Re: [PATCH V2 2/5] mm/khugepaged: count small VMAs towards scan limit

["Garg, Shivank" <shivankg@amd.com>]

[-- Attachment #1: Type: text/plain, Size: 3196 bytes --]



On 12/24/2025 8:19 PM, Wei Yang wrote:
> On Wed, Dec 24, 2025 at 07:51:36PM +0800, Lance Yang wrote:
>>
>>
>> On 2025/12/24 19:13, Shivank Garg wrote:
>>> The khugepaged_scan_mm_slot() uses a 'progress' counter to limit the
>>> amount of work performed and consists of three components:
>>> 1. Transitioning to a new mm (+1).
> 
> Hmm... maybe not only a new mm, but also we start another scan from last mm.
> 
> Since default khugepaged_pages_to_scan is 8 PMD, it looks very possible.
> 
It makes sense, will correct this.

>>> 2. Skipping an unsuitable VMA (+1).
>>> 3. Scanning a PMD-sized range (+HPAGE_PMD_NR).
>>>
>>> Consider a 1MB VMA sitting between two 2MB alignment boundaries:
>>>
>>>       vma1       vma2   vma3
>>>      +----------+------+----------+
>>>      |2M        |1M    |2M        |
>>>      +----------+------+----------+
>>>                 ^      ^
>>>                 start  end
>>>                 ^
>>>            hstart,hend
>>>
>>> In this case, for vma2:
>>>    hstart = round_up(start, HPAGE_PMD_SIZE)  -> Next 2MB alignment
>>>    hend   = round_down(end, HPAGE_PMD_SIZE) -> Prev 2MB alignment
>>>
>>> Currently, since `hend <= hstart`, VMAs that are too small or unaligned
>>> to contain a hugepage are skipped without incrementing 'progress'.
>>> A process containing a large number of such small VMAs will unfairly
>>> consume more CPU cycles before yielding compared to a process with
>>> fewer, larger, or aligned VMAs.
>>>
>>> Fix this by incrementing progress when the `hend <= hstart` condition
>>> is met.
>>>
>>> Additionally, change 'progress' type to `unsigned int` to match both
>>> the 'pages' type and the function return value.
>>>
>>> Suggested-by: Wei Yang <richard.weiyang@gmail.com>
>>> Signed-off-by: Shivank Garg <shivankg@amd.com>
>>> ---
>>>   mm/khugepaged.c | 4 ++--
>>>   1 file changed, 2 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/mm/khugepaged.c b/mm/khugepaged.c
>>> index 107146f012b1..0b549c3250f9 100644
>>> --- a/mm/khugepaged.c
>>> +++ b/mm/khugepaged.c
>>> @@ -2403,7 +2403,7 @@ static unsigned int khugepaged_scan_mm_slot(unsigned int pages, int *result,
>>>   	struct mm_slot *slot;
>>>   	struct mm_struct *mm;
>>>   	struct vm_area_struct *vma;
>>> -	int progress = 0;
>>> +	unsigned int progress = 0;
>>>   	VM_BUG_ON(!pages);
>>>   	lockdep_assert_held(&khugepaged_mm_lock);
>>> @@ -2447,7 +2447,7 @@ static unsigned int khugepaged_scan_mm_slot(unsigned int pages, int *result,
>>>   		}
>>>   		hstart = round_up(vma->vm_start, HPAGE_PMD_SIZE);
>>>   		hend = round_down(vma->vm_end, HPAGE_PMD_SIZE);
>>> -		if (khugepaged_scan.address > hend) {
>>
>> Maybe add a short comment explaining why we increment progress for small VMAs
>> ;)
>>
>> Something like this:
>>
>> 		/* Count small VMAs that can't hold a hugepage towards scan limit */

I'll add explanation.

>>> +		if (khugepaged_scan.address > hend || hend <= hstart) {
>>>   			progress++;
>>>   			continue;
>>>   		}
>>
>> Otherwise, looks good to me.
>>
>> Reviewed-by: Lance Yang <lance.yang@linux.dev>
>>
> 
> The code change LGTM.
> 
> Reviewed-by: Wei Yang <richard.weiyang@gmail.com>
> 

Thanks Lance and Wei. I have made suggested changes.


[-- Attachment #2: 0002-mm-khugepaged-count-small-VMAs-towards-scan-limit.patch --]
[-- Type: text/plain, Size: 2550 bytes --]

From d464604c09cef70f0f2aa0f9607a977b4bcd7081 Mon Sep 17 00:00:00 2001
From: Shivank Garg <shivankg@amd.com>
Date: Wed, 17 Dec 2025 19:43:26 +0000
Subject: [PATCH V2 2/5] mm/khugepaged: count small VMAs towards scan limit

The khugepaged_scan_mm_slot() uses a 'progress' counter to limit the
amount of work performed and consists of three components:
1. Starting/resuming scan of mm slot (+1).
2. Skipping an unsuitable VMA (+1).
3. Scanning a PMD-sized range (+HPAGE_PMD_NR).

Consider a 1MB VMA sitting between two 2MB alignment boundaries:

     vma1       vma2   vma3
    +----------+------+----------+
    |2M        |1M    |2M        |
    +----------+------+----------+
               ^      ^
               start  end
               ^
          hstart,hend

In this case, for vma2:
  hstart = round_up(start, HPAGE_PMD_SIZE)  -> Next 2MB alignment
  hend   = round_down(end, HPAGE_PMD_SIZE) -> Prev 2MB alignment

Currently, since `hend <= hstart`, VMAs that are too small or unaligned
to contain a hugepage are skipped without incrementing 'progress'.
A process containing a large number of such small VMAs will unfairly
consume more CPU cycles before yielding compared to a process with
fewer, larger, or aligned VMAs.

Fix this by incrementing progress when the `hend <= hstart` condition
is met.

Additionally, change 'progress' to `unsigned int`. This matches both
the 'pages' type and the function return value.

Suggested-by: Wei Yang <richard.weiyang@gmail.com>
Reviewed-by: Wei Yang <richard.weiyang@gmail.com>
Reviewed-by: Lance Yang <lance.yang@linux.dev>
Signed-off-by: Shivank Garg <shivankg@amd.com>
---
 mm/khugepaged.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index 107146f012b1..155281c49169 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -2403,7 +2403,7 @@ static unsigned int khugepaged_scan_mm_slot(unsigned int pages, int *result,
 	struct mm_slot *slot;
 	struct mm_struct *mm;
 	struct vm_area_struct *vma;
-	int progress = 0;
+	unsigned int progress = 0;
 
 	VM_BUG_ON(!pages);
 	lockdep_assert_held(&khugepaged_mm_lock);
@@ -2447,7 +2447,8 @@ static unsigned int khugepaged_scan_mm_slot(unsigned int pages, int *result,
 		}
 		hstart = round_up(vma->vm_start, HPAGE_PMD_SIZE);
 		hend = round_down(vma->vm_end, HPAGE_PMD_SIZE);
-		if (khugepaged_scan.address > hend) {
+		if (khugepaged_scan.address > hend || hend <= hstart) {
+			/* VMA already scanned or too small/unaligned for hugepage. */
 			progress++;
 			continue;
 		}
-- 
2.43.0