On Fri, May 24, 2019 at 09:04:45AM +1000, Dave Chinner wrote: > > +STATIC void > > +xlog_write_iclog( > > + struct xlog *log, > > + struct xlog_in_core *iclog, > > + struct xfs_buf *bp, > > + uint64_t bno, > > + bool flush) > > Can you rename this to need_flush here and in xlog_sync()? I kept > having to check whether it meant "we need a flush" or "we've > already done a flush" while reading the patch. Ok. > But on the extra buffer, we don't set it's I/O length at all.... > > Oh, the setting of the IO length is hidden inside > xfs_buf_associate_memory(). Can you add a comment indicating that > this is why we omit the setting of the io length for the extra > buffer? Well, for one this is how we've always done it, and second we remove this whole thing and xfs_buf_associate_memory a few patches down the road. I'd rather not bother writing a comment here.
