xfs: map an inode's offset to an exact physical block

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>
Reviewed-by: Christoph Hellwig <hch@lst.de>

2 files changed, 77 insertions, 1 deletions

diff --git a/fs/xfs/libxfs/xfs_bmap.c b/fs/xfs/libxfs/xfs_bmap.c
index 907d7b8df7bb..1ec3087bbb97 100644
--- a/fs/xfs/libxfs/xfs_bmap.c
+++ b/fs/xfs/libxfs/xfs_bmap.c
@@ -3877,6 +3877,63 @@ 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);
+	args.agbno = XFS_FSB_TO_AGBNO(mp, bno);
+	if (args.agno >= mp->m_sb.sb_agcount ||
+	    args.agbno >= mp->m_sb.sb_agblocks)
+		return -EFSCORRUPTED;
+
+	/* "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.datatype = ap->datatype;
+	args.pag = xfs_perag_get(args.mp, args.agno);
+	ASSERT(args.pag);
+
+	/*
+	 * The freelist fixing code will decline the allocation if
+	 * the size and shape of the free space doesn't allow for
+	 * allocating the extent and updating all the metadata that
+	 * happens during an allocation.  We're remapping, not
+	 * allocating, so skip that check by pretending to be freeing.
+	 */
+	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.
  */
@@ -3884,6 +3941,8 @@ 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) &&
 	    xfs_alloc_is_userdata(ap->datatype))
 		return xfs_bmap_rtalloc(ap);
@@ -4442,6 +4501,9 @@ xfs_bmapi_write(
 	ASSERT(len > 0);
 	ASSERT(XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_LOCAL);
 	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
+	ASSERT(!(flags & XFS_BMAPI_REMAP) || whichfork == XFS_DATA_FORK);
+	ASSERT(!(flags & XFS_BMAPI_PREALLOC) || !(flags & XFS_BMAPI_REMAP));
+	ASSERT(!(flags & XFS_BMAPI_CONVERT) || !(flags & XFS_BMAPI_REMAP));
 
 	/* zeroing is for currently only for data extents, not metadata */
 	ASSERT((flags & (XFS_BMAPI_METADATA | XFS_BMAPI_ZERO)) !=
@@ -4503,6 +4565,12 @@ 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/fs/xfs/libxfs/xfs_bmap.h b/fs/xfs/libxfs/xfs_bmap.h
index 7219207d58eb..8e69b5329629 100644
--- a/fs/xfs/libxfs/xfs_bmap.h
+++ b/fs/xfs/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)