2023-03-15 17:17:39

by Peter Xu

[permalink] [raw]

Subject: [PATCH] mm/thp: Rename TRANSPARENT_HUGEPAGE_NEVER_DAX to _UNSUPPORTED

---

TRANSPARENT_HUGEPAGE_NEVER_DAX has nothing to do with DAX. It's set when
has_transparent_hugepage() returns false, checked in hugepage_vma_check()
and will disable THP completely if false. Rename it to reflect its real
purpose.

Cc: Aneesh Kumar K.V <[email protected]>
Cc: Yang Shi <[email protected]>
Signed-off-by: Peter Xu <[email protected]>
---
include/linux/huge_mm.h | 2 +-
mm/huge_memory.c | 4 ++--
2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
index 70bd867eba94..9a3a3af2dd80 100644
--- a/include/linux/huge_mm.h
+++ b/include/linux/huge_mm.h
@@ -79,7 +79,7 @@ static inline vm_fault_t vmf_insert_pfn_pud(struct vm_fault *vmf, pfn_t pfn,
}

enum transparent_hugepage_flag {
- TRANSPARENT_HUGEPAGE_NEVER_DAX,
+ TRANSPARENT_HUGEPAGE_UNSUPPORTED,
TRANSPARENT_HUGEPAGE_FLAG,
TRANSPARENT_HUGEPAGE_REQ_MADV_FLAG,
TRANSPARENT_HUGEPAGE_DEFRAG_DIRECT_FLAG,
diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index b0ab247939e0..913e7dc32869 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -88,7 +88,7 @@ bool hugepage_vma_check(struct vm_area_struct *vma, unsigned long vm_flags,
/*
* If the hardware/firmware marked hugepage support disabled.
*/
- if (transparent_hugepage_flags & (1 << TRANSPARENT_HUGEPAGE_NEVER_DAX))
+ if (transparent_hugepage_flags & (1 << TRANSPARENT_HUGEPAGE_UNSUPPORTED))
return false;

/* khugepaged doesn't collapse DAX vma, but page fault is fine. */
@@ -464,7 +464,7 @@ static int __init hugepage_init(void)
* Hardware doesn't support hugepages, hence disable
* DAX PMD support.
*/
- transparent_hugepage_flags = 1 << TRANSPARENT_HUGEPAGE_NEVER_DAX;
+ transparent_hugepage_flags = 1 << TRANSPARENT_HUGEPAGE_UNSUPPORTED;
return -EINVAL;
}

--
2.39.1

2023-03-15 21:07:11

by Yang Shi

[permalink] [raw]

Subject: Re: [PATCH] mm/thp: Rename TRANSPARENT_HUGEPAGE_NEVER_DAX to _UNSUPPORTED

---

On Wed, Mar 15, 2023 at 10:16 AM Peter Xu <[email protected]> wrote:
>
> TRANSPARENT_HUGEPAGE_NEVER_DAX has nothing to do with DAX. It's set when
> has_transparent_hugepage() returns false, checked in hugepage_vma_check()
> and will disable THP completely if false. Rename it to reflect its real
> purpose.

Makes sense to me. Reviewed-by: Yang Shi <[email protected]>

>
> Cc: Aneesh Kumar K.V <[email protected]>
> Cc: Yang Shi <[email protected]>
> Signed-off-by: Peter Xu <[email protected]>
> ---
> include/linux/huge_mm.h | 2 +-
> mm/huge_memory.c | 4 ++--
> 2 files changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
> index 70bd867eba94..9a3a3af2dd80 100644
> --- a/include/linux/huge_mm.h
> +++ b/include/linux/huge_mm.h
> @@ -79,7 +79,7 @@ static inline vm_fault_t vmf_insert_pfn_pud(struct vm_fault *vmf, pfn_t pfn,
> }
>
> enum transparent_hugepage_flag {
> - TRANSPARENT_HUGEPAGE_NEVER_DAX,
> + TRANSPARENT_HUGEPAGE_UNSUPPORTED,
> TRANSPARENT_HUGEPAGE_FLAG,
> TRANSPARENT_HUGEPAGE_REQ_MADV_FLAG,
> TRANSPARENT_HUGEPAGE_DEFRAG_DIRECT_FLAG,
> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
> index b0ab247939e0..913e7dc32869 100644
> --- a/mm/huge_memory.c
> +++ b/mm/huge_memory.c
> @@ -88,7 +88,7 @@ bool hugepage_vma_check(struct vm_area_struct *vma, unsigned long vm_flags,
> /*
> * If the hardware/firmware marked hugepage support disabled.
> */
> - if (transparent_hugepage_flags & (1 << TRANSPARENT_HUGEPAGE_NEVER_DAX))
> + if (transparent_hugepage_flags & (1 << TRANSPARENT_HUGEPAGE_UNSUPPORTED))
> return false;
>
> /* khugepaged doesn't collapse DAX vma, but page fault is fine. */
> @@ -464,7 +464,7 @@ static int __init hugepage_init(void)
> * Hardware doesn't support hugepages, hence disable
> * DAX PMD support.
> */
> - transparent_hugepage_flags = 1 << TRANSPARENT_HUGEPAGE_NEVER_DAX;
> + transparent_hugepage_flags = 1 << TRANSPARENT_HUGEPAGE_UNSUPPORTED;
> return -EINVAL;
> }
>
> --
> 2.39.1
>

2023-03-16 14:58:09

by David Hildenbrand

[permalink] [raw]

Subject: Re: [PATCH] mm/thp: Rename TRANSPARENT_HUGEPAGE_NEVER_DAX to _UNSUPPORTED

---

On 15.03.23 18:16, Peter Xu wrote:
> TRANSPARENT_HUGEPAGE_NEVER_DAX has nothing to do with DAX. It's set when
> has_transparent_hugepage() returns false, checked in hugepage_vma_check()
> and will disable THP completely if false. Rename it to reflect its real
> purpose.
>
> Cc: Aneesh Kumar K.V <[email protected]>
> Cc: Yang Shi <[email protected]>
> Signed-off-by: Peter Xu <[email protected]>
> ---
> include/linux/huge_mm.h | 2 +-
> mm/huge_memory.c | 4 ++--
> 2 files changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
> index 70bd867eba94..9a3a3af2dd80 100644
> --- a/include/linux/huge_mm.h
> +++ b/include/linux/huge_mm.h
> @@ -79,7 +79,7 @@ static inline vm_fault_t vmf_insert_pfn_pud(struct vm_fault *vmf, pfn_t pfn,
> }
>
> enum transparent_hugepage_flag {
> - TRANSPARENT_HUGEPAGE_NEVER_DAX,
> + TRANSPARENT_HUGEPAGE_UNSUPPORTED,
> TRANSPARENT_HUGEPAGE_FLAG,
> TRANSPARENT_HUGEPAGE_REQ_MADV_FLAG,
> TRANSPARENT_HUGEPAGE_DEFRAG_DIRECT_FLAG,
> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
> index b0ab247939e0..913e7dc32869 100644
> --- a/mm/huge_memory.c
> +++ b/mm/huge_memory.c
> @@ -88,7 +88,7 @@ bool hugepage_vma_check(struct vm_area_struct *vma, unsigned long vm_flags,
> /*
> * If the hardware/firmware marked hugepage support disabled.
> */
> - if (transparent_hugepage_flags & (1 << TRANSPARENT_HUGEPAGE_NEVER_DAX))
> + if (transparent_hugepage_flags & (1 << TRANSPARENT_HUGEPAGE_UNSUPPORTED))
> return false;
>
> /* khugepaged doesn't collapse DAX vma, but page fault is fine. */
> @@ -464,7 +464,7 @@ static int __init hugepage_init(void)
> * Hardware doesn't support hugepages, hence disable
> * DAX PMD support.


We should also fixup that comment then, no?

Acked-by: David Hildenbrand <[email protected]>

--
Thanks,

David / dhildenb

2023-03-16 15:19:24

by Peter Xu

[permalink] [raw]

Subject: Re: [PATCH] mm/thp: Rename TRANSPARENT_HUGEPAGE_NEVER_DAX to _UNSUPPORTED

---

On Thu, Mar 16, 2023 at 03:57:12PM +0100, David Hildenbrand wrote:
> On 15.03.23 18:16, Peter Xu wrote:
> > TRANSPARENT_HUGEPAGE_NEVER_DAX has nothing to do with DAX. It's set when
> > has_transparent_hugepage() returns false, checked in hugepage_vma_check()
> > and will disable THP completely if false. Rename it to reflect its real
> > purpose.
> >
> > Cc: Aneesh Kumar K.V <[email protected]>
> > Cc: Yang Shi <[email protected]>
> > Signed-off-by: Peter Xu <[email protected]>
> > ---
> > include/linux/huge_mm.h | 2 +-
> > mm/huge_memory.c | 4 ++--
> > 2 files changed, 3 insertions(+), 3 deletions(-)
> >
> > diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
> > index 70bd867eba94..9a3a3af2dd80 100644
> > --- a/include/linux/huge_mm.h
> > +++ b/include/linux/huge_mm.h
> > @@ -79,7 +79,7 @@ static inline vm_fault_t vmf_insert_pfn_pud(struct vm_fault *vmf, pfn_t pfn,
> > }
> > enum transparent_hugepage_flag {
> > - TRANSPARENT_HUGEPAGE_NEVER_DAX,
> > + TRANSPARENT_HUGEPAGE_UNSUPPORTED,
> > TRANSPARENT_HUGEPAGE_FLAG,
> > TRANSPARENT_HUGEPAGE_REQ_MADV_FLAG,
> > TRANSPARENT_HUGEPAGE_DEFRAG_DIRECT_FLAG,
> > diff --git a/mm/huge_memory.c b/mm/huge_memory.c
> > index b0ab247939e0..913e7dc32869 100644
> > --- a/mm/huge_memory.c
> > +++ b/mm/huge_memory.c
> > @@ -88,7 +88,7 @@ bool hugepage_vma_check(struct vm_area_struct *vma, unsigned long vm_flags,
> > /*
> > * If the hardware/firmware marked hugepage support disabled.
> > */
> > - if (transparent_hugepage_flags & (1 << TRANSPARENT_HUGEPAGE_NEVER_DAX))
> > + if (transparent_hugepage_flags & (1 << TRANSPARENT_HUGEPAGE_UNSUPPORTED))
> > return false;
> > /* khugepaged doesn't collapse DAX vma, but page fault is fine. */
> > @@ -464,7 +464,7 @@ static int __init hugepage_init(void)
> > * Hardware doesn't support hugepages, hence disable
> > * DAX PMD support.
>
>
> We should also fixup that comment then, no?

Yes, I'd drop the comment since the code explains itself. One fixup
attached. Thanks.

--
Peter Xu

---

Attachments:

(No filename) (2.02 kB)
0001-fixup-mm-thp-rename-TRANSPARENT_HUGEPAGE_NEVER_DAX-t.patch (804.00 B)
Download all attachments

2023-03-16 15:40:53

by David Hildenbrand

[permalink] [raw]

Subject: Re: [PATCH] mm/thp: Rename TRANSPARENT_HUGEPAGE_NEVER_DAX to _UNSUPPORTED

---

On 16.03.23 16:18, Peter Xu wrote:
> On Thu, Mar 16, 2023 at 03:57:12PM +0100, David Hildenbrand wrote:
>> On 15.03.23 18:16, Peter Xu wrote:
>>> TRANSPARENT_HUGEPAGE_NEVER_DAX has nothing to do with DAX. It's set when
>>> has_transparent_hugepage() returns false, checked in hugepage_vma_check()
>>> and will disable THP completely if false. Rename it to reflect its real
>>> purpose.
>>>
>>> Cc: Aneesh Kumar K.V <[email protected]>
>>> Cc: Yang Shi <[email protected]>
>>> Signed-off-by: Peter Xu <[email protected]>
>>> ---
>>> include/linux/huge_mm.h | 2 +-
>>> mm/huge_memory.c | 4 ++--
>>> 2 files changed, 3 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
>>> index 70bd867eba94..9a3a3af2dd80 100644
>>> --- a/include/linux/huge_mm.h
>>> +++ b/include/linux/huge_mm.h
>>> @@ -79,7 +79,7 @@ static inline vm_fault_t vmf_insert_pfn_pud(struct vm_fault *vmf, pfn_t pfn,
>>> }
>>> enum transparent_hugepage_flag {
>>> - TRANSPARENT_HUGEPAGE_NEVER_DAX,
>>> + TRANSPARENT_HUGEPAGE_UNSUPPORTED,
>>> TRANSPARENT_HUGEPAGE_FLAG,
>>> TRANSPARENT_HUGEPAGE_REQ_MADV_FLAG,
>>> TRANSPARENT_HUGEPAGE_DEFRAG_DIRECT_FLAG,
>>> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
>>> index b0ab247939e0..913e7dc32869 100644
>>> --- a/mm/huge_memory.c
>>> +++ b/mm/huge_memory.c
>>> @@ -88,7 +88,7 @@ bool hugepage_vma_check(struct vm_area_struct *vma, unsigned long vm_flags,
>>> /*
>>> * If the hardware/firmware marked hugepage support disabled.
>>> */
>>> - if (transparent_hugepage_flags & (1 << TRANSPARENT_HUGEPAGE_NEVER_DAX))
>>> + if (transparent_hugepage_flags & (1 << TRANSPARENT_HUGEPAGE_UNSUPPORTED))
>>> return false;
>>> /* khugepaged doesn't collapse DAX vma, but page fault is fine. */
>>> @@ -464,7 +464,7 @@ static int __init hugepage_init(void)
>>> * Hardware doesn't support hugepages, hence disable
>>> * DAX PMD support.
>>
>>
>> We should also fixup that comment then, no?
>
> Yes, I'd drop the comment since the code explains itself. One fixup
> attached. Thanks.

LGTM

--
Thanks,

David / dhildenb