On Tue, Jun 10, 2025 at 10:19:35AM -0400, Brian Foster wrote: > > On thing that the batch would be extremely useful for is making > > iomap_file_unshare not totally suck by reading in all folios for a > > range (not just the dirty ones) similar to the filemap_read path > > instead of synchronously reading one block at a time. > > > > I can add it to the list to look into. On a quick look though any reason > we wouldn't want to just invoke readahead or something somewhere in that > loop, particularly if that is mainly a performance issue..? I was planning to look into it once your series lands. Yes, doing readahead is the main thing. But once we start doing that we might as well try to reuse the entire folio_batch optimization done in the file read path, which has shown to be pretty effective.
