Re: [akpm-mm:mm-new] BUILD SUCCESS WITH UNVERIFIED WARNING e0ffd277b09a7c9f1a436e5f75be37ac490bc59f

[Stanislav Kinsburskii <skinsburskii@gmail.com>]

On Tue, Aug 11, 2026 at 05:27:13PM -0700, Andrew Morton wrote:
> On Tue, 11 Aug 2026 16:50:15 -0700 Stanislav Kinsburskii <skinsburskii@gmail.com> wrote:
> 
> > > Caused by
> > > 
> > > 	121170831228 ("mm/hmm: add hmm_range_fault_unlocked_timeout() for mmap lock-drop support")
> > > 
> > > 
> > > Stanislaw, you've previously said that this is a can't-happen, but boy
> > > I'm getting tired of seeing this warning report.  Would the below change
> > > hurt anything?  I can't immediately think of a cost-free way of doing
> > > this.
> > > 
> > 
> > I think this change doesn't do any harm.
> > Perhaps it worth a WARN_ON_ONCE to expose a wrong fault handler
> > behavior?
> 
> Like this?
> 
> 		if (ret & (VM_FAULT_COMPLETED | VM_FAULT_RETRY)) {
> 			if (hmm_vma_walk->locked)    /* needed by sparse */
> 				*hmm_vma_walk->locked = false;
> 			else
> 				WARN_ON_ONCE(1);    /* broken fault handler */
> 			return HMM_FAULT_UNLOCKED;
> 		}
> 

Yes, or even like this:

 		if (ret & (VM_FAULT_COMPLETED | VM_FAULT_RETRY)) {
 			if (WARN_ON_ONCE(hmm_vma_walk->locked))    /* needed by sparse */
 				*hmm_vma_walk->locked = false;
 			return HMM_FAULT_UNLOCKED;
 		}
 

Thanks,
Stanislav