (c = b; c < end; c += cnr) {
+					state = get_bmap_ext(agno, c, end, &cnr);
+					switch (state) {
+					case XR_E_UNKNOWN:
+					case XR_E_COW:
+						do_warn(
+_("leftover CoW extent (%u/%u) len %u\n"),
+						agno, c, cnr);
+						set_bmap_ext(agno, c, cnr, XR_E_FREE);
+						break;
+					default:
+						do_warn(
+_("extent (%u/%u) len %u claimed, state is %d\n"),
+						agno, c, cnr, state);
+						break;
+					}
+				}
+			} else if (nr < 2 || nr > MAXREFCOUNT) {
 				do_warn(
 	_("invalid reference count %u in record %u of %s btree block %u/%u\n"),
 					nr, i, name, agno, bno);

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 68/71] xfs_repair: check the CoW extent size hint

[Darrick J. Wong]

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 repair/dinode.c |   55 +++++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 55 insertions(+)


diff --git a/repair/dinode.c b/repair/dinode.c
index 64fc983..11b60ce 100644
--- a/repair/dinode.c
+++ b/repair/dinode.c
@@ -2531,6 +2531,38 @@ _("bad (negative) size %" PRId64 " on inode %" PRIu64 "\n"),
 			goto clear_bad_out;
 		}
 
+		if ((flags2 & XFS_DIFLAG2_COWEXTSIZE) &&
+		    !xfs_sb_version_hasreflink(&mp->m_sb)) {
+			if (!uncertain) {
+				do_warn(
+	_("inode %" PRIu64 " has CoW extent size hint but file system does not support reflink\n"),
+					lino);
+			}
+			flags2 &= ~XFS_DIFLAG2_COWEXTSIZE;
+		}
+
+		if (flags2 & XFS_DIFLAG2_COWEXTSIZE) {
+			/* must be a directory or file */
+			if (di_mode && !S_ISDIR(di_mode) && !S_ISREG(di_mode)) {
+				if (!uncertain) {
+					do_warn(
+	_("CoW extent size flag set on non-file, non-directory inode %" PRIu64 "\n" ),
+						lino);
+				}
+				flags2 &= ~XFS_DIFLAG2_COWEXTSIZE;
+			}
+		}
+
+		if ((flags2 & XFS_DIFLAG2_COWEXTSIZE) &&
+		    (flags & (XFS_DIFLAG_REALTIME | XFS_DIFLAG_RTINHERIT))) {
+			if (!uncertain) {
+				do_warn(
+	_("Cannot have CoW extent size hint on a realtime inode %" PRIu64 "\n"),
+					lino);
+			}
+			flags2 &= ~XFS_DIFLAG2_COWEXTSIZE;
+		}
+
 		if (!verify_mode && flags2 != be64_to_cpu(dino->di_flags2)) {
 			if (!no_modify) {
 				do_warn(_("fixing bad flags2.\n"));
@@ -2624,6 +2656,29 @@ _("bad non-zero extent size %u for non-realtime/extsize inode %" PRIu64 ", "),
 	}
 
 	/*
+	 * Only (regular files and directories) with COWEXTSIZE flags
+	 * set can have extsize set.
+	 */
+	if (dino->di_version >= 3 &&
+	    be32_to_cpu(dino->di_cowextsize) != 0) {
+		if ((type == XR_INO_DIR || type == XR_INO_DATA) &&
+		    (be64_to_cpu(dino->di_flags2) &
+					XFS_DIFLAG2_COWEXTSIZE)) {
+			/* s'okay */ ;
+		} else {
+			do_warn(
+_("Cannot have non-zero CoW extent size %u on non-cowextsize inode %" PRIu64 ", "),
+					be32_to_cpu(dino->di_cowextsize), lino);
+			if (!no_modify)  {
+				do_warn(_("resetting to zero\n"));
+				dino->di_cowextsize = 0;
+				*dirty = 1;
+			} else
+				do_warn(_("would reset to zero\n"));
+		}
+	}
+
+	/*
 	 * general size/consistency checks:
 	 */
 	if (process_check_inode_sizes(mp, dino, lino, type) != 0)

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 69/71] xfs_repair: use range query when while checking rmaps

[Darrick J. Wong]

For shared extents, we ought to use a range query on the rmapbt to
find the corresponding rmap.  However, most of the time the observed
rmap will be an exact match for the rmapbt rmap, in which case we
could have used the (much faster) regular lookup.  Therefore, try the
regular lookup first and resort to the range lookup if that doesn't
get us what we want.  This can cut the run time of the rmap check of
xfs_repair in half.

Theoretically, the only reason why an observed rmap wouldn't be an
exact match for an rmapbt rmap is because we modified some file on
account of a metadata error.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 libxfs/libxfs_api_defs.h |    1 +
 repair/rmap.c            |   26 ++++++++++++++++++++++++++
 2 files changed, 27 insertions(+)


diff --git a/libxfs/libxfs_api_defs.h b/libxfs/libxfs_api_defs.h
index 28bac11..8026522 100644
--- a/libxfs/libxfs_api_defs.h
+++ b/libxfs/libxfs_api_defs.h
@@ -140,5 +140,6 @@
 #define xfs_refcountbt_init_cursor	libxfs_refcountbt_init_cursor
 #define xfs_refcount_lookup_le		libxfs_refcount_lookup_le
 #define xfs_refcount_get_rec		libxfs_refcount_get_rec
+#define xfs_rmap_lookup_le_range	libxfs_rmap_lookup_le_range
 
 #endif /* __LIBXFS_API_DEFS_H__ */
diff --git a/repair/rmap.c b/repair/rmap.c
index c53d6b7..597655b 100644
--- a/repair/rmap.c
+++ b/repair/rmap.c
@@ -909,6 +909,20 @@ rmap_lookup(
 	return -libxfs_rmap_get_rec(bt_cur, tmp, have);
 }
 
+/* Look for an rmap in the rmapbt that matches a given rmap. */
+static int
+rmap_lookup_overlapped(
+	struct xfs_btree_cur	*bt_cur,
+	struct xfs_rmap_irec	*rm_rec,
+	struct xfs_rmap_irec	*tmp,
+	int			*have)
+{
+	/* Have to use our fancy version for overlapped */
+	return -libxfs_rmap_lookup_le_range(bt_cur, rm_rec->rm_startblock,
+				rm_rec->rm_owner, rm_rec->rm_offset,
+				rm_rec->rm_flags, tmp, have);
+}
+
 /* Does the btree rmap cover the observed rmap? */
 #define NEXTP(x)	((x)->rm_startblock + (x)->rm_blockcount)
 #define NEXTL(x)	((x)->rm_offset + (x)->rm_blockcount)
@@ -997,6 +1011,18 @@ rmaps_verify_btree(
 		error = rmap_lookup(bt_cur, rm_rec, &tmp, &have);
 		if (error)
 			goto err;
+		/*
+		 * Using the range query is expensive, so only do it if
+		 * the regular lookup doesn't find anything or if it doesn't
+		 * match the observed rmap.
+		 */
+		if (xfs_sb_version_hasreflink(&bt_cur->bc_mp->m_sb) &&
+				(!have || !rmap_is_good(rm_rec, &tmp))) {
+			error = rmap_lookup_overlapped(bt_cur, rm_rec,
+					&tmp, &have);
+			if (error)
+				goto err;
+		}
 		if (!have) {
 			do_warn(
 _("Missing reverse-mapping record for (%u/%u) %slen %u owner %"PRId64" \

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 70/71] xfs_repair: check for mergeable refcount records

[Darrick J. Wong]

Make sure there aren't adjacent refcount records that could be merged;
this is a sign that the refcount tree algorithms aren't working
correctly.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 repair/scan.c |   32 +++++++++++++++++++++++++++++++-
 1 file changed, 31 insertions(+), 1 deletion(-)


diff --git a/repair/scan.c b/repair/scan.c
index 499253e..f4d6f89 100644
--- a/repair/scan.c
+++ b/repair/scan.c
@@ -1215,6 +1215,12 @@ out:
 		rmap_avoid_check();
 }
 
+struct refc_priv {
+	struct xfs_refcount_irec	last_rec;
+	xfs_agblock_t			nr_blocks;
+};
+
+
 static void
 scan_refcbt(
 	struct xfs_btree_block	*block,
@@ -1234,6 +1240,7 @@ scan_refcbt(
 	int			numrecs;
 	int			state;
 	xfs_agblock_t		lastblock = 0;
+	struct refc_priv	*refc_priv = priv;
 
 	if (magic != XFS_REFC_CRC_MAGIC) {
 		name = "(unknown)";
@@ -1258,6 +1265,8 @@ scan_refcbt(
 			goto out;
 	}
 
+	refc_priv->nr_blocks++;
+
 	/* check for btree blocks multiply claimed */
 	state = get_bmap(agno, bno);
 	if (!(state == XR_E_UNKNOWN || state == XR_E_REFC))  {
@@ -1349,6 +1358,20 @@ _("extent (%u/%u) len %u claimed, state is %d\n"),
 				lastblock = b;
 			}
 
+			/* Is this record mergeable with the last one? */
+			if (refc_priv->last_rec.rc_startblock +
+			    refc_priv->last_rec.rc_blockcount == b &&
+			    refc_priv->last_rec.rc_refcount == nr) {
+				do_warn(
+	_("record %d in block (%u/%u) of %s tree should be merged with previous record\n"),
+					i, agno, bno, name);
+				refc_priv->last_rec.rc_blockcount += len;
+			} else {
+				refc_priv->last_rec.rc_startblock = b;
+				refc_priv->last_rec.rc_blockcount = len;
+				refc_priv->last_rec.rc_refcount = nr;
+			}
+
 			/* XXX: probably want to mark the reflinked areas? */
 		}
 		goto out;
@@ -2192,10 +2215,17 @@ validate_agf(
 	if (xfs_sb_version_hasreflink(&mp->m_sb)) {
 		bno = be32_to_cpu(agf->agf_refcount_root);
 		if (bno != 0 && verify_agbno(mp, agno, bno)) {
+			struct refc_priv	priv;
+
+			memset(&priv, 0, sizeof(priv));
 			scan_sbtree(bno,
 				    be32_to_cpu(agf->agf_refcount_level),
 				    agno, 0, scan_refcbt, 1, XFS_REFC_CRC_MAGIC,
-				    agcnts, &xfs_refcountbt_buf_ops);
+				    &priv, &xfs_refcountbt_buf_ops);
+			if (be32_to_cpu(agf->agf_refcount_blocks) != priv.nr_blocks)
+				do_warn(_("bad refcountbt block count %u, saw %u\n"),
+					priv.nr_blocks,
+					be32_to_cpu(agf->agf_refcount_blocks));
 		} else  {
 			do_warn(_("bad agbno %u for refcntbt root, agno %d\n"),
 				bno, agno);

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

[PATCH 71/71] mkfs.xfs: format reflink enabled filesystems

[Darrick J. Wong]

Create the refcount btree at mkfs time and set the feature flag.

v2: Turn on the reflink feature when calculating the minimum log size.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 include/xfs_multidisk.h  |    3 +-
 libxfs/libxfs_api_defs.h |    1 +
 man/man8/mkfs.xfs.8      |   28 ++++++++++++++++++
 mkfs/maxtrres.c          |    5 +++
 mkfs/xfs_mkfs.c          |   70 ++++++++++++++++++++++++++++++++++++++++++----
 5 files changed, 99 insertions(+), 8 deletions(-)


diff --git a/include/xfs_multidisk.h b/include/xfs_multidisk.h
index 8dc3027..ce9bbce 100644
--- a/include/xfs_multidisk.h
+++ b/include/xfs_multidisk.h
@@ -68,6 +68,7 @@ extern void res_failed (int err);
 /* maxtrres.c */
 extern int max_trans_res(unsigned long agsize, int crcs_enabled, int dirversion,
 		int sectorlog, int blocklog, int inodelog, int dirblocklog,
-		int logversion, int log_sunit, int finobt, int rmapbt);
+		int logversion, int log_sunit, int finobt, int rmapbt,
+		int reflink);
 
 #endif	/* __XFS_MULTIDISK_H__ */
diff --git a/libxfs/libxfs_api_defs.h b/libxfs/libxfs_api_defs.h
index 8026522..8d57a66 100644
--- a/libxfs/libxfs_api_defs.h
+++ b/libxfs/libxfs_api_defs.h
@@ -141,5 +141,6 @@
 #define xfs_refcount_lookup_le		libxfs_refcount_lookup_le
 #define xfs_refcount_get_rec		libxfs_refcount_get_rec
 #define xfs_rmap_lookup_le_range	libxfs_rmap_lookup_le_range
+#define xfs_refc_block			libxfs_refc_block
 
 #endif /* __LIBXFS_API_DEFS_H__ */
diff --git a/man/man8/mkfs.xfs.8 b/man/man8/mkfs.xfs.8
index 9578c4d..6dab46a 100644
--- a/man/man8/mkfs.xfs.8
+++ b/man/man8/mkfs.xfs.8
@@ -213,6 +213,34 @@ for filesystems created with the (default)
 option set. When the option
 .B \-m crc=0
 is used, the reverse mapping btree feature is not supported and is disabled.
+.TP
+.BI reflink= value
+This option enables the use of a separate reference count btree index in each
+allocation group. The value is either 0 to disable the feature, or 1 to create
+a reference count btree in each allocation group.
+.IP
+The reference count btree enables the sharing of physical extents between
+the data forks of different files, which is commonly known as "reflink".
+Unlike traditional Unix filesystems which assume that every inode and
+logical block pair map to a unique physical block, a reflink-capable
+XFS filesystem removes the uniqueness requirement, allowing up to four
+billion arbitrary inode/logical block pairs to map to a physical block.
+If a program tries to write to a multiply-referenced block in a file, the write
+will be redirected to a new block, and that file's logical-to-physical
+mapping will be changed to the new block ("copy on write").  This feature
+enables the creation of per-file snapshots and deduplication.  It is only
+available for the data forks of regular files.
+.IP
+By default,
+.B mkfs.xfs
+will not create reference count btrees and therefore will not enable the
+reflink feature.  This feature is only available for filesystems created with
+the (default)
+.B \-m crc=1
+option set. When the option
+.B \-m crc=0
+is used, the reference count btree feature is not supported and reflink is
+disabled.
 .RE
 .TP
 .BI \-d " data_section_options"
diff --git a/mkfs/maxtrres.c b/mkfs/maxtrres.c
index d7978b6..fba7818 100644
--- a/mkfs/maxtrres.c
+++ b/mkfs/maxtrres.c
@@ -39,7 +39,8 @@ max_trans_res(
 	int		logversion,
 	int		log_sunit,
 	int		finobt,
-	int		rmapbt)
+	int		rmapbt,
+	int		reflink)
 {
 	xfs_sb_t	*sbp;
 	xfs_mount_t	mount;
@@ -75,6 +76,8 @@ max_trans_res(
 		sbp->sb_features_ro_compat |= XFS_SB_FEAT_RO_COMPAT_FINOBT;
 	if (rmapbt)
 		sbp->sb_features_ro_compat |= XFS_SB_FEAT_RO_COMPAT_RMAPBT;
+	if (reflink)
+		sbp->sb_features_ro_compat |= XFS_SB_FEAT_RO_COMPAT_REFLINK;
 
 	libxfs_mount(&mount, sbp, 0,0,0,0);
 	maxfsb = libxfs_log_calc_minimum_size(&mount);
diff --git a/mkfs/xfs_mkfs.c b/mkfs/xfs_mkfs.c
index 580119e..fc565c0 100644
--- a/mkfs/xfs_mkfs.c
+++ b/mkfs/xfs_mkfs.c
@@ -672,6 +672,8 @@ struct opt_params mopts = {
 		"uuid",
 #define M_RMAPBT	3
 		"rmapbt",
+#define M_REFLINK	4
+		"reflink",
 		NULL
 	},
 	.subopt_params = {
@@ -697,6 +699,12 @@ struct opt_params mopts = {
 		  .maxval = 1,
 		  .defaultval = 0,
 		},
+		{ .index = M_REFLINK,
+		  .conflicts = { LAST_CONFLICT },
+		  .minval = 0,
+		  .maxval = 1,
+		  .defaultval = 0,
+		},
 	},
 };
 
@@ -1155,6 +1163,7 @@ struct sb_feat_args {
 	bool	dirftype;
 	bool	parent_pointers;
 	bool	rmapbt;
+	bool	reflink;
 };
 
 static void
@@ -1227,6 +1236,8 @@ sb_set_features(
 		sbp->sb_features_ro_compat = XFS_SB_FEAT_RO_COMPAT_FINOBT;
 	if (fp->rmapbt)
 		sbp->sb_features_ro_compat |= XFS_SB_FEAT_RO_COMPAT_RMAPBT;
+	if (fp->reflink)
+		sbp->sb_features_ro_compat |= XFS_SB_FEAT_RO_COMPAT_REFLINK;
 
 	/*
 	 * Sparse inode chunk support has two main inode alignment requirements.
@@ -1488,6 +1499,7 @@ main(
 		.dirftype = true,
 		.parent_pointers = false,
 		.rmapbt = false,
+		.reflink = false,
 	};
 
 	platform_uuid_generate(&uuid);
@@ -1776,6 +1788,10 @@ main(
 					sb_feat.rmapbt = getnum(
 						value, &mopts, M_RMAPBT);
 					break;
+				case M_REFLINK:
+					sb_feat.reflink = getnum(
+						value, &mopts, M_REFLINK);
+					break;
 				default:
 					unknown('m', value);
 				}
@@ -2115,6 +2131,13 @@ _("rmapbt not supported without CRC support\n"));
 			usage();
 		}
 		sb_feat.rmapbt = false;
+
+		if (sb_feat.reflink) {
+			fprintf(stderr,
+_("reflink not supported without CRC support\n"));
+			usage();
+		}
+		sb_feat.reflink = false;
 	}
 
 
@@ -2599,7 +2622,7 @@ an AG size that is one stripe unit smaller, for example %llu.\n"),
 				   sb_feat.crcs_enabled, sb_feat.dir_version,
 				   sectorlog, blocklog, inodelog, dirblocklog,
 				   sb_feat.log_version, lsunit, sb_feat.finobt,
-				   sb_feat.rmapbt);
+				   sb_feat.rmapbt, sb_feat.reflink);
 	ASSERT(min_logblocks);
 	min_logblocks = MAX(XFS_MIN_LOG_BLOCKS, min_logblocks);
 	if (!logsize && dblocks >= (1024*1024*1024) >> blocklog)
@@ -2734,7 +2757,7 @@ _("size %s specified for log subvolume is too large, maximum is %lld blocks\n"),
 		printf(_(
 		   "meta-data=%-22s isize=%-6d agcount=%lld, agsize=%lld blks\n"
 		   "         =%-22s sectsz=%-5u attr=%u, projid32bit=%u\n"
-		   "         =%-22s crc=%-8u finobt=%u, sparse=%u, rmapbt=%u\n"
+		   "         =%-22s crc=%-8u finobt=%u, sparse=%u, rmapbt=%u, reflink=%u\n"
 		   "data     =%-22s bsize=%-6u blocks=%llu, imaxpct=%u\n"
 		   "         =%-22s sunit=%-6u swidth=%u blks\n"
 		   "naming   =version %-14u bsize=%-6u ascii-ci=%d ftype=%d\n"
@@ -2745,7 +2768,7 @@ _("size %s specified for log subvolume is too large, maximum is %lld blocks\n"),
 			"", sectorsize, sb_feat.attr_version,
 				    !sb_feat.projid16bit,
 			"", sb_feat.crcs_enabled, sb_feat.finobt, sb_feat.spinodes,
-			sb_feat.rmapbt,
+			sb_feat.rmapbt, sb_feat.reflink,
 			"", blocksize, (long long)dblocks, imaxpct,
 			"", dsunit, dswidth,
 			sb_feat.dir_version, dirblocksize, sb_feat.nci,
@@ -2933,7 +2956,12 @@ _("size %s specified for log subvolume is too large, maximum is %lld blocks\n"),
 			agf->agf_levels[XFS_BTNUM_RMAPi] = cpu_to_be32(1);
 			agf->agf_rmap_blocks = cpu_to_be32(1);
 		}
-
+		if (xfs_sb_version_hasreflink(&mp->m_sb)) {
+			agf->agf_refcount_root = cpu_to_be32(
+					libxfs_refc_block(mp));
+			agf->agf_refcount_level = cpu_to_be32(1);
+			agf->agf_refcount_blocks = cpu_to_be32(1);
+		}
 		agf->agf_flfirst = 0;
 		agf->agf_fllast = cpu_to_be32(XFS_AGFL_SIZE(mp) - 1);
 		agf->agf_flcount = 0;
@@ -3102,6 +3130,24 @@ _("size %s specified for log subvolume is too large, maximum is %lld blocks\n"),
 		libxfs_writebuf(buf, LIBXFS_EXIT_ON_FAILURE);
 
 		/*
+		 * refcount btree root block
+		 */
+		if (xfs_sb_version_hasreflink(&mp->m_sb)) {
+			buf = libxfs_getbuf(mp->m_ddev_targp,
+					XFS_AGB_TO_DADDR(mp, agno,
+						libxfs_refc_block(mp)),
+					bsize);
+			buf->b_ops = &xfs_refcountbt_buf_ops;
+
+			block = XFS_BUF_TO_BLOCK(buf);
+			memset(block, 0, blocksize);
+			libxfs_btree_init_block(mp, buf, XFS_REFC_CRC_MAGIC, 0,
+						0, agno, XFS_BTREE_CRC_BLOCKS);
+
+			libxfs_writebuf(buf, LIBXFS_EXIT_ON_FAILURE);
+		}
+
+		/*
 		 * INO btree root block
 		 */
 		buf = libxfs_getbuf(mp->m_ddev_targp,
@@ -3189,9 +3235,21 @@ _("size %s specified for log subvolume is too large, maximum is %lld blocks\n"),
 			rrec->rm_offset = 0;
 			be16_add_cpu(&block->bb_numrecs, 1);
 
+			/* account for refcount btree root */
+			if (xfs_sb_version_hasreflink(&mp->m_sb)) {
+				rrec = XFS_RMAP_REC_ADDR(block, 5);
+				rrec->rm_startblock = cpu_to_be32(
+							libxfs_refc_block(mp));
+				rrec->rm_blockcount = cpu_to_be32(1);
+				rrec->rm_owner = cpu_to_be64(XFS_RMAP_OWN_REFC);
+				rrec->rm_offset = 0;
+				be16_add_cpu(&block->bb_numrecs, 1);
+			}
+
 			/* account for the log space */
 			if (loginternal && agno == logagno) {
-				rrec = XFS_RMAP_REC_ADDR(block, 5);
+				rrec = XFS_RMAP_REC_ADDR(block,
+					be16_to_cpu(block->bb_numrecs) + 1);
 				rrec->rm_startblock = cpu_to_be32(
 						XFS_FSB_TO_AGBNO(mp, logstart));
 				rrec->rm_blockcount = cpu_to_be32(logblocks);
@@ -3446,7 +3504,7 @@ usage( void )
 {
 	fprintf(stderr, _("Usage: %s\n\
 /* blocksize */		[-b log=n|size=num]\n\
-/* metadata */		[-m crc=0|1,finobt=0|1,uuid=xxx,rmapbt=0|1]\n\
+/* metadata */		[-m crc=0|1,finobt=0|1,uuid=xxx,rmapbt=0|1,reflink=0|1]\n\
 /* data subvol */	[-d agcount=n,agsize=n,file,name=xxx,size=num,\n\
 			    (sunit=value,swidth=value|su=num,sw=num|noalign),\n\
 			    sectlog=n|sectsize=num\n\

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs