On Mon, Nov 28, 2016 at 11:15:51AM -0800, Darrick J. Wong wrote:
> On Mon, Nov 28, 2016 at 02:09:36PM -0500, Brian Foster wrote:
> > The XFS reflink implementation adds a copy-on-write inode fork to track
> > newly allocated extents used to replace shared blocks on write. While,
> > in principle, these extents are tracked by the cow fork temporarily,
> > fragmentation avoidance mechanisms like the cowextsize hint and COW fork
> > speculative preallocation allocate additional blocks outside of the
> > range of the write. This means that blocks in the COW fork can linger
> > for some time until written to and remapped to the data fork or reaped
> > by the background cow fork reclaimer.
> >
> > To facilitate development and debugging, define and wire up a fiemap
> > flag to query the cow fork extent list of an inode. Note that fiemap
> > triggers writeback, which means all COW fork extents that are the target
> > of I/O are remapped to the data fork as part of the query. As a result,
> > the cow fork fiemap request returns only the blocks that have been
> > allocated and not yet written to or reclaimed.
>
> I don't think it's a good idea to expose implementation details through
> a VFS level API. Since this is only for debugging purposes...
>
> $ xfs_io -c 'bmap -c' /somefile
>
> ...should suffice and doesn't have the side effect of flushing dirty
> file data.
>
I didn't realize we had a bmap variant available. That suits my needs
just fine, but I do see the following on a quick test against a
reflinked, partially COW'd file:
# xfs_io -c "bmap -c" /mnt/file
xfs_io: xfsctl(XFS_IOC_GETBMAPX) iflags=0x28 ["/mnt/file"]: Invalid
argument
Is this under construction or expected to work as is?
Brian
> FWIW, all the other fiemap implementations need to reject
> FIEMAP_FLAG_COW if they don't know what that is.
>
> --D
>
> >
> > Signed-off-by: Brian Foster <bfoster@xxxxxxxxxx>
> > ---
> > fs/xfs/xfs_iomap.c | 45 +++++++++++++++++++++++++++++++++++++++++++++
> > fs/xfs/xfs_iomap.h | 1 +
> > fs/xfs/xfs_iops.c | 4 ++++
> > include/uapi/linux/fiemap.h | 1 +
> > 4 files changed, 51 insertions(+)
> >
> > diff --git a/fs/xfs/xfs_iomap.c b/fs/xfs/xfs_iomap.c
> > index 15a83813..4f46f49 100644
> > --- a/fs/xfs/xfs_iomap.c
> > +++ b/fs/xfs/xfs_iomap.c
> > @@ -1159,3 +1159,48 @@ xfs_xattr_iomap_begin(
> > struct iomap_ops xfs_xattr_iomap_ops = {
> > .iomap_begin = xfs_xattr_iomap_begin,
> > };
> > +
> > +static int
> > +xfs_cow_iomap_begin(
> > + struct inode *inode,
> > + loff_t offset,
> > + loff_t length,
> > + unsigned flags,
> > + struct iomap *iomap)
> > +{
> > + struct xfs_inode *ip = XFS_I(inode);
> > + struct xfs_mount *mp = ip->i_mount;
> > + xfs_fileoff_t offset_fsb = XFS_B_TO_FSBT(mp, offset);
> > + xfs_fileoff_t end_fsb = XFS_B_TO_FSB(mp, offset + length);
> > + struct xfs_bmbt_irec imap;
> > + int error = 0;
> > + int nimaps = 1;
> > + unsigned lockmode;
> > +
> > + if (XFS_FORCED_SHUTDOWN(mp))
> > + return -EIO;
> > +
> > + lockmode = xfs_ilock_data_map_shared(ip);
> > +
> > + if (!xfs_is_reflink_inode(ip)) {
> > + error = -ENOENT;
> > + goto out_unlock;
> > + }
> > +
> > + error = xfs_bmapi_read(ip, offset_fsb, end_fsb - offset_fsb, &imap,
> > + &nimaps, XFS_BMAPI_ENTIRE | XFS_BMAPI_COWFORK);
> > +
> > +out_unlock:
> > + xfs_iunlock(ip, lockmode);
> > +
> > + if (!error) {
> > + ASSERT(nimaps);
> > + xfs_bmbt_to_iomap(ip, iomap, &imap);
> > + }
> > +
> > + return error;
> > +}
> > +
> > +struct iomap_ops xfs_cow_iomap_ops = {
> > + .iomap_begin = xfs_cow_iomap_begin,
> > +};
> > diff --git a/fs/xfs/xfs_iomap.h b/fs/xfs/xfs_iomap.h
> > index 6d45cf0..69b62e4 100644
> > --- a/fs/xfs/xfs_iomap.h
> > +++ b/fs/xfs/xfs_iomap.h
> > @@ -35,5 +35,6 @@ xfs_extlen_t xfs_eof_alignment(struct xfs_inode *ip, xfs_extlen_t extsize);
> >
> > extern struct iomap_ops xfs_iomap_ops;
> > extern struct iomap_ops xfs_xattr_iomap_ops;
> > +extern struct iomap_ops xfs_cow_iomap_ops;
> >
> > #endif /* __XFS_IOMAP_H__*/
> > diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c
> > index 405a65c..517eeed 100644
> > --- a/fs/xfs/xfs_iops.c
> > +++ b/fs/xfs/xfs_iops.c
> > @@ -1043,6 +1043,10 @@ xfs_vn_fiemap(
> > fieinfo->fi_flags &= ~FIEMAP_FLAG_XATTR;
> > error = iomap_fiemap(inode, fieinfo, start, length,
> > &xfs_xattr_iomap_ops);
> > + } else if (fieinfo->fi_flags & FIEMAP_FLAG_COW) {
> > + fieinfo->fi_flags &= ~FIEMAP_FLAG_COW;
> > + error = iomap_fiemap(inode, fieinfo, start, length,
> > + &xfs_cow_iomap_ops);
> > } else {
> > error = iomap_fiemap(inode, fieinfo, start, length,
> > &xfs_iomap_ops);
> > diff --git a/include/uapi/linux/fiemap.h b/include/uapi/linux/fiemap.h
> > index 0c51d61..7014b4c 100644
> > --- a/include/uapi/linux/fiemap.h
> > +++ b/include/uapi/linux/fiemap.h
> > @@ -41,6 +41,7 @@ struct fiemap {
> > #define FIEMAP_FLAG_SYNC 0x00000001 /* sync file data before map */
> > #define FIEMAP_FLAG_XATTR 0x00000002 /* map extended attribute tree */
> > #define FIEMAP_FLAG_CACHE 0x00000004 /* request caching of the extents */
> > +#define FIEMAP_FLAG_COW 0x00000010 /* map cow fork extents */
> >
> > #define FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_XATTR)
> >
> > --
> > 2.7.4
> >
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
> > the body of a message to majordomo@xxxxxxxxxxxxxxx
> > More majordomo info at http://vger.kernel.org/majordomo-info.html
> --
> To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
