[RFCv2,0/4] iomap: zero range folio batch processing prototype

Message

Brian Foster Dec. 13, 2024, 3:05 p.m. UTC

Hi all,

This is an update of the folio_batch prototype for zero range. The idea
here is first to provide a mechanism to correctly handle dirty folios
over unwritten extents on zero range without needing to flush, and
second to add it in a way such that it can be extended to other
operations in the future. For example, seek data over unwritten extents
could use similar treatment as zero range, and multiple people have
brought up the idea of using this for buffered writes in the future.

The primary change in RFCv2 is this is now ported on top of the
incremental iter advance patches, which eliminates the need for patch 1
in RFCv1 and allows plumbing into the normal folio get path.

This is still a WIP and I'm posting mainly as a reference for the
incremental advance patches. Patch 1 is a squash of 7 or 8 prep patches
that aren't all that interesting, patch 2 implements batch support,
patch 3 is a prep patch for XFS, and patch 4 updates XFS to use the iter
batch on zero range.

Thoughts, reviews, flames appreciated.

Brian

RFCv2:
- Port onto incremental advance, drop patch 1 from RFCv1.
- Moved batch into iomap_iter, dynamically allocate and drop flag.
- Tweak XFS patch to always trim zero range on EOF boundary.
RFCv1: https://lore.kernel.org/linux-fsdevel/[e-mail]/

Brian Foster (4):
  iomap: prep work for folio_batch support
  iomap: optional zero range dirty folio processing
  xfs: always trim mapping to requested range for zero range
  xfs: fill dirty folios on zero range of unwritten mappings

 fs/iomap/buffered-io.c | 177 ++++++++++++++++++++++++++++++-----------
 fs/iomap/iter.c        |   6 ++
 fs/xfs/xfs_iomap.c     |  25 ++++--
 include/linux/iomap.h  |   4 +
 4 files changed, 158 insertions(+), 54 deletions(-)