Re: [PATCH 1/4] mm/gup: Add FOLL_INTERRUPTIBLE

[Peter Xu]

On Tue, Jun 28, 2022 at 02:40:53PM -0700, John Hubbard wrote:
> On 6/28/22 12:31, Peter Xu wrote:
> > > > diff --git a/mm/gup.c b/mm/gup.c
> > > > index 551264407624..ad74b137d363 100644
> > > > --- a/mm/gup.c
> > > > +++ b/mm/gup.c
> > > > @@ -933,8 +933,17 @@ static int faultin_page(struct vm_area_struct *vma,
> > > >    		fault_flags |= FAULT_FLAG_WRITE;
> > > >    	if (*flags & FOLL_REMOTE)
> > > >    		fault_flags |= FAULT_FLAG_REMOTE;
> > > > -	if (locked)
> > > > +	if (locked) {
> > > >    		fault_flags |= FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
> > > > +		/*
> > > > +		 * We should only grant FAULT_FLAG_INTERRUPTIBLE when we're
> > > > +		 * (at least) killable.  It also mostly means we're not
> > > > +		 * with NOWAIT.  Otherwise ignore FOLL_INTERRUPTIBLE since
> > > > +		 * it won't make a lot of sense to be used alone.
> > > > +		 */
> > > 
> > > This comment seems a little confusing due to its location. We've just
> > > checked "locked", but the comment is talking about other constraints.
> > > 
> > > Not sure what to suggest. Maybe move it somewhere else?
> > 
> > I put it here to be after FAULT_FLAG_KILLABLE we just set.
> > 
> > Only if we have "locked" will we set FAULT_FLAG_KILLABLE.  That's also the
> > key we grant "killable" attribute to this GUP.  So I thought it'll be good
> > to put here because I want to have FOLL_INTERRUPTIBLE dependent on "locked"
> > being set.
> > 
> 
> The key point is the connection between "locked" and killable. If the comment
> explained why "locked" means "killable", that would help clear this up. The
> NOWAIT sentence is also confusing to me, and adding "mostly NOWAIT" does not
> clear it up either... :)

Sorry to have a comment that makes it feels confusing.  I tried to
explicitly put the comment to be after setting FAULT_FLAG_KILLABLE but
obviously I didn't do my job well..

Maybe that NOWAIT thing adds more complexity but not even necessary.

Would below one more acceptable?

		/*
		 * We'll only be able to respond to signals when "locked !=
		 * NULL".  When with it, we'll always respond to SIGKILL
		 * (as implied by FAULT_FLAG_KILLABLE above), and we'll
		 * respond to non-fatal signals only if the GUP user has
		 * specified FOLL_INTERRUPTIBLE.
		 */

Thanks,

-- 
Peter Xu