Re: [PATCH v9 6/6] NFSD: drop a shared boundary page only once both WRITEs have written it

[Mike Snitzer <snitzer@kernel.org>]

On Sun, Oct 25, 2026 at 10:12:10PM +0000, sashiko-bot@kernel.org wrote:
> However, in fs/nfsd/vfs.c:nfsd_write_dio_iters_init(), the boundary
> between the buffered prefix or suffix and the direct middle segment is
> still cut at max_t(u32, offset_align, PAGE_SIZE):
[...]
> Should cut in nfsd_write_dio_iters_init() also take
> mapping_min_folio_nrbytes(nf->nf_file->f_mapping) into account so that
> buffered prefix and suffix segments are cut on unit boundaries?

Yes, and a separate patch already does that.  "NFSD: cut a misaligned
direct-mode WRITE at folio boundaries" replaces the commit in
cel/nfsd-testing that added this cut:

  https://lore.kernel.org/linux-nfs/at0DFod8CcUfus2S@kernel.org/

With it, a misaligned end is cut at

	cut = max_t(unsigned long, offset_align,
		    mapping_min_folio_nrbytes(nf->nf_file->f_mapping));

so the buffered prefix and suffix end on a boundary of the smallest folio,
the same unit that nfsd_write_dio_seg_hold() uses.  The cover letter of
v9 says that v8 1/7 went out separately.  This review ran on the v9 base
(cel/nfsd-testing), which does not have that replacement yet.  v9 applies
on top of the replacement without change.

An end that is already aligned to offset_align keeps no buffered segment,
mid-folio or not.  This keeps such a WRITE direct, on purpose.

Mike