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@xxxxxxxxxx>
---
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 d3b2486..6277b53 100644
--- a/include/xfs_trace.h
+++ b/include/xfs_trace.h
@@ -256,6 +256,9 @@
#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 7ef1d18..58f730e 100644
--- a/libxfs/xfs_bmap.c
+++ b/libxfs/xfs_bmap.c
@@ -3864,6 +3864,55 @@ 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.
*/
@@ -3871,6 +3920,8 @@ 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);
@@ -4507,6 +4558,12 @@ 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)) !=
@@ -4568,6 +4625,12 @@ 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 62a66d0..fb2fd4c 100644
--- a/libxfs/xfs_bmap.h
+++ b/libxfs/xfs_bmap.h
@@ -97,6 +97,13 @@ struct xfs_bmap_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_bmap_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)
|