1 files changed, 351 insertions, 233 deletions

diff --git a/fs/xfs/xfs_buf_item.c b/fs/xfs/xfs_buf_item.c
index 45df2b857d48..bfc4e0c26fd3 100644
--- a/fs/xfs/xfs_buf_item.c
+++ b/fs/xfs/xfs_buf_item.c
@@ -37,109 +37,6 @@ static inline struct xfs_buf_log_item *BUF_ITEM(struct xfs_log_item *lip)
 	return container_of(lip, struct xfs_buf_log_item, bli_item);
 }
 
-
-#ifdef XFS_TRANS_DEBUG
-/*
- * This function uses an alternate strategy for tracking the bytes
- * that the user requests to be logged.  This can then be used
- * in conjunction with the bli_orig array in the buf log item to
- * catch bugs in our callers' code.
- *
- * We also double check the bits set in xfs_buf_item_log using a
- * simple algorithm to check that every byte is accounted for.
- */
-STATIC void
-xfs_buf_item_log_debug(
-	xfs_buf_log_item_t	*bip,
-	uint			first,
-	uint			last)
-{
-	uint	x;
-	uint	byte;
-	uint	nbytes;
-	uint	chunk_num;
-	uint	word_num;
-	uint	bit_num;
-	uint	bit_set;
-	uint	*wordp;
-
-	ASSERT(bip->bli_logged != NULL);
-	byte = first;
-	nbytes = last - first + 1;
-	bfset(bip->bli_logged, first, nbytes);
-	for (x = 0; x < nbytes; x++) {
-		chunk_num = byte >> XFS_BLF_SHIFT;
-		word_num = chunk_num >> BIT_TO_WORD_SHIFT;
-		bit_num = chunk_num & (NBWORD - 1);
-		wordp = &(bip->bli_format.blf_data_map[word_num]);
-		bit_set = *wordp & (1 << bit_num);
-		ASSERT(bit_set);
-		byte++;
-	}
-}
-
-/*
- * This function is called when we flush something into a buffer without
- * logging it.  This happens for things like inodes which are logged
- * separately from the buffer.
- */
-void
-xfs_buf_item_flush_log_debug(
-	xfs_buf_t	*bp,
-	uint		first,
-	uint		last)
-{
-	xfs_buf_log_item_t	*bip = bp->b_fspriv;
-	uint			nbytes;
-
-	if (bip == NULL || (bip->bli_item.li_type != XFS_LI_BUF))
-		return;
-
-	ASSERT(bip->bli_logged != NULL);
-	nbytes = last - first + 1;
-	bfset(bip->bli_logged, first, nbytes);
-}
-
-/*
- * This function is called to verify that our callers have logged
- * all the bytes that they changed.
- *
- * It does this by comparing the original copy of the buffer stored in
- * the buf log item's bli_orig array to the current copy of the buffer
- * and ensuring that all bytes which mismatch are set in the bli_logged
- * array of the buf log item.
- */
-STATIC void
-xfs_buf_item_log_check(
-	xfs_buf_log_item_t	*bip)
-{
-	char		*orig;
-	char		*buffer;
-	int		x;
-	xfs_buf_t	*bp;
-
-	ASSERT(bip->bli_orig != NULL);
-	ASSERT(bip->bli_logged != NULL);
-
-	bp = bip->bli_buf;
-	ASSERT(bp->b_length > 0);
-	ASSERT(bp->b_addr != NULL);
-	orig = bip->bli_orig;
-	buffer = bp->b_addr;
-	for (x = 0; x < BBTOB(bp->b_length); x++) {
-		if (orig[x] != buffer[x] && !btst(bip->bli_logged, x)) {
-			xfs_emerg(bp->b_mount,
-				"%s: bip %x buffer %x orig %x index %d",
-				__func__, bip, bp, orig, x);
-			ASSERT(0);
-		}
-	}
-}
-#else
-#define		xfs_buf_item_log_debug(x,y,z)
-#define		xfs_buf_item_log_check(x)
-#endif
-
 STATIC void	xfs_buf_do_callbacks(struct xfs_buf *bp);
 
 /*
@@ -153,33 +50,25 @@ STATIC void	xfs_buf_do_callbacks(struct xfs_buf *bp);
  * If the XFS_BLI_STALE flag has been set, then log nothing.
  */
 STATIC uint
-xfs_buf_item_size(
-	struct xfs_log_item	*lip)
+xfs_buf_item_size_segment(
+	struct xfs_buf_log_item	*bip,
+	struct xfs_buf_log_format *blfp)
 {
-	struct xfs_buf_log_item	*bip = BUF_ITEM(lip);
 	struct xfs_buf		*bp = bip->bli_buf;
 	uint			nvecs;
 	int			next_bit;
 	int			last_bit;
 
-	ASSERT(atomic_read(&bip->bli_refcount) > 0);
-	if (bip->bli_flags & XFS_BLI_STALE) {
-		/*
-		 * The buffer is stale, so all we need to log
-		 * is the buf log format structure with the
-		 * cancel flag in it.
-		 */
-		trace_xfs_buf_item_size_stale(bip);
-		ASSERT(bip->bli_format.blf_flags & XFS_BLF_CANCEL);
-		return 1;
-	}
+	last_bit = xfs_next_bit(blfp->blf_data_map, blfp->blf_map_size, 0);
+	if (last_bit == -1)
+		return 0;
+
+	/*
+	 * initial count for a dirty buffer is 2 vectors - the format structure
+	 * and the first dirty region.
+	 */
+	nvecs = 2;
 
-	ASSERT(bip->bli_flags & XFS_BLI_LOGGED);
-	nvecs = 1;
-	last_bit = xfs_next_bit(bip->bli_format.blf_data_map,
-					 bip->bli_format.blf_map_size, 0);
-	ASSERT(last_bit != -1);
-	nvecs++;
 	while (last_bit != -1) {
 		/*
 		 * This takes the bit number to start looking from and
@@ -187,16 +76,15 @@ xfs_buf_item_size(
 		 * if there are no more bits set or the start bit is
 		 * beyond the end of the bitmap.
 		 */
-		next_bit = xfs_next_bit(bip->bli_format.blf_data_map,
-						 bip->bli_format.blf_map_size,
-						 last_bit + 1);
+		next_bit = xfs_next_bit(blfp->blf_data_map, blfp->blf_map_size,
+					last_bit + 1);
 		/*
 		 * If we run out of bits, leave the loop,
 		 * else if we find a new set of bits bump the number of vecs,
 		 * else keep scanning the current set of bits.
 		 */
 		if (next_bit == -1) {
-			last_bit = -1;
+			break;
 		} else if (next_bit != last_bit + 1) {
 			last_bit = next_bit;
 			nvecs++;
@@ -210,22 +98,83 @@ xfs_buf_item_size(
 		}
 	}
 
-	trace_xfs_buf_item_size(bip);
 	return nvecs;
 }
 
 /*
- * This is called to fill in the vector of log iovecs for the
- * given log buf item.  It fills the first entry with a buf log
- * format structure, and the rest point to contiguous chunks
- * within the buffer.
+ * This returns the number of log iovecs needed to log the given buf log item.
+ *
+ * It calculates this as 1 iovec for the buf log format structure and 1 for each
+ * stretch of non-contiguous chunks to be logged.  Contiguous chunks are logged
+ * in a single iovec.
+ *
+ * Discontiguous buffers need a format structure per region that that is being
+ * logged. This makes the changes in the buffer appear to log recovery as though
+ * they came from separate buffers, just like would occur if multiple buffers
+ * were used instead of a single discontiguous buffer. This enables
+ * discontiguous buffers to be in-memory constructs, completely transparent to
+ * what ends up on disk.
+ *
+ * If the XFS_BLI_STALE flag has been set, then log nothing but the buf log
+ * format structures.
  */
-STATIC void
-xfs_buf_item_format(
-	struct xfs_log_item	*lip,
-	struct xfs_log_iovec	*vecp)
+STATIC uint
+xfs_buf_item_size(
+	struct xfs_log_item	*lip)
 {
 	struct xfs_buf_log_item	*bip = BUF_ITEM(lip);
+	uint			nvecs;
+	int			i;
+
+	ASSERT(atomic_read(&bip->bli_refcount) > 0);
+	if (bip->bli_flags & XFS_BLI_STALE) {
+		/*
+		 * The buffer is stale, so all we need to log
+		 * is the buf log format structure with the
+		 * cancel flag in it.
+		 */
+		trace_xfs_buf_item_size_stale(bip);
+		ASSERT(bip->__bli_format.blf_flags & XFS_BLF_CANCEL);
+		return bip->bli_format_count;
+	}
+
+	ASSERT(bip->bli_flags & XFS_BLI_LOGGED);
+
+	if (bip->bli_flags & XFS_BLI_ORDERED) {
+		/*
+		 * The buffer has been logged just to order it.
+		 * It is not being included in the transaction
+		 * commit, so no vectors are used at all.
+		 */
+		trace_xfs_buf_item_size_ordered(bip);
+		return XFS_LOG_VEC_ORDERED;
+	}
+
+	/*
+	 * the vector count is based on the number of buffer vectors we have
+	 * dirty bits in. This will only be greater than one when we have a
+	 * compound buffer with more than one segment dirty. Hence for compound
+	 * buffers we need to track which segment the dirty bits correspond to,
+	 * and when we move from one segment to the next increment the vector
+	 * count for the extra buf log format structure that will need to be
+	 * written.
+	 */
+	nvecs = 0;
+	for (i = 0; i < bip->bli_format_count; i++) {
+		nvecs += xfs_buf_item_size_segment(bip, &bip->bli_formats[i]);
+	}
+
+	trace_xfs_buf_item_size(bip);
+	return nvecs;
+}
+
+static struct xfs_log_iovec *
+xfs_buf_item_format_segment(
+	struct xfs_buf_log_item	*bip,
+	struct xfs_log_iovec	*vecp,
+	uint			offset,
+	struct xfs_buf_log_format *blfp)
+{
 	struct xfs_buf	*bp = bip->bli_buf;
 	uint		base_size;
 	uint		nvecs;
@@ -235,40 +184,33 @@ xfs_buf_item_format(
 	uint		nbits;
 	uint		buffer_offset;
 
-	ASSERT(atomic_read(&bip->bli_refcount) > 0);
-	ASSERT((bip->bli_flags & XFS_BLI_LOGGED) ||
-	       (bip->bli_flags & XFS_BLI_STALE));
+	/* copy the flags across from the base format item */
+	blfp->blf_flags = bip->__bli_format.blf_flags;
 
 	/*
-	 * The size of the base structure is the size of the
-	 * declared structure plus the space for the extra words
-	 * of the bitmap.  We subtract one from the map size, because
-	 * the first element of the bitmap is accounted for in the
-	 * size of the base structure.
+	 * Base size is the actual size of the ondisk structure - it reflects
+	 * the actual size of the dirty bitmap rather than the size of the in
+	 * memory structure.
 	 */
-	base_size =
-		(uint)(sizeof(xfs_buf_log_format_t) +
-		       ((bip->bli_format.blf_map_size - 1) * sizeof(uint)));
-	vecp->i_addr = &bip->bli_format;
+	base_size = offsetof(struct xfs_buf_log_format, blf_data_map) +
+			(blfp->blf_map_size * sizeof(blfp->blf_data_map[0]));
+
+	nvecs = 0;
+	first_bit = xfs_next_bit(blfp->blf_data_map, blfp->blf_map_size, 0);
+	if (!(bip->bli_flags & XFS_BLI_STALE) && first_bit == -1) {
+		/*
+		 * If the map is not be dirty in the transaction, mark
+		 * the size as zero and do not advance the vector pointer.
+		 */
+		goto out;
+	}
+
+	vecp->i_addr = blfp;
 	vecp->i_len = base_size;
 	vecp->i_type = XLOG_REG_TYPE_BFORMAT;
 	vecp++;
 	nvecs = 1;
 
-	/*
-	 * If it is an inode buffer, transfer the in-memory state to the
-	 * format flags and clear the in-memory state. We do not transfer
-	 * this state if the inode buffer allocation has not yet been committed
-	 * to the log as setting the XFS_BLI_INODE_BUF flag will prevent
-	 * correct replay of the inode allocation.
-	 */
-	if (bip->bli_flags & XFS_BLI_INODE_BUF) {
-		if (!((bip->bli_flags & XFS_BLI_INODE_ALLOC_BUF) &&
-		      xfs_log_item_in_current_chkpt(lip)))
-			bip->bli_format.blf_flags |= XFS_BLF_INODE_BUF;
-		bip->bli_flags &= ~XFS_BLI_INODE_BUF;
-	}
-
 	if (bip->bli_flags & XFS_BLI_STALE) {
 		/*
 		 * The buffer is stale, so all we need to log
@@ -276,17 +218,15 @@ xfs_buf_item_format(
 		 * cancel flag in it.
 		 */
 		trace_xfs_buf_item_format_stale(bip);
-		ASSERT(bip->bli_format.blf_flags & XFS_BLF_CANCEL);
-		bip->bli_format.blf_size = nvecs;
-		return;
+		ASSERT(blfp->blf_flags & XFS_BLF_CANCEL);
+		goto out;
 	}
 
+
 	/*
 	 * Fill in an iovec for each set of contiguous chunks.
 	 */
-	first_bit = xfs_next_bit(bip->bli_format.blf_data_map,
-					 bip->bli_format.blf_map_size, 0);
-	ASSERT(first_bit != -1);
+
 	last_bit = first_bit;
 	nbits = 1;
 	for (;;) {
@@ -296,9 +236,8 @@ xfs_buf_item_format(
 		 * if there are no more bits set or the start bit is
 		 * beyond the end of the bitmap.
 		 */
-		next_bit = xfs_next_bit(bip->bli_format.blf_data_map,
-						 bip->bli_format.blf_map_size,
-						 (uint)last_bit + 1);
+		next_bit = xfs_next_bit(blfp->blf_data_map, blfp->blf_map_size,
+					(uint)last_bit + 1);
 		/*
 		 * If we run out of bits fill in the last iovec and get
 		 * out of the loop.
@@ -309,14 +248,14 @@ xfs_buf_item_format(
 		 * keep counting and scanning.
 		 */
 		if (next_bit == -1) {
-			buffer_offset = first_bit * XFS_BLF_CHUNK;
+			buffer_offset = offset + first_bit * XFS_BLF_CHUNK;
 			vecp->i_addr = xfs_buf_offset(bp, buffer_offset);
 			vecp->i_len = nbits * XFS_BLF_CHUNK;
 			vecp->i_type = XLOG_REG_TYPE_BCHUNK;
 			nvecs++;
 			break;
 		} else if (next_bit != last_bit + 1) {
-			buffer_offset = first_bit * XFS_BLF_CHUNK;
+			buffer_offset = offset + first_bit * XFS_BLF_CHUNK;
 			vecp->i_addr = xfs_buf_offset(bp, buffer_offset);
 			vecp->i_len = nbits * XFS_BLF_CHUNK;
 			vecp->i_type = XLOG_REG_TYPE_BCHUNK;
@@ -325,18 +264,16 @@ xfs_buf_item_format(
 			first_bit = next_bit;
 			last_bit = next_bit;
 			nbits = 1;
-		} else if (xfs_buf_offset(bp, next_bit << XFS_BLF_SHIFT) !=
-			   (xfs_buf_offset(bp, last_bit << XFS_BLF_SHIFT) +
+		} else if (xfs_buf_offset(bp, offset +
+					      (next_bit << XFS_BLF_SHIFT)) !=
+			   (xfs_buf_offset(bp, offset +
+					       (last_bit << XFS_BLF_SHIFT)) +
 			    XFS_BLF_CHUNK)) {
-			buffer_offset = first_bit * XFS_BLF_CHUNK;
+			buffer_offset = offset + first_bit * XFS_BLF_CHUNK;
 			vecp->i_addr = xfs_buf_offset(bp, buffer_offset);
 			vecp->i_len = nbits * XFS_BLF_CHUNK;
 			vecp->i_type = XLOG_REG_TYPE_BCHUNK;
-/* You would think we need to bump the nvecs here too, but we do not
- * this number is used by recovery, and it gets confused by the boundary
- * split here
- *			nvecs++;
- */
+			nvecs++;
 			vecp++;
 			first_bit = next_bit;
 			last_bit = next_bit;
@@ -346,13 +283,73 @@ xfs_buf_item_format(
 			nbits++;
 		}
 	}
-	bip->bli_format.blf_size = nvecs;
+out:
+	blfp->blf_size = nvecs;
+	return vecp;
+}
+
+/*
+ * This is called to fill in the vector of log iovecs for the
+ * given log buf item.  It fills the first entry with a buf log
+ * format structure, and the rest point to contiguous chunks
+ * within the buffer.
+ */
+STATIC void
+xfs_buf_item_format(
+	struct xfs_log_item	*lip,
+	struct xfs_log_iovec	*vecp)
+{
+	struct xfs_buf_log_item	*bip = BUF_ITEM(lip);
+	struct xfs_buf		*bp = bip->bli_buf;
+	uint			offset = 0;
+	int			i;
+
+	ASSERT(atomic_read(&bip->bli_refcount) > 0);
+	ASSERT((bip->bli_flags & XFS_BLI_LOGGED) ||
+	       (bip->bli_flags & XFS_BLI_STALE));
+
+	/*
+	 * If it is an inode buffer, transfer the in-memory state to the
+	 * format flags and clear the in-memory state.
+	 *
+	 * For buffer based inode allocation, we do not transfer
+	 * this state if the inode buffer allocation has not yet been committed
+	 * to the log as setting the XFS_BLI_INODE_BUF flag will prevent
+	 * correct replay of the inode allocation.
+	 *
+	 * For icreate item based inode allocation, the buffers aren't written
+	 * to the journal during allocation, and hence we should always tag the
+	 * buffer as an inode buffer so that the correct unlinked list replay
+	 * occurs during recovery.
+	 */
+	if (bip->bli_flags & XFS_BLI_INODE_BUF) {
+		if (xfs_sb_version_hascrc(&lip->li_mountp->m_sb) ||
+		    !((bip->bli_flags & XFS_BLI_INODE_ALLOC_BUF) &&
+		      xfs_log_item_in_current_chkpt(lip)))
+			bip->__bli_format.blf_flags |= XFS_BLF_INODE_BUF;
+		bip->bli_flags &= ~XFS_BLI_INODE_BUF;
+	}
+
+	if ((bip->bli_flags & (XFS_BLI_ORDERED|XFS_BLI_STALE)) ==
+							XFS_BLI_ORDERED) {
+		/*
+		 * The buffer has been logged just to order it.  It is not being
+		 * included in the transaction commit, so don't format it.
+		 */
+		trace_xfs_buf_item_format_ordered(bip);
+		return;
+	}
+
+	for (i = 0; i < bip->bli_format_count; i++) {
+		vecp = xfs_buf_item_format_segment(bip, vecp, offset,
+						&bip->bli_formats[i]);
+		offset += bp->b_maps[i].bm_len;
+	}
 
 	/*
 	 * Check to make sure everything is consistent.
 	 */
 	trace_xfs_buf_item_format(bip);
-	xfs_buf_item_log_check(bip);
 }
 
 /*
@@ -372,6 +369,7 @@ xfs_buf_item_pin(
 
 	ASSERT(atomic_read(&bip->bli_refcount) > 0);
 	ASSERT((bip->bli_flags & XFS_BLI_LOGGED) ||
+	       (bip->bli_flags & XFS_BLI_ORDERED) ||
 	       (bip->bli_flags & XFS_BLI_STALE));
 
 	trace_xfs_buf_item_pin(bip);
@@ -418,7 +416,7 @@ xfs_buf_item_unpin(
 		ASSERT(bip->bli_flags & XFS_BLI_STALE);
 		ASSERT(xfs_buf_islocked(bp));
 		ASSERT(XFS_BUF_ISSTALE(bp));
-		ASSERT(bip->bli_format.blf_flags & XFS_BLF_CANCEL);
+		ASSERT(bip->__bli_format.blf_flags & XFS_BLF_CANCEL);
 
 		trace_xfs_buf_item_unpin_stale(bip);
 
@@ -459,7 +457,25 @@ xfs_buf_item_unpin(
 		}
 		xfs_buf_relse(bp);
 	} else if (freed && remove) {
+		/*
+		 * There are currently two references to the buffer - the active
+		 * LRU reference and the buf log item. What we are about to do
+		 * here - simulate a failed IO completion - requires 3
+		 * references.
+		 *
+		 * The LRU reference is removed by the xfs_buf_stale() call. The
+		 * buf item reference is removed by the xfs_buf_iodone()
+		 * callback that is run by xfs_buf_do_callbacks() during ioend
+		 * processing (via the bp->b_iodone callback), and then finally
+		 * the ioend processing will drop the IO reference if the buffer
+		 * is marked XBF_ASYNC.
+		 *
+		 * Hence we need to take an additional reference here so that IO
+		 * completion processing doesn't free the buffer prematurely.
+		 */
 		xfs_buf_lock(bp);
+		xfs_buf_hold(bp);
+		bp->b_flags |= XBF_ASYNC;
 		xfs_buf_ioerror(bp, EIO);
 		XFS_BUF_UNDONE(bp);
 		xfs_buf_stale(bp);
@@ -478,8 +494,18 @@ xfs_buf_item_push(
 
 	if (xfs_buf_ispinned(bp))
 		return XFS_ITEM_PINNED;
-	if (!xfs_buf_trylock(bp))
+	if (!xfs_buf_trylock(bp)) {
+		/*
+		 * If we have just raced with a buffer being pinned and it has
+		 * been marked stale, we could end up stalling until someone else
+		 * issues a log force to unpin the stale buffer. Check for the
+		 * race condition here so xfsaild recognizes the buffer is pinned
+		 * and queues a log force to move it along.
+		 */
+		if (xfs_buf_ispinned(bp))
+			return XFS_ITEM_PINNED;
 		return XFS_ITEM_LOCKED;
+	}
 
 	ASSERT(!(bip->bli_flags & XFS_BLI_STALE));
 
@@ -516,8 +542,9 @@ xfs_buf_item_unlock(
 {
 	struct xfs_buf_log_item	*bip = BUF_ITEM(lip);
 	struct xfs_buf		*bp = bip->bli_buf;
-	int			aborted;
-	uint			hold;
+	bool			clean;
+	bool			aborted;
+	int			flags;
 
 	/* Clear the buffer's association with this transaction. */
 	bp->b_transp = NULL;
@@ -528,25 +555,23 @@ xfs_buf_item_unlock(
 	 * (cancelled) buffers at unpin time, but we'll never go through the
 	 * pin/unpin cycle if we abort inside commit.
 	 */
-	aborted = (lip->li_flags & XFS_LI_ABORTED) != 0;
-
+	aborted = (lip->li_flags & XFS_LI_ABORTED) ? true : false;
 	/*
-	 * Before possibly freeing the buf item, determine if we should
-	 * release the buffer at the end of this routine.
+	 * Before possibly freeing the buf item, copy the per-transaction state
+	 * so we can reference it safely later after clearing it from the
+	 * buffer log item.
 	 */
-	hold = bip->bli_flags & XFS_BLI_HOLD;
-
-	/* Clear the per transaction state. */
-	bip->bli_flags &= ~(XFS_BLI_LOGGED | XFS_BLI_HOLD);
+	flags = bip->bli_flags;
+	bip->bli_flags &= ~(XFS_BLI_LOGGED | XFS_BLI_HOLD | XFS_BLI_ORDERED);
 
 	/*
 	 * If the buf item is marked stale, then don't do anything.  We'll
 	 * unlock the buffer and free the buf item when the buffer is unpinned
 	 * for the last time.
 	 */
-	if (bip->bli_flags & XFS_BLI_STALE) {
+	if (flags & XFS_BLI_STALE) {
 		trace_xfs_buf_item_unlock_stale(bip);
-		ASSERT(bip->bli_format.blf_flags & XFS_BLF_CANCEL);
+		ASSERT(bip->__bli_format.blf_flags & XFS_BLF_CANCEL);
 		if (!aborted) {
 			atomic_dec(&bip->bli_refcount);
 			return;
@@ -557,15 +582,36 @@ xfs_buf_item_unlock(
 
 	/*
 	 * If the buf item isn't tracking any data, free it, otherwise drop the
-	 * reference we hold to it.
+	 * reference we hold to it. If we are aborting the transaction, this may
+	 * be the only reference to the buf item, so we free it anyway
+	 * regardless of whether it is dirty or not. A dirty abort implies a
+	 * shutdown, anyway.
+	 *
+	 * Ordered buffers are dirty but may have no recorded changes, so ensure
+	 * we only release clean items here.
 	 */
-	if (xfs_bitmap_empty(bip->bli_format.blf_data_map,
-			     bip->bli_format.blf_map_size))
+	clean = (flags & XFS_BLI_DIRTY) ? false : true;
+	if (clean) {
+		int i;
+		for (i = 0; i < bip->bli_format_count; i++) {
+			if (!xfs_bitmap_empty(bip->bli_formats[i].blf_data_map,
+				     bip->bli_formats[i].blf_map_size)) {
+				clean = false;
+				break;
+			}
+		}
+	}
+	if (clean)
 		xfs_buf_item_relse(bp);
-	else
+	else if (aborted) {
+		if (atomic_dec_and_test(&bip->bli_refcount)) {
+			ASSERT(XFS_FORCED_SHUTDOWN(lip->li_mountp));
+			xfs_buf_item_relse(bp);
+		}
+	} else
 		atomic_dec(&bip->bli_refcount);
 
-	if (!hold)
+	if (!(flags & XFS_BLI_HOLD))
 		xfs_buf_relse(bp);
 }
 
@@ -622,6 +668,35 @@ static const struct xfs_item_ops xfs_buf_item_ops = {
 	.iop_committing = xfs_buf_item_committing
 };
 
+STATIC int
+xfs_buf_item_get_format(
+	struct xfs_buf_log_item	*bip,
+	int			count)
+{
+	ASSERT(bip->bli_formats == NULL);
+	bip->bli_format_count = count;
+
+	if (count == 1) {
+		bip->bli_formats = &bip->__bli_format;
+		return 0;
+	}
+
+	bip->bli_formats = kmem_zalloc(count * sizeof(struct xfs_buf_log_format),
+				KM_SLEEP);
+	if (!bip->bli_formats)
+		return ENOMEM;
+	return 0;
+}
+
+STATIC void
+xfs_buf_item_free_format(
+	struct xfs_buf_log_item	*bip)
+{
+	if (bip->bli_formats != &bip->__bli_format) {
+		kmem_free(bip->bli_formats);
+		bip->bli_formats = NULL;
+	}
+}
 
 /*
  * Allocate a new buf log item to go with the given buffer.
@@ -639,6 +714,8 @@ xfs_buf_item_init(
 	xfs_buf_log_item_t	*bip;
 	int			chunks;
 	int			map_size;
+	int			error;
+	int			i;
 
 	/*
 	 * Check to see if there is already a buf log item for
@@ -650,25 +727,33 @@ xfs_buf_item_init(
 	if (lip != NULL && lip->li_type == XFS_LI_BUF)
 		return;
 
-	/*
-	 * chunks is the number of XFS_BLF_CHUNK size pieces
-	 * the buffer can be divided into. Make sure not to
-	 * truncate any pieces.  map_size is the size of the
-	 * bitmap needed to describe the chunks of the buffer.
-	 */
-	chunks = (int)((BBTOB(bp->b_length) + (XFS_BLF_CHUNK - 1)) >>
-								XFS_BLF_SHIFT);
-	map_size = (int)((chunks + NBWORD) >> BIT_TO_WORD_SHIFT);
-
-	bip = (xfs_buf_log_item_t*)kmem_zone_zalloc(xfs_buf_item_zone,
-						    KM_SLEEP);
+	bip = kmem_zone_zalloc(xfs_buf_item_zone, KM_SLEEP);
 	xfs_log_item_init(mp, &bip->bli_item, XFS_LI_BUF, &xfs_buf_item_ops);
 	bip->bli_buf = bp;
 	xfs_buf_hold(bp);
-	bip->bli_format.blf_type = XFS_LI_BUF;
-	bip->bli_format.blf_blkno = (__int64_t)XFS_BUF_ADDR(bp);
-	bip->bli_format.blf_len = (ushort)bp->b_length;
-	bip->bli_format.blf_map_size = map_size;
+
+	/*
+	 * chunks is the number of XFS_BLF_CHUNK size pieces the buffer
+	 * can be divided into. Make sure not to truncate any pieces.
+	 * map_size is the size of the bitmap needed to describe the
+	 * chunks of the buffer.
+	 *
+	 * Discontiguous buffer support follows the layout of the underlying
+	 * buffer. This makes the implementation as simple as possible.
+	 */
+	error = xfs_buf_item_get_format(bip, bp->b_map_count);
+	ASSERT(error == 0);
+
+	for (i = 0; i < bip->bli_format_count; i++) {
+		chunks = DIV_ROUND_UP(BBTOB(bp->b_maps[i].bm_len),
+				      XFS_BLF_CHUNK);
+		map_size = DIV_ROUND_UP(chunks, NBWORD);
+
+		bip->bli_formats[i].blf_type = XFS_LI_BUF;
+		bip->bli_formats[i].blf_blkno = bp->b_maps[i].bm_bn;
+		bip->bli_formats[i].blf_len = bp->b_maps[i].bm_len;
+		bip->bli_formats[i].blf_map_size = map_size;
+	}
 
 #ifdef XFS_TRANS_DEBUG
 	/*
@@ -699,10 +784,11 @@ xfs_buf_item_init(
  * item's bitmap.
  */
 void
-xfs_buf_item_log(
-	xfs_buf_log_item_t	*bip,
+xfs_buf_item_log_segment(
+	struct xfs_buf_log_item	*bip,
 	uint			first,
-	uint			last)
+	uint			last,
+	uint			*map)
 {
 	uint		first_bit;
 	uint		last_bit;
@@ -715,12 +801,6 @@ xfs_buf_item_log(
 	uint		mask;
 
 	/*
-	 * Mark the item as having some dirty data for
-	 * quick reference in xfs_buf_item_dirty.
-	 */
-	bip->bli_flags |= XFS_BLI_DIRTY;
-
-	/*
 	 * Convert byte offsets to bit numbers.
 	 */
 	first_bit = first >> XFS_BLF_SHIFT;
@@ -736,7 +816,7 @@ xfs_buf_item_log(
 	 * to set a bit in.
 	 */
 	word_num = first_bit >> BIT_TO_WORD_SHIFT;
-	wordp = &(bip->bli_format.blf_data_map[word_num]);
+	wordp = &map[word_num];
 
 	/*
 	 * Calculate the starting bit in the first word.
@@ -779,13 +859,50 @@ xfs_buf_item_log(
 		mask = (1 << end_bit) - 1;
 		*wordp |= mask;
 	}
+}
 
-	xfs_buf_item_log_debug(bip, first, last);
+/*
+ * Mark bytes first through last inclusive as dirty in the buf
+ * item's bitmap.
+ */
+void
+xfs_buf_item_log(
+	xfs_buf_log_item_t	*bip,
+	uint			first,
+	uint			last)
+{
+	int			i;
+	uint			start;
+	uint			end;
+	struct xfs_buf		*bp = bip->bli_buf;
+
+	/*
+	 * walk each buffer segment and mark them dirty appropriately.
+	 */
+	start = 0;
+	for (i = 0; i < bip->bli_format_count; i++) {
+		if (start > last)
+			break;
+		end = start + BBTOB(bp->b_maps[i].bm_len);
+		if (first > end) {
+			start += BBTOB(bp->b_maps[i].bm_len);
+			continue;
+		}
+		if (first < start)
+			first = start;
+		if (end > last)
+			end = last;
+
+		xfs_buf_item_log_segment(bip, first, end,
+					 &bip->bli_formats[i].blf_data_map[0]);
+
+		start += bp->b_maps[i].bm_len;
+	}
 }
 
 
 /*
- * Return 1 if the buffer has some data that has been logged (at any
+ * Return 1 if the buffer has been logged or ordered in a transaction (at any
  * point, not just the current transaction) and 0 if not.
  */
 uint
@@ -804,6 +921,7 @@ xfs_buf_item_free(
 	kmem_free(bip->bli_logged);
 #endif /* XFS_TRANS_DEBUG */
 
+	xfs_buf_item_free_format(bip);
 	kmem_zone_free(xfs_buf_item_zone, bip);
 }
 
@@ -818,11 +936,11 @@ void
 xfs_buf_item_relse(
 	xfs_buf_t	*bp)
 {
-	xfs_buf_log_item_t	*bip;
+	xfs_buf_log_item_t	*bip = bp->b_fspriv;
 
 	trace_xfs_buf_item_relse(bp, _RET_IP_);
+	ASSERT(!(bip->bli_item.li_flags & XFS_LI_IN_AIL));
 
-	bip = bp->b_fspriv;
 	bp->b_fspriv = bip->bli_item.li_bio_list;
 	if (bp->b_fspriv == NULL)
 		bp->b_iodone = NULL;
@@ -954,7 +1072,7 @@ xfs_buf_iodone_callbacks(
 
 		if (!XFS_BUF_ISSTALE(bp)) {
 			bp->b_flags |= XBF_WRITE | XBF_ASYNC | XBF_DONE;
-			xfs_bdstrat_cb(bp);
+			xfs_buf_iorequest(bp);
 		} else {
 			xfs_buf_relse(bp);
 		}