Re: [PATCH] support "THPeligible" semantics for mTHP with anonymous shmem

["Bang Li" <libang.li@antgroup.com>]

On 2024/7/1 19:12, Baolin Wang wrote:
> 
> 
> On 2024/7/1 17:43, Bang Li wrote:
>> Hi, Baolin
>>
>> On 2024/7/1 16:33, Baolin Wang wrote:
>>>
>>>
>>> On 2024/7/1 15:55, Ryan Roberts wrote:
>>>> On 28/06/2024 11:49, Bang Li wrote:
>>>>> After the commit 7fb1b252afb5 ("mm: shmem: add mTHP support for
>>>>> anonymous shmem"), we can configure different policies through
>>>>> the multi-size THP sysfs interface for anonymous shmem. But
>>>>> currently "THPeligible" indicates only whether the mapping is
>>>>> eligible for allocating THP-pages as well as the THP is PMD
>>>>> mappable or not for anonymous shmem, we need to support semantics
>>>>> for mTHP with anonymous shmem similar to those for mTHP with
>>>>> anonymous memory.
>>>>>
>>>>> Signed-off-by: Bang Li <libang.li@antgroup.com>
>>>>> ---
>>>>>   fs/proc/task_mmu.c      | 10 +++++++---
>>>>>   include/linux/huge_mm.h | 11 +++++++++++
>>>>>   mm/shmem.c              |  9 +--------
>>>>>   3 files changed, 19 insertions(+), 11 deletions(-)
>>>>>
>>>>> diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
>>>>> index 93fb2c61b154..09b5db356886 100644
>>>>> --- a/fs/proc/task_mmu.c
>>>>> +++ b/fs/proc/task_mmu.c
>>>>> @@ -870,6 +870,7 @@ static int show_smap(struct seq_file *m, void *v)
>>>>>   {
>>>>>       struct vm_area_struct *vma = v;
>>>>>       struct mem_size_stats mss = {};
>>>>> +    bool thp_eligible;
>>>>>       smap_gather_stats(vma, &mss, 0);
>>>>> @@ -882,9 +883,12 @@ static int show_smap(struct seq_file *m, void *v)
>>>>>       __show_smap(m, &mss, false);
>>>>> -    seq_printf(m, "THPeligible:    %8u\n",
>>>>> -           !!thp_vma_allowable_orders(vma, vma->vm_flags,
>>>>> -               TVA_SMAPS | TVA_ENFORCE_SYSFS, THP_ORDERS_ALL));
>>>>> +    thp_eligible = !!thp_vma_allowable_orders(vma, vma->vm_flags,
>>>>> +                        TVA_SMAPS | TVA_ENFORCE_SYSFS, 
>>>>> THP_ORDERS_ALL);
>>>>> +    if (vma_is_anon_shmem(vma))
>>>>> +        thp_eligible = 
>>>>> !!shmem_allowable_huge_orders(file_inode(vma->vm_file),
>>>>> +                            vma, vma->vm_pgoff, thp_eligible);
>>>>
>>>> Afraid I haven't been following the shmem mTHP support work as much 
>>>> as I would
>>>> have liked, but is there a reason why we need a separate function 
>>>> for shmem?
>>>
>>> Since shmem_allowable_huge_orders() only uses shmem specific logic to 
>>> determine if huge orders are allowable, there is no need to 
>>> complicate the thp_vma_allowable_orders() function by adding more 
>>> shmem related logic, making it more bloated. In my view, providing a 
>>> dedicated helper shmem_allowable_huge_orders(), specifically for 
>>> shmem, simplifies the logic.
>>>
>>> IIUC, I agree with David's suggestion that the 
>>> shmem_allowable_huge_orders() helper function could be used in 
>>> thp_vma_allowable_orders() to support shmem mTHP. Something like:
>>>
>>> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
>>> index c7ce28f6b7f3..9677fe6cf478 100644
>>> --- a/mm/huge_memory.c
>>> +++ b/mm/huge_memory.c
>>> @@ -151,10 +151,13 @@ unsigned long __thp_vma_allowable_orders(struct 
>>> vm_area_struct *vma,
>>>           * Must be done before hugepage flags check since shmem has its
>>>           * own flags.
>>>           */
>>> -       if (!in_pf && shmem_file(vma->vm_file))
>>> -               return shmem_is_huge(file_inode(vma->vm_file), 
>>> vma->vm_pgoff,
>>> -                                    !enforce_sysfs, vma->vm_mm, 
>>> vm_flags)
>>> -                       ? orders : 0;
>>> +       if (!in_pf && shmem_file(vma->vm_file)) {
>>> +               bool global_huge = 
>>> shmem_is_huge(file_inode(vma->vm_file), vma->vm_pgoff,
>>> +                                    !enforce_sysfs, vma->vm_mm, 
>>> vm_flags);
>>> +
>>> +               return 
>>> shmem_allowable_huge_orders(file_inode(vma->vm_file),
>>> +                                       vma, vma->vm_pgoff, 
>>> global_huge);
>>> +       }
>>>
>>>          if (!vma_is_anonymous(vma)) {
>>>                  /*
>>>
>>>> Couldn't (shouldn't) thp_vma_allowable_orders() be taught to handle 
>>>> shmem too?
>>>>
>>>>> +    seq_printf(m, "THPeligible:    %8u\n", thp_eligible);
>>>>>       if (arch_pkeys_enabled())
>>>>>           seq_printf(m, "ProtectionKey:  %8u\n", vma_pkey(vma));
>>>>> diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
>>>>> index 212cca384d7e..f87136f38aa1 100644
>>>>> --- a/include/linux/huge_mm.h
>>>>> +++ b/include/linux/huge_mm.h
>>>>> @@ -267,6 +267,10 @@ unsigned long thp_vma_allowable_orders(struct 
>>>>> vm_area_struct *vma,
>>>>>       return __thp_vma_allowable_orders(vma, vm_flags, tva_flags, 
>>>>> orders);
>>>>>   }
>>>>> +unsigned long shmem_allowable_huge_orders(struct inode *inode,
>>>>> +                struct vm_area_struct *vma, pgoff_t index,
>>>>> +                bool global_huge);
>>>>> +
>>>>>   struct thpsize {
>>>>>       struct kobject kobj;
>>>>>       struct list_head node;
>>>>> @@ -460,6 +464,13 @@ static inline unsigned long 
>>>>> thp_vma_allowable_orders(struct vm_area_struct *vma,
>>>>>       return 0;
>>>>>   }
>>>>> +static inline unsigned long shmem_allowable_huge_orders(struct 
>>>>> inode *inode,
>>>>> +                struct vm_area_struct *vma, pgoff_t index,
>>>>> +                bool global_huge)
>>>>> +{
>>>>> +    return 0;
>>>>> +}
>>>>> +
>>>>>   #define transparent_hugepage_flags 0UL
>>>>>   #define thp_get_unmapped_area    NULL
>>>>> diff --git a/mm/shmem.c b/mm/shmem.c
>>>>> index d495c0701a83..aa85df9c662a 100644
>>>>> --- a/mm/shmem.c
>>>>> +++ b/mm/shmem.c
>>>>> @@ -1622,7 +1622,7 @@ static gfp_t limit_gfp_mask(gfp_t huge_gfp, 
>>>>> gfp_t limit_gfp)
>>>>>   }
>>>>>   #ifdef CONFIG_TRANSPARENT_HUGEPAGE
>>>>> -static unsigned long shmem_allowable_huge_orders(struct inode *inode,
>>>>> +unsigned long shmem_allowable_huge_orders(struct inode *inode,
>>>>>                   struct vm_area_struct *vma, pgoff_t index,
>>>>>                   bool global_huge)
>>>>>   {
>>>>> @@ -1707,13 +1707,6 @@ static unsigned long 
>>>>> shmem_suitable_orders(struct inode *inode, struct vm_fault
>>>>>       return orders;
>>>>>   }
>>>>>   #else
>>>>> -static unsigned long shmem_allowable_huge_orders(struct inode *inode,
>>>>> -                struct vm_area_struct *vma, pgoff_t index,
>>>>> -                bool global_huge)
>>>>> -{
>>>>> -    return 0;
>>>>> -}
>>>>> -
>>>>>   static unsigned long shmem_suitable_orders(struct inode *inode, 
>>>>> struct vm_fault *vmf,
>>>>>                          struct address_space *mapping, pgoff_t index,
>>>>>                          unsigned long orders)
>>
>> Thanks for the reference code. Currently, we only implement the mTHP of
>> anonymous shmem, so we only need to handle anonymous shmem specially. As
>> shown in the following code:
>>
>> --- a/mm/huge_memory.c
>> +++ b/mm/huge_memory.c
>> @@ -151,10 +151,14 @@ unsigned long __thp_vma_allowable_orders(struct 
>> vm_area_struct *vma,
>>           * Must be done before hugepage flags check since shmem has its
>>           * own flags.
>>           */
>> -       if (!in_pf && shmem_file(vma->vm_file))
>> -               return shmem_is_huge(file_inode(vma->vm_file), 
>> vma->vm_pgoff,
>> -                                    !enforce_sysfs, vma->vm_mm, 
>> vm_flags)
>> -                       ? orders : 0;
>> +       if (!in_pf && shmem_file(vma->vm_file)) {
>> +               bool global_huge = 
>> shmem_is_huge(file_inode(vma->vm_file), vma->vm_pgoff,
>> +                                    !enforce_sysfs, vma->vm_mm, 
>> vm_flags);
> 
> Nit: add a blank line after the declaration. Otherwise looks good to me.

It doesn't matter to me whether I add spaces or not, thanks for your
suggestion anyway.

Thanks,
Bang

> 
>> +               if (!vma_is_anon_shmem(vma))
>> +                       return global_huge? orders : 0;
>> +               return 
>> shmem_allowable_huge_orders(file_inode(vma->vm_file),
>> +                                               vma, vma->vm_pgoff, 
>> global_huge);
>> +       }
>>
>>          if (!vma_is_anonymous(vma)) {
>>
>> Thanks,
>> Bang