Re: [stable-6.6.y] mm: khugepaged refuses to freeze

[Baolin Wang]

On 2/10/26 11:21 AM, Sergey Senozhatsky wrote:
> On (26/02/06 10:00), David Hildenbrand (Arm) wrote:
> > > > I recall that there is a notifier when the system is preparing to
> > > > sleep (pm notifier or something). Could we simply hook into that to
> > > > tell khugepaged to suspend+resume?
> > >
> > > Do you mean “struct dev_pm_ops”, which is used to register PM callbacks
> > > for devices? However, I don’t know how to use it with a kernel thread.
> > >
> > > Also look at how kswapd does it, kswapd also uses
> > > kthread_freezable_should_stop() to check the freeze state.
> >
> > Right, mimicking what kswapd does sound reasonable!
>
> I may be missing something, as I'm not seeing dev_pm_ops in vmscan code.
> Would something like this work?
>
> ---
>
> diff --git a/mm/khugepaged.c b/mm/khugepaged.c
> index fa6a018b20a8..c5d89ec223d3 100644
> --- a/mm/khugepaged.c
> +++ b/mm/khugepaged.c
> @@ -394,8 +394,12 @@ static inline int hpage_collapse_test_exit(struct mm_struct *mm)
>   
>   static inline int hpage_collapse_test_exit_or_disable(struct mm_struct *mm)
>   {
> +	bool was_frozen;
> +	int ret = kthread_freezable_should_stop(&was_frozen);
> +
>   	return hpage_collapse_test_exit(mm) ||
> -		mm_flags_test(MMF_DISABLE_THP_COMPLETELY, mm);
> +		mm_flags_test(MMF_DISABLE_THP_COMPLETELY, mm) ||
> +		was_frozen || ret;
>   }

Since the hpage_collapse_test_exit_or_disable() can be called by 
madvise_callapse(), which is not a kernel thread. So I think using the 
try_to_freeze() is enough? or pass the cc->is_khugepaged to check if 
current thread is khugepaged.