On Fri, Jun 29, 2018 at 04:40:40PM +0200, Andreas Gruenbacher wrote: > On 29 June 2018 at 10:56, Christoph Hellwig <hch@xxxxxx> wrote: > > This looks generally fine. But I think it might be worth refactoring > > iomap_dio_actor a bit first, e.g. something like this new patch > > before yours, which would also nicely solve your alignmnet concern > > (entirely untested for now): > > This looks correct. I've rebased my patches on top of it and I ran the > xfstest auto group on gfs2 and xfs on top. > > Can you push this to your gfs2-iomap branch? I'll then repost an > updated version of "iomap: Direct I/O for inline data". Darrick now has a real iomap merge branch which replaced it. Before formally submitting the patch I'd also like to verify that it does not regress for XFS by doing a full xfstests run.
