| Message ID | 20221114150240.198648-1-jlayton@kernel.org (mailing list archive) |
|---|---|
| Headers | show |
| Series | filelock: remove redundant filp arguments from API | expand |
> On Nov 14, 2022, at 10:02 AM, Jeff Layton <jlayton@kernel.org> wrote: > > Some of the exported functions in fs/locks.c take both a struct file > argument and a struct file_lock. struct file_lock has a dedicated field > to record which file it was set on (fl_file). This is redundant, and > there have been some cases where the two didn't match [1], leading to > bugs. Hi Jeff, doesn't the same argument apply to f_ops->lock ? Do you have a plan for updating that API as well? > This patchset is intended to remove this ambiguity by eliminating the > separate struct file argument from vfs_lock_file, vfs_test_lock and > vfs_cancel_lock. > > Most callers are easy to vet to ensure that they set this correctly, but > lockd had a few places where it wasn't doing the right thing. This > series depends on the lockd patches I sent late last week [2]. > > I'm targeting this series for v6.3. I'll plan to get it into linux-next > soon unless there are objections. > > [1]: https://bugzilla.kernel.org/show_bug.cgi?id=216582 > [2]: https://lore.kernel.org/linux-nfs/20221111215538.356543-1-jlayton@kernel.org/T/#t > > Jeff Layton (3): > filelock: remove redundant filp argument from vfs_lock_file > filelock: remove redundant filp argument from vfs_test_lock > filelock: remove redundant filp arg from vfs_cancel_lock > > fs/ksmbd/smb2pdu.c | 4 ++-- > fs/lockd/svclock.c | 21 +++++++-------------- > fs/lockd/svcsubs.c | 4 ++-- > fs/locks.c | 29 ++++++++++++++--------------- > fs/nfsd/nfs4state.c | 6 +++--- > include/linux/fs.h | 14 +++++++------- > 6 files changed, 35 insertions(+), 43 deletions(-) > > -- > 2.38.1 > -- Chuck Lever
On Mon, 2022-11-14 at 15:08 +0000, Chuck Lever III wrote: > > > On Nov 14, 2022, at 10:02 AM, Jeff Layton <jlayton@kernel.org> wrote: > > > > Some of the exported functions in fs/locks.c take both a struct file > > argument and a struct file_lock. struct file_lock has a dedicated field > > to record which file it was set on (fl_file). This is redundant, and > > there have been some cases where the two didn't match [1], leading to > > bugs. > > Hi Jeff, doesn't the same argument apply to f_ops->lock ? Do you > have a plan for updating that API as well? > > It does apply to fops->lock. I don't have a real plan as of yet. I figure we'll get this set in first and then we can look at changing that API as well. > > This patchset is intended to remove this ambiguity by eliminating the > > separate struct file argument from vfs_lock_file, vfs_test_lock and > > vfs_cancel_lock. > > > > Most callers are easy to vet to ensure that they set this correctly, but > > lockd had a few places where it wasn't doing the right thing. This > > series depends on the lockd patches I sent late last week [2]. > > > > I'm targeting this series for v6.3. I'll plan to get it into linux-next > > soon unless there are objections. > > > > [1]: https://bugzilla.kernel.org/show_bug.cgi?id=216582 > > [2]: https://lore.kernel.org/linux-nfs/20221111215538.356543-1-jlayton@kernel.org/T/#t > > > > Jeff Layton (3): > > filelock: remove redundant filp argument from vfs_lock_file > > filelock: remove redundant filp argument from vfs_test_lock > > filelock: remove redundant filp arg from vfs_cancel_lock > > > > fs/ksmbd/smb2pdu.c | 4 ++-- > > fs/lockd/svclock.c | 21 +++++++-------------- > > fs/lockd/svcsubs.c | 4 ++-- > > fs/locks.c | 29 ++++++++++++++--------------- > > fs/nfsd/nfs4state.c | 6 +++--- > > include/linux/fs.h | 14 +++++++------- > > 6 files changed, 35 insertions(+), 43 deletions(-) > > > > -- > > 2.38.1 > > > > -- > Chuck Lever > > >