On Thu, Dec 16, 2021 at 08:31:19AM -0800, Linus Torvalds wrote:
> On Thu, Dec 16, 2021 at 8:22 AM David Howells <dhowells@xxxxxxxxxx> wrote:
> >
> > With transparent huge pages, in the future, write_begin() and write_end()
> > may be passed a length parameter that, in combination with the offset into
> > the page, exceeds the length of that page. This allows
> > grab_cache_page_write_begin() to better choose the size of THP to allocate.
>
> I still think this is a fundamental bug in the caller. That
> "explanation" is weak, and the whole concept smells like week-old fish
> to me.
You're right that ->write_end can't be called with more bytes than fit
in the folio. That makes no sense at all.
I haven't finished fully fleshing this out yet (and as a result we still
only create PAGE_SIZE folios on writes), but my basic plan was:
generic_perform_write:
- bytes = min_t(unsigned long, PAGE_SIZE - offset,
+ bytes = min_t(unsigned long, FOLIO_MAX_PAGE_CACHE_SIZE - offset,
iov_iter_count(i));
...
status = a_ops->write_begin(file, mapping, pos, bytes, flags,
- &page, &fsdata);
+ &folio, &fsdata);
+ offset = offset_in_folio(folio, pos);
+ bytes = folio_size(folio - offset);
...
status = a_ops->write_end(file, mapping, pos, bytes, copied,
- page, fsdata);
+ folio, fsdata);
Since ->write_begin is the place where we actually create folios, it
needs to know what size folio to create. Unless you'd rather we do
something to actually create the folio before calling ->write_begin?
