Skip to content

fs: dlm: lock cancellation feature

Alexander Aring cluster-devel 2 versions – New

Get series with b4
b4 am [email protected]

Versions newest first · linked by author and title

  1. v4 Latest3 patches 0 repliesNew
  2. v33 patches 4 repliesSuperseded

Changes in v4 from the cover letter

- drop patch "fs: dlm: ignore DLM_PLOCK_FL_CLOSE flag results", we assume
that plock ops with DLM_PLOCK_FL_CLOSE flag set will never fail.
- Let DLM_PLOCK_OP_CANCEL to always send a reply back. This is useful when
the op fails, e.g. older dlm_controld will return -EINVAL and we can implement F_CANCELLK which does not have a reference of the plock_op instance.
- remove DLM_PLOCK_OP_FLAG_SENT as it was only a optimization for a rare case. That DLM_PLOCK_OP_CANCEL sends a reply back will synchronize it now.
- remove DLM_PLOCK_OP_FLAG_INTERRUPTED as it's not necessary anymore because waiting for a reply of DLM_PLOCK_OP_CANCEL we don't need to handle this special case anymore.
- add "fs: dlm: remove twice newline" because I saw this while doing nfs lockd experiments.

Patches in v4 No applies recorded yet

#Subject and filesSizeRepliesApplied
1/3fs: dlm: remove twice newlinefs/dlm/plock.c+2 -20 replies
2/3fs: dlm: allow to F_SETLKW getting interruptedfs/dlm/plock.c include/uapi/linux/dlm_plock.h+37 -200 replies
3/3fs: dlm: fix F_CANCELLK to cancel pending requestfs/dlm/plock.c fs/gfs2/file.c fs/ocfs2/stack_user.c +1 more+98 -291 reply
3/3fs: dlm: fix F_CANCELLK to cancel pending request+0 -01 reply
3/3fs: dlm: fix F_CANCELLK to cancel pending request+0 -00 replies
3/3fs: dlm: fix F_CANCELLK to cancel pending request+0 -00 replies

Reviews trailers per patch

No Reviewed-by, Acked-by or Tested-by trailers for v4 yet.

Read the v4 thread

Cover letter [Cluster-devel] [PATCHv4 v6.5-rc2 0/3] fs: dlm: lock cancellation feature

Hi,

this patch series implements lock cancellation feature. Either if a POSIX F_SETLKW got interrupted by a signal (most common use case) or lockd sends a F_CANCELLK request.

As I note: the current nfs handling seems to be broken. This patch assumes that the current behaviour works. Future patches will fix the nfs handling which seems to be broken anyway.

- Alex

changes since v4:

- zero memory of info structure in dlm_posix_cancel()

changes since v3:

- drop patch "fs: dlm: ignore DLM_PLOCK_FL_CLOSE flag results", we assume
that plock ops with DLM_PLOCK_FL_CLOSE flag set will never fail.
- Let DLM_PLOCK_OP_CANCEL to always send a reply back. This is useful when
the op fails, e.g. older dlm_controld will return -EINVAL and we can implement F_CANCELLK which does not have a reference of the plock_op instance.
- remove DLM_PLOCK_OP_FLAG_SENT as it was only a optimization for a rare case. That DLM_PLOCK_OP_CANCEL sends a reply back will synchronize it now.
- remove DLM_PLOCK_OP_FLAG_INTERRUPTED as it's not necessary anymore because waiting for a reply of DLM_PLOCK_OP_CANCEL we don't need to handle this special case anymore.
- add "fs: dlm: remove twice newline" because I saw this while doing nfs lockd experiments.

Alexander Aring (3):
fs: dlm: remove twice newline
fs: dlm: allow to F_SETLKW getting interrupted
fs: dlm: fix F_CANCELLK to cancel pending request

 fs/dlm/plock.c                 | 163 ++++++++++++++++++++++++++-------
 fs/gfs2/file.c                 |   9 +-
 fs/ocfs2/stack_user.c          |  13 +--
 include/linux/dlm_plock.h      |   2 +
 include/uapi/linux/dlm_plock.h |   1 +
 5 files changed, 137 insertions(+), 51 deletions(-)
Signature · 1 line
2.31.1