On Thu, Feb 25, 2021 at 10:15:59AM +0100, Christoph Hellwig wrote:
> > diff --git a/fs/xfs/xfs_log_cil.c b/fs/xfs/xfs_log_cil.c
> > index e8c674b291f3..145f1e847f82 100644
> > --- a/fs/xfs/xfs_log_cil.c
> > +++ b/fs/xfs/xfs_log_cil.c
> > @@ -652,14 +652,22 @@ xlog_cil_process_committed(
> > }
> >
> > struct xlog_cil_trans_hdr {
> > + struct xlog_op_header oph[2];
> > struct xfs_trans_header thdr;
> > - struct xfs_log_iovec lhdr;
> > + struct xfs_log_iovec lhdr[2];
>
> oph and lhdr aren't really used as individual arrays. What about
> splitting them into separate fields and giving them descriptive names?
Because the next steps with this series are to start combining
adjacent regions in a log vector buffer into a single regions to
reduce the number of regions that we have to process in
xlog_write().
I did it this way here just to maintain the current "start rec is
spearate from trans_header" because of all the special treatment of
the start rec in the existing code. Now that all that special
treatment or start records in xlog_write() is gone, we can collapse
this down to just a single region again. I plan to do that in a
future patchset that is just aggregating regions where it is
appropriate for various log items.
Cheers,
Dave.
--
Dave Chinner
david@xxxxxxxxxxxxx
