Re: [PATCH] mm: madvise: make MADV_NOHUGEPAGE a no-op if !THP

[Ignacio Moreno Gonzalez <Ignacio.MorenoGonzalez@kuka.com>]

On 5/15/2025 3:03 PM, Lorenzo Stoakes wrote:
> Hi Igancio,
> 
> I can package this up in a series with the fix below and with the
> appropriate tags (signed-off by you) to enforce the correct ordering
> between the s390 patch and this one if that works for you?
> 
> Let me know if my sending this revised patch in series with these changes
> and your signed-off works for you?

Hi Lorenzo,

Yes, that's fine :) Thanks!

> Cheers, Lorenzo
> 
> On Thu, May 15, 2025 at 09:03:19AM +0200, Ignacio Moreno Gonzalez wrote:
>> On 5/14/2025 10:15 PM, James Houghton wrote:
>>> On Thu, May 8, 2025 at 3:20 AM Ignacio Moreno Gonzalez via B4 Relay
>>> <devnull+Ignacio.MorenoGonzalez.kuka.com@kernel.org> wrote:
>>>>
>>>> From: Ignacio Moreno Gonzalez <Ignacio.MorenoGonzalez@kuka.com>
>>>>
>>>> VM_NOHUGEPAGE is a no-op if CONFIG_TRANSPARENT_HUGEPAGE is disabled. So
>>>> it makes no sense to return an error when calling madvise() with
>>>> MADV_NOHUGEPAGE in that case.
>>>>
>>>> Suggested-by: Matthew Wilcox <willy@infradead.org>
>>>> Signed-off-by: Ignacio Moreno Gonzalez <Ignacio.MorenoGonzalez@kuka.com>
>>>> ---
>>>> https://lore.kernel.org/linux-mm/20250502-map-map_stack-to-vm_nohugepage-only-if-thp-is-enabled-v1-1-113cc634cd51@kuka.com
>>>>
>>>> Here it is presented as a separate thread to avoid mixing stable and
>>>> non-stable patches.
>>>>
>>>> This change makes calling madvise(addr, size, MADV_NOHUGEPAGE) on !THP
>>>> kernels to return 0 instead of -EINVAL.
>>>> ---
>>>>  include/linux/huge_mm.h | 5 +++++
>>>>  1 file changed, 5 insertions(+)
>>>>
>>>> diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
>>>> index e893d546a49f464f7586db639fe216231f03651a..5fca742dc5ba784ffccea055b07247707d16cc67 100644
>>>> --- a/include/linux/huge_mm.h
>>>> +++ b/include/linux/huge_mm.h
>>>> @@ -509,6 +509,8 @@ bool unmap_huge_pmd_locked(struct vm_area_struct *vma, unsigned long addr,
>>>>
>>>>  #else /* CONFIG_TRANSPARENT_HUGEPAGE */
>>>>
>>>> +#include <uapi/asm/mman.h>
>>>> +
>>>>  static inline bool folio_test_pmd_mappable(struct folio *folio)
>>>>  {
>>>>         return false;
>>>> @@ -598,6 +600,9 @@ static inline bool unmap_huge_pmd_locked(struct vm_area_struct *vma,
>>>>  static inline int hugepage_madvise(struct vm_area_struct *vma,
>>>>                                    unsigned long *vm_flags, int advice)
>>>>  {
>>>> +       /* On a !THP kernel, MADV_NOHUGEPAGE is a no-op, but MADV_NOHUGEPAGE is not supported */
>>>
>>> Do you mean "but MADV_HUGEPAGE is not supported"? Just want to make
>>> sure; it seems like no one else has asked about this yet.
>>>
>>
>> Yes, this is a typo. It should be MADV_HUGEPAGE. Thanks!
>>
>>>> +       if (advice == MADV_NOHUGEPAGE)
>>>> +               return 0;
>>>>         return -EINVAL;
>>>>  }
>>>
>>
>