On Thu, May 09, 2024 at 08:05:24AM -0700, Christoph Hellwig wrote: > On Thu, May 09, 2024 at 07:32:50AM -0700, Darrick J. Wong wrote: > > On Thu, May 09, 2024 at 05:58:23AM -0700, Christoph Hellwig wrote: > > > On Thu, May 09, 2024 at 12:55:14PM +0000, Pankaj Raghav (Samsung) wrote: > > > > We might still fail here during mount. My question is: do we also fail > > > > the mount if folio_alloc fails? > > > > > > Yes. Like any other allocation that fails at mount time. > > > > How hard is it to fallback to regular zero-page if you can't allocate > > the zero-hugepage? > > We'd need the bio allocation and bio_add_page loop. Not the end > of the world, but also a bit annoying. If we do that we might as > well just do it unconditionally. > > > I think most sysadmins would rather mount with > > reduced zeroing performance than get an ENOMEM. > > If you don't have 2MB free for the zero huge folio, are you going to > do useful things with your large block size XFS file system which > only makes sense for giant storage sizes? Oh. Right, this is for bs>ps. For that case it makes sense to fail the mount. I was only thinking about bs<=ps with large folios, where it doesn't. (Would we use the zero-hugepage for large folios on a 4k fsblock fs?) --D
