On Mon, Mar 07, 2016 at 08:49:46AM +1100, Dave Chinner wrote: > From: "Darrick J. Wong" <darrick.wong@xxxxxxxxxx> > > We need to create a new ioend if the current writepage call isn't > logically contiguous with the range contained in the previous ioend. > Hopefully writepage gets called in order of increasing file offset. This looks reasonable, but how did we manage to get away without this for so long? I think as-is we do not actually require it - for setting i_size we just care about the highest offset, and for unwritten extent conversion we just need the lowest and highest offset, and we were making use of that fact in direct I/O extensively before I rewrote that code not to use ioends. So this looks fine to me, but the description could use some better wording. Signed-off-by: Christoph Hellwig <hch@xxxxxx> _______________________________________________ xfs mailing list xfs@xxxxxxxxxxx http://oss.sgi.com/mailman/listinfo/xfs
