Re: [PATCH v2 2/2] fuse: remove tmp folio for writebacks and internal rb tree

[Miklos Szeredi <miklos@szeredi.hu>]

On Fri, 25 Oct 2024 at 03:38, Jingbo Xu <jefflexu@linux.alibaba.com> wrote:

> Actually as for FUSE, IIUC the writeback is not guaranteed to be
> completed when sync(2) returns since the temp page mechanism.  When
> sync(2) returns, PG_writeback is indeed cleared for all original pages
> (in the address_space), while the real writeback work (initiated from
> temp page) may be still in progress.

Correct, this is the current behavior of fuse.

I'm not against changing this for the privileged server case.  I.e. a
server running with certain privileges (e.g. CAP_SYS_ADMIN in the
global namespace) then it should be able to opt in to waiting sync(2)
behavior.

> I think this is also what Miklos means in:
> https://lore.kernel.org/all/CAJfpegsJKD4YT5R5qfXXE=hyqKvhpTRbD4m1wsYNbGB6k4rC2A@mail.gmail.com/
>
> Though we need special handling for AS_NO_WRITEBACK_RECLAIM marked pages
> in sync(2) codepath similar to what we have done for the direct reclaim
> in patch 1.

I'd love to get rid of the tmp page thing completely if the same
guarantees can be implemented in the mm.

Thanks,
Miklos