From: Darrick J. Wong <djwong@xxxxxxxxxx>
Add the necessary alignment checking code to the reflink remap code to
ensure that remap requests are aligned to rt extent boundaries if the
realtime extent size isn't a power of two. The VFS helpers assume that
they can use the usual (blocksize - 1) masking to avoid slow 64-bit
division, but since XFS is special we won't make everyone pay that cost
for our weird edge case.
Signed-off-by: "Darrick J. Wong" <djwong@xxxxxxxxxx>
---
fs/xfs/xfs_reflink.c | 119 +++++++++++++++++++++++++++++++++++++-------------
fs/xfs/xfs_reflink.h | 2 -
fs/xfs/xfs_rtalloc.c | 4 --
fs/xfs/xfs_super.c | 9 ----
4 files changed, 90 insertions(+), 44 deletions(-)
diff --git a/fs/xfs/xfs_reflink.c b/fs/xfs/xfs_reflink.c
index 0222b78dedd92d..6ceb00565bab24 100644
--- a/fs/xfs/xfs_reflink.c
+++ b/fs/xfs/xfs_reflink.c
@@ -1638,6 +1638,83 @@ xfs_reflink_adjust_rtbigalloc_len(
# define xfs_reflink_adjust_rtbigalloc_len(...) (0)
#endif /* CONFIG_XFS_RT */
+/*
+ * Check the alignment of a remap request when the allocation unit size isn't a
+ * power of two. The VFS helpers use (fast) bitmask-based alignment checks,
+ * but here we have to use slow long division.
+ */
+static int
+xfs_reflink_remap_check_rtalign(
+ struct xfs_inode *ip_in,
+ loff_t pos_in,
+ struct xfs_inode *ip_out,
+ loff_t pos_out,
+ loff_t *req_len,
+ unsigned int remap_flags)
+{
+ struct xfs_mount *mp = ip_in->i_mount;
+ uint32_t rextbytes;
+ loff_t in_size, out_size;
+ loff_t new_length, length = *req_len;
+ loff_t blen;
+
+ rextbytes = XFS_FSB_TO_B(mp, mp->m_sb.sb_rextsize);
+ in_size = i_size_read(VFS_I(ip_in));
+ out_size = i_size_read(VFS_I(ip_out));
+
+ /* The start of both ranges must be aligned to a rt extent. */
+ if (!isaligned_64(pos_in, rextbytes) ||
+ !isaligned_64(pos_out, rextbytes))
+ return -EINVAL;
+
+ if (length == 0)
+ length = in_size - pos_in;
+
+ /*
+ * If the user wanted us to exchange up to the infile's EOF, round up
+ * to the next block boundary for this check.
+ *
+ * Otherwise, reject the range length if it's not extent aligned. We
+ * already confirmed the starting offsets' extent alignment.
+ */
+ if (pos_in + length == in_size)
+ blen = roundup_64(in_size, rextbytes) - pos_in;
+ else
+ blen = rounddown_64(length, rextbytes);
+
+ /* Don't allow overlapped remappings within the same file. */
+ if (ip_in == ip_out &&
+ pos_out + blen > pos_in &&
+ pos_in + blen > pos_out)
+ return -EINVAL;
+
+ /*
+ * Ensure that we don't exchange a partial EOF extent into the middle
+ * of another file.
+ */
+ if (isaligned_64(length, rextbytes))
+ return 0;
+
+ new_length = length;
+ if (pos_out + length < out_size)
+ new_length = rounddown_64(new_length, rextbytes);
+
+ if (new_length == length)
+ return 0;
+
+ /*
+ * Return the shortened request if the caller permits it. If the
+ * request was shortened to zero rt extents, we know that the original
+ * arguments weren't valid in the first place.
+ */
+ if ((remap_flags & REMAP_FILE_CAN_SHORTEN) && new_length > 0) {
+ *req_len = new_length;
+ return 0;
+ }
+
+ return (remap_flags & REMAP_FILE_DEDUP) ? -EBADE : -EINVAL;
+}
+
/*
* Prepare two files for range cloning. Upon a successful return both inodes
* will have the iolock and mmaplock held, the page cache of the out file will
@@ -1681,6 +1758,7 @@ xfs_reflink_remap_prep(
struct inode *inode_out = file_inode(file_out);
struct xfs_inode *dest = XFS_I(inode_out);
const struct iomap_ops *dax_read_ops = NULL;
+ unsigned int alloc_unit = xfs_inode_alloc_unitsize(dest);
int ret;
/* Lock both files against IO */
@@ -1698,14 +1776,22 @@ xfs_reflink_remap_prep(
if (IS_DAX(inode_in) != IS_DAX(inode_out))
goto out_unlock;
- ASSERT(is_power_of_2(xfs_inode_alloc_unitsize(dest)));
+ /* Check non-power of two alignment issues, if necessary. */
+ if (XFS_IS_REALTIME_INODE(dest) && !is_power_of_2(alloc_unit)) {
+ ret = xfs_reflink_remap_check_rtalign(src, pos_in, dest,
+ pos_out, len, remap_flags);
+ if (ret)
+ goto out_unlock;
+
+ /* Do the VFS checks with the regular block alignment. */
+ alloc_unit = src->i_mount->m_sb.sb_blocksize;
+ }
if (IS_DAX(inode_in))
dax_read_ops = &xfs_read_iomap_ops;
ret = __generic_remap_file_range_prep(file_in, pos_in, file_out,
- pos_out, len, remap_flags, dax_read_ops,
- xfs_inode_alloc_unitsize(dest));
+ pos_out, len, remap_flags, dax_read_ops, alloc_unit);
if (ret || *len == 0)
goto out_unlock;
@@ -1949,30 +2035,3 @@ xfs_reflink_unshare(
trace_xfs_reflink_unshare_error(ip, error, _RET_IP_);
return error;
}
-
-/*
- * Can we use reflink with this realtime extent size? Note that we don't check
- * for rblocks > 0 here because this can be called as part of attaching a new
- * rt section.
- */
-bool
-xfs_reflink_supports_rextsize(
- struct xfs_mount *mp,
- unsigned int rextsize)
-{
- /* reflink on the realtime device requires rtgroups */
- if (!xfs_has_rtgroups(mp))
- return false;
-
- /*
- * Reflink doesn't support file allocation units larger than a single
- * block and not a power of two because we would have to perform
- * CoW-around for unaligned write requests to guarantee that we always
- * remap entire allocation units and the reflink code cannot yet handle
- * rounding ranges to align to non powers of two.
- */
- if (!is_power_of_2(rextsize))
- return false;
-
- return true;
-}
diff --git a/fs/xfs/xfs_reflink.h b/fs/xfs/xfs_reflink.h
index cc4e92278279b6..3bfd7ab9e1148a 100644
--- a/fs/xfs/xfs_reflink.h
+++ b/fs/xfs/xfs_reflink.h
@@ -62,6 +62,4 @@ extern int xfs_reflink_remap_blocks(struct xfs_inode *src, loff_t pos_in,
extern int xfs_reflink_update_dest(struct xfs_inode *dest, xfs_off_t newlen,
xfs_extlen_t cowextsize, unsigned int remap_flags);
-bool xfs_reflink_supports_rextsize(struct xfs_mount *mp, unsigned int rextsize);
-
#endif /* __XFS_REFLINK_H */
diff --git a/fs/xfs/xfs_rtalloc.c b/fs/xfs/xfs_rtalloc.c
index d8e6d073d64dc9..586da450cc44b4 100644
--- a/fs/xfs/xfs_rtalloc.c
+++ b/fs/xfs/xfs_rtalloc.c
@@ -1295,9 +1295,7 @@ xfs_growfs_rt(
goto out_unlock;
if (xfs_has_reflink(mp))
goto out_unlock;
- } else if (xfs_has_reflink(mp) &&
- !xfs_reflink_supports_rextsize(mp, in->extsize))
- goto out_unlock;
+ }
error = xfs_sb_validate_fsb_count(&mp->m_sb, in->newblocks);
if (error)
diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
index c91b9467a3eef8..8050fea541140a 100644
--- a/fs/xfs/xfs_super.c
+++ b/fs/xfs/xfs_super.c
@@ -1754,15 +1754,6 @@ xfs_fs_fill_super(
xfs_warn_experimental(mp, XFS_EXPERIMENTAL_METADIR);
if (xfs_has_reflink(mp)) {
- if (xfs_has_realtime(mp) &&
- !xfs_reflink_supports_rextsize(mp, mp->m_sb.sb_rextsize)) {
- xfs_alert(mp,
- "reflink not compatible with non-power-of-2 realtime extent size %u!",
- mp->m_sb.sb_rextsize);
- error = -EINVAL;
- goto out_filestream_unmount;
- }
-
/*
* always-cow mode is not supported on filesystems with rt
* extent sizes larger than a single block because we'd have
