On Tue, Sep 06, 2016 at 08:25:53AM -0700, Christoph Hellwig wrote: > FYI, this has some major conflicts with the iomap updates - we'll need > to know from Dave which one should go in first. While we're combining patches, I think it makes sense to modify _iomap_write_allocate so that when it's converting a data fork delalloc reservation into a real extent that it should look for an adjacent (file offset-wise) real extent in the CoW fork and feed that to the allocator in the hopes of allocating all the blocks in the same part of the disk. Or better yet as one big(ger) extent record. --D _______________________________________________ xfs mailing list xfs@xxxxxxxxxxx http://oss.sgi.com/mailman/listinfo/xfs
