Re: [PATCH 06/12] blkcg: use q and plid instead of opaque void * for blkio_group association

[Vivek Goyal <vgoyal@redhat.com>]

On Thu, Jan 19, 2012 at 07:55:45AM -0800, Tejun Heo wrote:

[..]
> > > +static void cfq_unlink_blkio_group(struct request_queue *q,
> > > +				   struct blkio_group *blkg)
> > >  {
> > > -	unsigned long  flags;
> > > -	struct cfq_data *cfqd = key;
> > > +	struct cfq_data *cfqd = q->elevator->elevator_data;
> > > +	unsigned long flags;
> > >  
> > > -	spin_lock_irqsave(cfqd->queue->queue_lock, flags);
> > > +	spin_lock_irqsave(q->queue_lock, flags);
> > >  	cfq_destroy_cfqg(cfqd, cfqg_of_blkg(blkg));
> > > -	spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
> > > +	spin_unlock_irqrestore(q->queue_lock, flags);
> > 
> > I think this code will create problem where both old elevator group and
> > new elevator group is on blkcg list and upon cgroup removal one can not
> > rely that q->elevator->elevator_data will give us old elevator's cfqd.
> 
> Again, if I didn't botch up earlier elevator switch code, it shouldn't.

I think I am missing something. IIUC, following is new elevator switch
sequence.

1. elv_quiesce_start
2. unregister old elevator
3. ioc_clear_queue
4. allocate new elevator
5. init new elevator
6. exit old elevator

So any groups on old elevator, will be cleaned up in step 6. So till step
5 these groups are still present on blkcg list. Now assume between step 5
and step 6, if a cgroup removal takes place and blkcg tries to call into
elevator to remove that group, will it not be accessing the wrong cfqd
in cfq_destroy_cfqg() (cfqd of new elevator instead of old elevator).

What am I missing?

Thanks
Vivek