Re: [PATCH v3 00/12] ceph: convert writeback path to folios

Yes, its best to separate any mm/ work from ceph when submitting a
ceph centric patchset.
These fall under different maintainers' purview. The one exception is
when a core API changes, requiring a tree-wide change that also
updates the users. So any mm/ changes shouldnt be a subset incide a
ceph centric patchset.

On Thu, Aug 13, 2026 at 2:51 PM Tal Zussman <[email protected]> wrote:
>
> [adding the mm maintainers back to CC]
>
> On 8/13/26 7:20 AM, Alex Markuze wrote:
> > Hi Tal,
> >
> > Patches 5 and 8 remove wait_on_page_writeback(), thp_size() and
> > thp_order() from mm/, but those helpers are still used by f2fs,
> > mm/filemap.c, and mm/folio-compat.c — removing them would break those
> > users.
> >
> > Could you split this into two series? The ceph-only conversion (patches
> > 1-4, 6-7, 9-12) can go through the ceph tree once we drop those two
> > patches. The mm/ helper removal is a separate effort that needs to
> > coordinate with f2fs and the mm maintainers after all other users are
> > converted.
> >
>
> If this is the preferred approach, I'm happy to split it up this way, but
> to clarify:
>
>  - This series removes the definitions and exports in mm/folio-compat.c
>  - The mm/filemap.c use of thp_order is a local variable, not a call to the
>    function
>  - And I don't see any f2fs users of any of these... it looks like f2fs
>    defines a folio-based wrapper with a similar name to
>    wait_on_page_writeback(), but doesn't actually use it directly.
>
> Given that all users should be removed and the relevant mm maintainers have
> provided Acks for the two mm patches, I think this could all go in together
> safely and avoid delaying this an extra cycle.
>
> Let me know what you prefer.
>
> Thanks!
>
> Tal
>