On 26.08.22 12:07, David Hildenbrand wrote: > On 26.08.22 12:02, Marc-André Lureau wrote: >> Hi >> >> >> On Fri, Aug 26, 2022 at 2:01 PM David Hildenbrand <[email protected] >> <mailto:[email protected]>> wrote: >> >> On 25.08.22 15:21, [email protected] >> <mailto:[email protected]> wrote: >> > From: Marc-André Lureau <[email protected] >> <mailto:[email protected]>> >> > >> > Rewrite get_next_page() to work over non-aligned blocks. When it >> > encounters non aligned addresses, it will allocate a zero-page and try >> > to fill it. >> >> Could we simplify by using one global helper page (or caller provided >> page) and avoiding the allocation/freeing? >> >> >> I don't think that makes a big difference, but certainly doable. > > If we're using one central page, I guess we'd have to pass "flag_sync = > true" to write_cache() in case that page is used. Or we simply specify > on the single global page in there and force a sync. Changes would be > limited to get_next_page() and write_cache() then.
I might be wrong. I think we might not have to touch write_cache() at all -- it will copy the data into the DataCache buffer. -- Thanks, David / dhildenb
