Re: [PATCH 2/2] Replace pid_t in autofs with struct pid reference

[Ian Kent <raven@themaw.net>]

On Thu, 2007-03-22 at 08:31 -0500, Serge E. Hallyn wrote:
> Quoting Ian Kent (raven@themaw.net):
> > On Wed, 2007-03-21 at 21:19 -0500, Serge E. Hallyn wrote:
> > > Quoting Ian Kent (raven@themaw.net):
> > > > On Tue, 2007-03-20 at 16:01 -0600, Eric W. Biederman wrote:
> > > > > "Serge E. Hallyn" <serue@us.ibm.com> writes:
> > > > > 
> > > > > >> >  void autofs4_dentry_release(struct dentry *);
> > > > > >> >  extern void autofs4_kill_sb(struct super_block *);
> > > > > >> > diff --git a/fs/autofs4/waitq.c b/fs/autofs4/waitq.c
> > > > > >> > index 9857543..4a9ad9b 100644
> > > > > >> > --- a/fs/autofs4/waitq.c
> > > > > >> > +++ b/fs/autofs4/waitq.c
> > > > > >> > @@ -141,8 +141,8 @@ static void autofs4_notify_daemon(struct
> > > > > >> >  		packet->ino = wq->ino;
> > > > > >> >  		packet->uid = wq->uid;
> > > > > >> >  		packet->gid = wq->gid;
> > > > > >> > -		packet->pid = wq->pid;
> > > > > >> > -		packet->tgid = wq->tgid;
> > > > > >> > +		packet->pid = pid_nr(wq->pid);
> > > > > >> > +		packet->tgid = pid_nr(wq->tgid);
> > > > > >> >  		break;
> > > > > >> 
> > > > > >> I'm assuming we build the packet in the process context of the
> > > > > >> daemon we are sending it to.  If not we have a problem here.
> > > > > >
> > > > > > Yes this is data being sent to a userspace daemon (Ian pls correct me if
> > > > > > I'm wrong) so the pid_nr is the only thing we can send.
> > > > > 
> > > > > Agreed.  The question is are we in the user space daemon's process when
> > > > > we generate the pid_nr.  Or do we stuff this in some kind of socket,
> > > > > and the socket switch locations of the packet.
> > > > 
> > > > The context here is the automount daemon only for expire runs.
> > > > 
> > > > Mount request packets are triggered by user processes walking over an
> > > > autofs mount point directory. So "current" in this case isn't the autofs
> > > > daemon.
> > > > 
> > > > Requests are sent via a pipe to the daemon.
> > > 
> > > So is the pid used for anything other than debugging?
> > > 
> > > In any case, here is a replacement patch which sends the pid number
> > > in the pid_namespace of the process which did the autofs4 mount.
> > > 
> > > Still not sure whether that is actually what makes sense...
> > > 
> > > From: "Serge E. Hallyn" <serue@us.ibm.com>
> > > Subject: [PATCH] autofs: prevent pid wraparound in waitqs
> > > 
> > > Instead of storing pid numbers for waitqs, store references
> > > to struct pids.  Also store a reference to the mounter's pid
> > > namespace in the autofs4 sb info so that pid numbers for
> > > mount miss and expiry msgs can send the pid# in the mounter's
> > > pidns.
> > 
> > I think this amounts to what I suggested in my previous replies.
> > Hopefully my comments are enough to clear up any questions on
> > correctness of this approach.
> > 
> > Sorry to be a pain but I'm having a little trouble reviewing the patch
> > because I'm not clear on where the code to handle the automount process
> > group (so called oz_pgrp), from the first patch, fits in with this.
> 
> It also has pidspace infrastructure code in it, so I think we will just
> hold off on this until we have that infrastructure merged into the
> pidspace code and into -mm.  Then we can send you a single, more concise
> patch.
> 
> > Is this patch in addition to the original?
> 
> Yes.
> 
> > If so are the references to pid_nr still OK?
> 
> I think so, because AIUI the rest are all executed in a context where
> current is both the actor and recipient.

Not sure about that but lets not worry about it until we have a full
patch to look at.

> Thanks for your help.

My pleasure.

Ian