Here's an updated patch queue that implements iomap in gfs2 and switches gfs2 fiemap over to use iomap. Following that is lseek SEEK_HOLE and SEEK_DATA support for gfs2 and xfs. Finally, xfs's internal use of lseek SEEK_HOLE / SEEK_DATA is converted to iomap. With the approach taken by this patch queue, filesystems that support iomap still need to implement fiemap and lseek SEEK_HOLE / SEEK_DATA separately. Those implementations will all be identical except for the filesystem specific locking / unlocking code at the beginning / end of the operations. Wouldn't it make sense to add iomap_lock and iomap_unlock operations, expose the iomap ops at the vfs layer, and implement fiemap and lseek SEEK_HOLE / SEEK_DATA generically in a next step? Thanks, Andreas Andreas Gruenbacher (4): vfs: Add iomap_seek_hole_data helper gfs2: Implement lseek SEEK_HOLE / SEEK_DATA via iomap xfs: Switch to iomap for lseek SEEK_HOLE / SEEK_DATA xfs: Switch to iomap for seeking in quota files Bob Peterson (3): GFS2: Make height info part of metapath GFS2: Implement iomap for block_map GFS2: Switch fiemap implementation to use iomap fs/gfs2/Kconfig | 1 + fs/gfs2/bmap.c | 288 ++++++++++++++++++++++++++++++------------ fs/gfs2/bmap.h | 4 + fs/gfs2/file.c | 14 ++- fs/gfs2/inode.c | 54 ++++++-- fs/gfs2/inode.h | 1 + fs/gfs2/trace_gfs2.h | 65 ++++++++++ fs/iomap.c | 81 ++++++++++++ fs/xfs/xfs_dquot.c | 5 +- fs/xfs/xfs_file.c | 337 +------------------------------------------------- fs/xfs/xfs_inode.h | 2 - include/linux/iomap.h | 9 +- 12 files changed, 429 insertions(+), 432 deletions(-) -- 2.7.5
