[Cluster-devel] [5.15 REGRESSION] iomap: Fix inline extent handling in iomap_readpage

[Andreas Gruenbacher <agruenba@redhat.com>]

On Wed, Nov 10, 2021 at 1:55 PM Christoph Hellwig <hch@lst.de> wrote:
> On Wed, Nov 10, 2021 at 12:38:42PM +0100, Andreas Gruenbacher wrote:
> > +     if (iomap->type == IOMAP_INLINE) {
> > +             /*
> > +              * The filesystem sets iomap->length to the size of the inline
> > +              * data.  We're at the end of the file, so we know that the
> > +              * rest of the page needs to be zeroed out.
> > +              */
> > +             iomap->length = iomap_read_inline_data(iter, page);
> > +             return iomap->length;
>
> You can't just change iomap->length here.  Fix the file system to
> return the right length, please.

Hmm, that doesn't make sense to me: the filesystem doesn't know that
iomap_readpage will pad to page boundaries. This happens at the iomap
layer, so the iomap layer should also deal with the consequences.
We're using different alignment rules here and for direct I/O, so that
makes fake-aligning the extent size in iomap_begin even more
questionable.

"Fixing" the extent size the filesystem returns would also break
direct I/O. We could add some additional padding code to
iomap_dio_inline_iter to deal with that, but currently, there's no
need for that.

Thanks,
Andreas