On Sun, May 22, 2022 at 08:28:42AM -0700, Darrick J. Wong wrote: > From: Darrick J. Wong <djwong@xxxxxxxxxx> > > libxfs itself should never be messing with whether or not to enable > logging for extended attribute updates -- this decision should be made > on a case-by-case basis by libxfs callers. Move the code that actually > enables the log features to xfs_xattr.c, and adjust the callers. > > This removes an awkward coupling point between libxfs and what would be > libxlog, if the XFS log were actually its own library. Furthermore, it > makes bulk attribute updates and inode security initialization a tiny > bit more efficient, since they now avoid cycling the log feature between > every single xattr. > > Signed-off-by: Darrick J. Wong <djwong@xxxxxxxxxx> > --- > fs/xfs/libxfs/xfs_attr.c | 12 +------- > fs/xfs/xfs_acl.c | 10 +++++++ > fs/xfs/xfs_ioctl.c | 22 +++++++++++++--- > fs/xfs/xfs_ioctl.h | 2 + > fs/xfs/xfs_ioctl32.c | 4 ++- > fs/xfs/xfs_iops.c | 25 ++++++++++++++---- > fs/xfs/xfs_log.c | 45 -------------------------------- > fs/xfs/xfs_log.h | 1 - > fs/xfs/xfs_super.h | 2 + > fs/xfs/xfs_xattr.c | 65 ++++++++++++++++++++++++++++++++++++++++++++++ > 10 files changed, 120 insertions(+), 68 deletions(-) This seems like the wrong way to approach this. I would have defined a wrapper function for xfs_attr_set() to do the log state futzing, not moved it all into callers that don't need (or want) to know anything about how attrs are logged internally.... Cheers, Dave. -- Dave Chinner david@xxxxxxxxxxxxx
