[PATCH v2 0/3] NFS: fixes for delegated timestamps

[Jeff Layton <jlayton@kernel.org>]

This version fixes some issues that Sashiko pointed out.

These are some client-side fixes for problems that I was seeing with
delegated timestamps. Mostly this was evident in the git regression
testsuite (which is particularly sensitive to timestamp issues).

Signed-off-by: Jeff Layton <jlayton@kernel.org>
---
Changes in v2:
- "don't let delegated c/mtime stamps go backward": compute the stamp
  before storing it, so lockless readers never see the coarse time
- "stamp after the server's ctime when truncating under delegated times":
  stamp before nfs_vmtruncate() drops i_lock
- "stamp after the server's ctime following a CLONE":
  - stamp under i_lock
  - apply the reply with nfs_refresh_inode() under delegated times, so
    its ctime doesn't overwrite the stamp
- Link to v1: https://lore.kernel.org/r/20261106-delegts-nfs-v1-0-1d5c5e364a07@kernel.org

---
Jeff Layton (3):
      NFS: don't let delegated c/mtime stamps go backward
      NFS: stamp after the server's ctime when truncating under delegated times
      NFS: stamp after the server's ctime following a CLONE

 fs/nfs/delegation.h |  2 ++
 fs/nfs/inode.c      | 65 ++++++++++++++++++++++++++++++++++++++++-------------
 fs/nfs/nfs42proc.c  | 25 ++++++++++++++++-----
 3 files changed, 72 insertions(+), 20 deletions(-)
---
base-commit: 5e810a57a1cf2fc785d89e711ff40f1112f6404c
change-id: 20261106-delegts-nfs-25dafe47b434

Best regards,
-- 
Jeff Layton <jlayton@kernel.org>