On Thu, Dec 19, 2024 at 01:08:15PM -0500, Jeff Layton wrote:
> > + tmp = copy_folio_from_iter_atomic(folio, folio_offset, bytes, ii);
>
> Just to save someone else going down the same rabbit hole:
>
> copy_folio_from_iter_atomic() is defined as:
>
> static inline size_t copy_folio_from_iter_atomic(struct folio *folio,
> size_t offset, size_t bytes, struct iov_iter *i)
> {
> return copy_page_from_iter_atomic(&folio->page, offset, bytes, i);
> }
>
> ...which _looks_ sort of like it's not fully baked yet and can't handle
> a large folio, but it turns out that copy_page_from_iter_atomic() can
> handle compound pages, so I think this is actually OK. Whew!
Yes, this is fine. I'd love to clean this up further. If someone wants
the kudos of doing that, ntfs_compress_write() is your challenge. It's
the only remaining caller of copy_page_from_iter_atomic() and
once it's converted to use folios, we can push folios deeper into
the iov_iter code.
