Re: [PATCH 2/3] fd/locks: allow get the lock owner by F_OFD_GETLK

[Jeff Layton <jlayton@kernel.org>]

On Tue, 2023-06-20 at 15:57 +0500, stsp wrote:
> Hello,
> 
> 20.06.2023 15:51, Jeff Layton пишет:
> > On Tue, 2023-06-20 at 14:55 +0500, Stas Sergeev wrote:
> > > Currently F_OFD_GETLK sets the pid of the lock owner to -1.
> > > Remove such behavior to allow getting the proper owner's pid.
> > > This may be helpful when you want to send some message (like SIGKILL)
> > > to the offending locker.
> > > 
> > > Signed-off-by: Stas Sergeev <stsp2@yandex.ru>
> > > 
> > > CC: Jeff Layton <jlayton@kernel.org>
> > > CC: Chuck Lever <chuck.lever@oracle.com>
> > > CC: Alexander Viro <viro@zeniv.linux.org.uk>
> > > CC: Christian Brauner <brauner@kernel.org>
> > > CC: linux-fsdevel@vger.kernel.org
> > > CC: linux-kernel@vger.kernel.org
> > > 
> > > ---
> > >   fs/locks.c | 2 --
> > >   1 file changed, 2 deletions(-)
> > > 
> > > diff --git a/fs/locks.c b/fs/locks.c
> > > index 210766007e63..ee265e166542 100644
> > > --- a/fs/locks.c
> > > +++ b/fs/locks.c
> > > @@ -2158,8 +2158,6 @@ static pid_t locks_translate_pid(struct file_lock *fl, struct pid_namespace *ns)
> > >   	pid_t vnr;
> > >   	struct pid *pid;
> > >   
> > > -	if (IS_OFDLCK(fl))
> > > -		return -1;
> > >   	if (IS_REMOTELCK(fl))
> > >   		return fl->fl_pid;
> > >   	/*
> > NACK on this one.
> > 
> > OFD locks are not owned by processes. They are owned by the file
> > description (hence the name). Because of this, returning a pid here is
> > wrong.
> 
> But fd is owned by a process.

No, it isn't.

fd's can be handed off between processes via unix descriptors.

Multithreaded processes are also a bit of a gray area here: Suppose I
open a file and set an OFD lock on it in one task, and then let that
task exit while the file is still open. What should l_pid say in that
case?

> PID has a meaning, you can send SIGKILL
> to the returned PID, and the lock is clear.
> Was there any reason to hide the PID at
> a first place?
> 

Yes, because OFD locks are not tied to a pid in the same way that
traditional POSIX locks are.

> 
> > This precedent comes from BSD, where flock() and POSIX locks can
> > conflict. BSD returns -1 for the pid if you call F_GETLK on a file
> > locked with flock(). Since OFD locks have similar ownership semantics to
> > flock() locks, we use the same convention here.
> OK if you insist I can drop this one and
> search the PID by some other means.
> Just a bit unsure what makes it so important
> to overwrite the potentially useful info
> with -1.
> 
> So in case you insist on that, then should
> I send a v2 or can you just drop the patch
> yourself?

-- 
Jeff Layton <jlayton@kernel.org>