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

[Jeff Layton]

On Tue, 2023-06-20 at 18:39 +0500, stsp wrote:
> 20.06.2023 18:19, Jeff Layton пишет:
> > The bottom line is that these locks are specifically not owned by a
> > process, so returning the l_pid field is unreliable (at best). There is
> > no guarantee that the pid returned will still represent the task that
> > set the lock.
> 
> Though it will, for sure, represent the
> task that _owns_ the lock.
> 
> > You may want to review this article. They're called "File-private" locks
> > here, but the name was later changed to "Open file description" (OFD)
> > locks:
> > 
> >      https://lwn.net/Articles/586904/
> > 
> > The rationale for why -1 is reported is noted there.
> Well, they point to fork() and SCM_RIGHTS.
> Yes, these 2 beasts can make the same lock
> owned by more than one process.
> Yet l_pid returned, is going to be always valid:
> it will still represent one of the valid owners.

No, it won't. The l_pid field is populated from the file_lock->fl_pid.
That field is set when the lock is set, and never updated. So it's quite
possible for F_GETLK to return the pid of a process that no longer
exists.

In principle, we could try to address that by changing how we track lock
ownership, but that's a fairly major overhaul, and I'm not clear on any
use-cases where that matters.

> So my call is to be brave and just re-consider
> the conclusion of that article, made 10 years
> ago! :)
> 

I think my foot has too many bullet wounds for that sort of bravery.

> Of course if returning just 1 of possibly multiple
> owners is a problem, then oh well, I'll drop
> this patch.


-- 
Jeff Layton <jlayton@xxxxxxxxxx>