[PATCH v6 0/9] vfs/nfsd: clean up handling of i_version counter

Jeff Layton posted 9 patches 2 months ago
fs/ceph/inode.c           | 16 +++++++++----
fs/ext4/file.c            | 20 +++++++++++++---
fs/libfs.c                | 36 +++++++++++++++++++++++++++++
fs/nfs/export.c           |  7 ------
fs/nfs/inode.c            | 10 ++++++--
fs/nfsd/nfs4xdr.c         |  4 +++-
fs/nfsd/nfsfh.c           | 40 ++++++++++++++++++++++++++++++++
fs/nfsd/nfsfh.h           | 29 +----------------------
fs/nfsd/vfs.h             |  7 +++++-
fs/stat.c                 |  7 ++++++
include/linux/exportfs.h  |  1 -
include/linux/iversion.h  | 48 ++++++++-------------------------------
include/linux/stat.h      |  2 +-
include/uapi/linux/stat.h |  6 +++--
mm/filemap.c              | 17 ++++++++++----
samples/vfs/test-statx.c  |  8 +++++--
16 files changed, 163 insertions(+), 95 deletions(-)
[PATCH v6 0/9] vfs/nfsd: clean up handling of i_version counter
Posted by Jeff Layton 2 months ago
v6: add support for STATX_ATTR_VERSION_MONOTONIC
    add patch to expose i_version counter to userland
    patches to update times/version after copying write data

An updated set of i_version handling changes! I've dropped the earlier
ext4 patches since Ted has picked up the relevant ext4 ones.

This set is based on linux-next, to make sure we don't collide with the
statx DIO alignment patches, and some other i_version cleanups that are
in flight.  I'm hoping those land in 6.1.

There are a few changes since v5, mostly centered around adding
STATX_ATTR_VERSION_MONOTONIC. I've also re-added the patch to expose
STATX_VERSION to userland via statx. What I'm proposing should now
(mostly) conform to the semantics I layed out in the manpage patch I
sent recently [1].

Finally, I've added two patches to make __generic_file_write_iter and
ext4 update the c/mtime after copying file data instead of before, which
Neil pointed out makes for better cache-coherency handling. Those should
take care of ext4 and tmpfs. xfs and btrfs will need to make the same
changes.

One thing I'm not sure of is what we should do if update_times fails
after an otherwise successful write. Should we just ignore that and move
on (and maybe WARN)? Return an error? Set a writeback error? What's the
right recourse there?

I'd like to go ahead and get the first 6 patches from this series into
linux-next fairly soon, so if anyone has objections, please speak up!

[1]: https://lore.kernel.org/linux-nfs/20220928134200.28741-1-jlayton@kernel.org/T/#u

Jeff Layton (9):
  iversion: move inode_query_iversion to libfs.c
  iversion: clarify when the i_version counter must be updated
  vfs: plumb i_version handling into struct kstat
  nfs: report the inode version in getattr if requested
  ceph: report the inode version in getattr if requested
  nfsd: use the getattr operation to fetch i_version
  vfs: expose STATX_VERSION to userland
  vfs: update times after copying data in __generic_file_write_iter
  ext4: update times after I/O in write codepaths

 fs/ceph/inode.c           | 16 +++++++++----
 fs/ext4/file.c            | 20 +++++++++++++---
 fs/libfs.c                | 36 +++++++++++++++++++++++++++++
 fs/nfs/export.c           |  7 ------
 fs/nfs/inode.c            | 10 ++++++--
 fs/nfsd/nfs4xdr.c         |  4 +++-
 fs/nfsd/nfsfh.c           | 40 ++++++++++++++++++++++++++++++++
 fs/nfsd/nfsfh.h           | 29 +----------------------
 fs/nfsd/vfs.h             |  7 +++++-
 fs/stat.c                 |  7 ++++++
 include/linux/exportfs.h  |  1 -
 include/linux/iversion.h  | 48 ++++++++-------------------------------
 include/linux/stat.h      |  2 +-
 include/uapi/linux/stat.h |  6 +++--
 mm/filemap.c              | 17 ++++++++++----
 samples/vfs/test-statx.c  |  8 +++++--
 16 files changed, 163 insertions(+), 95 deletions(-)

-- 
2.37.3