Re: [PATCH for-rc v1] IB/hfi1: Move cached value of mm into handler

[Jason Gunthorpe <jgg@nvidia.com>]

On Wed, Nov 04, 2020 at 04:12:45PM -0800, Ira Weiny wrote:
> > -int hfi1_mmu_rb_register(void *ops_arg, struct mm_struct *mm,
> > +int hfi1_mmu_rb_register(void *ops_arg,
> >  			 struct mmu_rb_ops *ops,
> >  			 struct workqueue_struct *wq,
> >  			 struct mmu_rb_handler **handler)
> > @@ -110,18 +98,20 @@ int hfi1_mmu_rb_register(void *ops_arg, struct mm_struct *mm,
> >  	INIT_HLIST_NODE(&handlr->mn.hlist);
> >  	spin_lock_init(&handlr->lock);
> >  	handlr->mn.ops = &mn_opts;
> > -	handlr->mm = mm;
> >  	INIT_WORK(&handlr->del_work, handle_remove);
> >  	INIT_LIST_HEAD(&handlr->del_list);
> >  	INIT_LIST_HEAD(&handlr->lru_list);
> >  	handlr->wq = wq;
> >  
> > -	ret = mmu_notifier_register(&handlr->mn, handlr->mm);
> > +	ret = mmu_notifier_register(&handlr->mn, current->mm);
> >  	if (ret) {
> >  		kfree(handlr);
> >  		return ret;
> >  	}
> >  
> > +	mmget(current->mm);
> 
> I flagged this initially but then reviewed the commit message for why you need
> this reference.  I think it is worth a comment here as well as below.
> Specifically mentioning the order of calls in do_exit().

Oh? a mmget should not be held for a long time, and a notifier already
holds a mmgrab while it is registered

If hfi later needs the mmget then the only the part that needs it
should convert the mmgrab to a mmget with mmget_not_zero

Jason