Re: [PATCH v2 1/2] mm: huge_memory: disallow hugepages if the system-wide THP sysfs settings are disabled

[Lorenzo Stoakes <lorenzo.stoakes@oracle.com>]

On Mon, Jun 09, 2025 at 02:10:12PM +0800, Baolin Wang wrote:
>
>
> On 2025/6/7 19:55, Lorenzo Stoakes wrote:
> > Not related to your patch at all, but man this whole thing (thp allowed orders)
> > needs significant improvement, it seems always perversely complicated for a
> > relatively simple operation.
> >
> > Overall I LOVE what you're doing here, but I feel we can clarify things a
> > little while we're at it to make it clear exactly what we're doing.
> >
> > This is a very important change so forgive my fiddling about here but I'm
> > hoping we can take the opportunity to make things a little simpler!
> >
> > On Thu, Jun 05, 2025 at 04:00:58PM +0800, Baolin Wang wrote:
> > > The MADV_COLLAPSE will ignore the system-wide Anon THP sysfs settings, which
> > > means that even though we have disabled the Anon THP configuration, MADV_COLLAPSE
> > > will still attempt to collapse into a Anon THP. This violates the rule we have
> > > agreed upon: never means never.
> > >
> > > Another rule for madvise, referring to David's suggestion: “allowing for collapsing
> > > in a VM without VM_HUGEPAGE in the "madvise" mode would be fine".
> >
> > I'm generally not sure it's worth talking only about MADV_COLLAPSE here when
> > you're changing what THP is permitted across the board, I may have missed some
> > discussion and forgive me if so, but what is special about MADV_COLLAPSE's use
> > of thp_vma_allowable_orders() that makes it ignore 'never's moreso than other
> > users?
>
> We found that MADV_COLLAPSE ignores the THP configuration, meaning that even
> when THP is set to 'never', MADV_COLLAPSE can still collapse into THPs (and
> mTHPs in the future). This is because when MADV_COLLAPSE calls
> thp_vma_allowable_orders(), it does not set the TVA_ENFORCE_SYSFS flag,
> which means it ignores the system-wide Anon THP sysfs settings.
>
> So this patch set is aimed to fix the THP policy for MADV_COLLAPSE.
>

Yeah of course, and this is exactly why, but what I mean is, the patch
doesn't explicitly address MADV_COLLAPSE, it addresses a case that
MADV_COLLAPSE uses (which is as you say the motivating cause for the
change).

So I think the commit message should rather open something like:

	If, when invoking thp_vma_allowable_orders(), the TVA_ENFORCE_SYSFS
	flag is not specified, we ignore sysfs TLB settings.

	Whilst it makes sense for the callers who do not specify this flag,
	it creates a odd and surprising situation where a sysadmin
	specifying 'never' for all THP sizes still observing THP pages
	being allocated and used on the system.

	The motivating case for this is MADV_COLLAPSE, <blah blah blah> :)

> > > To address this issue, should check whether the Anon THP configuration is disabled
> > > in thp_vma_allowable_orders(), even when the TVA_ENFORCE_SYSFS flag is set.
> > >
> > > In summary, the current strategy is:
> > >
> > > 1. If always & orders == 0, and madvise & orders == 0, and hugepage_global_enabled() == false
> > > (global THP settings are not enabled), it means mTHP of that orders are prohibited
> > > from being used, then madvise_collapse() is forbidden for that orders.
> > >
> > > 2. If always & orders == 0, and madvise & orders == 0, and hugepage_global_enabled() == true
> > > (global THP settings are enabled), and inherit & orders == 0, it means mTHP of that
> > > orders are still prohibited from being used, thus madvise_collapse() is not allowed
> > > for that orders.
> >
> > OK so it's already confusing that the global settings only impact 'inherit'
> > settings below, so they're not really global at all, but rather perhaps should
> > be called 'inherited'.
> >
> > Maybe I need to submit a patch to rename thp_inherited_enabled(), or perhaps
> > that'd just add to the confusion :P
> >
> > OK this is also not your fault just general commentary.
> >
> > Anyway, I feel points 1 and 2 can more succinctly be summed up as below,
> > also there's no need to refer to the code, it's actually clearer I think to
> > refer to the underlying logic:
> >
> > 	If no hugepage modes are enabled for the desired orders, nor can we
> > 	enable them by inheriting from a 'global' enabled setting - then it
> > 	must be the case that all desired orders either specify or inherit
> > 	'NEVER' - and we must abort.
>
> OK. Thanks for helping me make it simpler:)
>

Thanks :)

> > >
> > > Reviewed-by: Zi Yan <ziy@nvidia.com>
> > > Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com>
> > > ---
> > >   include/linux/huge_mm.h | 23 +++++++++++++++++++----
> > >   1 file changed, 19 insertions(+), 4 deletions(-)
> > >
> > > diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
> > > index 2f190c90192d..199ddc9f04a1 100644
> > > --- a/include/linux/huge_mm.h
> > > +++ b/include/linux/huge_mm.h
> > > @@ -287,20 +287,35 @@ unsigned long thp_vma_allowable_orders(struct vm_area_struct *vma,
> > >   				       unsigned long orders)
> > >   {
> > >   	/* Optimization to check if required orders are enabled early. */
> > > -	if ((tva_flags & TVA_ENFORCE_SYSFS) && vma_is_anonymous(vma)) {
> > > -		unsigned long mask = READ_ONCE(huge_anon_orders_always);
> > > +	if (vma_is_anonymous(vma)) {
> > > +		unsigned long always = READ_ONCE(huge_anon_orders_always);
> > > +		unsigned long madvise = READ_ONCE(huge_anon_orders_madvise);
> > > +		unsigned long inherit = READ_ONCE(huge_anon_orders_inherit);
> > > +		unsigned long mask = always | madvise;
> > > +
> > > +		/*
> > > +		 * If the system-wide THP/mTHP sysfs settings are disabled,
> > > +		 * then we should never allow hugepages.
> > > +		 */
> > > +		if (!(mask & orders) && !(hugepage_global_enabled() && (inherit & orders)))
> > > +			return 0;
> > > +
> > > +		if (!(tva_flags & TVA_ENFORCE_SYSFS))
> > > +			goto skip;
> > >
> > > +		mask = always;
> > >   		if (vm_flags & VM_HUGEPAGE)
> > > -			mask |= READ_ONCE(huge_anon_orders_madvise);
> > > +			mask |= madvise;
> > >   		if (hugepage_global_always() ||
> > >   		    ((vm_flags & VM_HUGEPAGE) && hugepage_global_enabled()))
> > > -			mask |= READ_ONCE(huge_anon_orders_inherit);
> > > +			mask |= inherit;
> > >
> > >   		orders &= mask;
> > >   		if (!orders)
> > >   			return 0;
> > >   	}
> > >
> > > +skip:
> > >   	return __thp_vma_allowable_orders(vma, vm_flags, tva_flags, orders);
> > >   }
> >
> > I feel this is compressing a lot of logic in a way that took me several
> > readings to understand (hey I might not be the smartest cookie in the jar,
> > but we need to account for all levels of kernel developer ;)
> >
> > I feel like we can make things a lot clearer here by separating out with a
> > helper function (means we can drop some indentation too), and also take
> > advantage of the fact that, if orders == 0, __thp_vma_allowable_orders()
> > exits with 0 early so no need for us to do so ourselves:
>
> Sure. Looks good to me. Thanks.

Great thanks!

>
> > /* Strictly mask requested anonymous orders according to sysfs settings. */
> > static inline unsigned long __thp_mask_anon_orders(unsigned long vm_flags,
> > 		unsigned long tva_flags, unsigned long orders)
> > {
> > 	unsigned long always = READ_ONCE(huge_anon_orders_always);
> > 	unsigned long madvise = READ_ONCE(huge_anon_orders_madvise);
> > 	unsigned long inherit = READ_ONCE(huge_anon_orders_inherit);;
> > 	bool inherit_enabled = hugepage_global_enabled();
> > 	bool has_madvise =  vm_flags & VM_HUGEPAGE;
> > 	unsigned long mask = always | madvise;
> >
> > 	mask = always | madvise;
> > 	if (inherit_enabled)
> > 		mask |= inherit;
> >
> > 	/* All set to/inherit NEVER - never means never globally, abort. */
> > 	if (!(mask & orders))
> > 		return 0;
> >
> > 	/* Otherwise, we only enforce sysfs settings if asked. */
> > 	if (!(tva_flags & TVA_ENFORCE_SYSFS))
> > 		return orders;
> >
> > 	mask = always;
> > 	if (has_madvise)
> > 		mask |= madvise;
> > 	if (hugepage_global_always() || (has_madvise && inherit_enabled))
> > 		mask |= inherit;
> >
> > 	return orders & mask;
> > }
> >
> > ...
> >
> > static inline
> > unsigned long thp_vma_allowable_orders(struct vm_area_struct *vma,
> > 				       unsigned long vm_flags,
> > 				       unsigned long tva_flags,
> > 				       unsigned long orders)
> > {
> > 	if (vma_is_anonymous(vma))
> > 		orders = __thp_mask_anon_orders(vm_flags, tva_flags, orders);
> >
> > 	return __thp_vma_allowable_orders(vma, vm_flags, tva_flags, orders);
> > }
>
>