[PATCH] mm: memory-failure: remove unneeded page state check in shake_page()

[PATCH] mm: memory-failure: remove unneeded page state check in shake_page()

[Miaohe Lin]

Remove unneeded PageLRU(p) and is_free_buddy_page(p) check as slab caches
are not shrunk now. This check can be added back when a lightweight range
based shrinker is available.

Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
---
 mm/memory-failure.c | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)

diff --git a/mm/memory-failure.c b/mm/memory-failure.c
index 5b663eca1f29..92f951df3e87 100644
--- a/mm/memory-failure.c
+++ b/mm/memory-failure.c
@@ -373,11 +373,10 @@ void shake_page(struct page *p)
 	if (PageHuge(p))
 		return;
 
-	if (!PageSlab(p)) {
-		lru_add_drain_all();
-		if (PageLRU(p) || is_free_buddy_page(p))
-			return;
-	}
+	if (PageSlab(p))
+		return;
+
+	lru_add_drain_all();
 
 	/*
 	 * TODO: Could shrink slab caches here if a lightweight range-based
-- 
2.27.0

Re: [PATCH] mm: memory-failure: remove unneeded page state check in shake_page()

[Matthew Wilcox]

On Sun, Jun 25, 2023 at 07:34:30PM +0800, Miaohe Lin wrote:
> Remove unneeded PageLRU(p) and is_free_buddy_page(p) check as slab caches
> are not shrunk now. This check can be added back when a lightweight range
> based shrinker is available.

I think your logic is faulty.  Based on the reasoning above, this patch
should be:

> -	if (!PageSlab(p)) {
> -		lru_add_drain_all();
> -		if (PageLRU(p) || is_free_buddy_page(p))
> -			return;
> -	}
> +	if (PageSlab(p))
> +		return;
> +
> +	lru_add_drain_all();
+	if (PageLRU(p) || is_free_buddy_page(p))
+		return;

but maybe it's the description that is wrong.

Re: [PATCH] mm: memory-failure: remove unneeded page state check in shake_page()

[Matthew Wilcox]

On Sun, Jun 25, 2023 at 07:34:30PM +0800, Miaohe Lin wrote:
> Remove unneeded PageLRU(p) and is_free_buddy_page(p) check as slab caches
> are not shrunk now. This check can be added back when a lightweight range
> based shrinker is available.

On further review, I think you've misunderstood the entire situation
here and no patch should be applied.

> Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
> ---
>  mm/memory-failure.c | 9 ++++-----
>  1 file changed, 4 insertions(+), 5 deletions(-)
> 
> diff --git a/mm/memory-failure.c b/mm/memory-failure.c
> index 5b663eca1f29..92f951df3e87 100644
> --- a/mm/memory-failure.c
> +++ b/mm/memory-failure.c
> @@ -373,11 +373,10 @@ void shake_page(struct page *p)
>  	if (PageHuge(p))
>  		return;
>  
> -	if (!PageSlab(p)) {
> -		lru_add_drain_all();
> -		if (PageLRU(p) || is_free_buddy_page(p))
> -			return;
> -	}
> +	if (PageSlab(p))
> +		return;
> +
> +	lru_add_drain_all();
>  
>  	/*
>  	 * TODO: Could shrink slab caches here if a lightweight range-based
> -- 
> 2.27.0
> 
> 

Re: [PATCH] mm: memory-failure: remove unneeded page state check in shake_page()

[Miaohe Lin]

On 2023/6/26 3:50, Matthew Wilcox wrote:
> On Sun, Jun 25, 2023 at 07:34:30PM +0800, Miaohe Lin wrote:
>> Remove unneeded PageLRU(p) and is_free_buddy_page(p) check as slab caches
>> are not shrunk now. This check can be added back when a lightweight range
>> based shrinker is available.
> 
> On further review, I think you've misunderstood the entire situation
> here and no patch should be applied.

Thanks for your review. I think you might misunderstood the patch. What I want
to do here is remove the unneeded if block:

 if (PageLRU(p) || is_free_buddy_page(p))
     return;

Because nothing is done after the check. So this check is useless and should be removed.
Thanks.

Re: [PATCH] mm: memory-failure: remove unneeded page state check in shake_page()

[Naoya Horiguchi]

On Sun, Jun 25, 2023 at 07:34:30PM +0800, Miaohe Lin wrote:
> Remove unneeded PageLRU(p) and is_free_buddy_page(p) check as slab caches
> are not shrunk now. This check can be added back when a lightweight range
> based shrinker is available.
> 
> Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>

This looks to me a good cleanup because the result of
"if (PageLRU(p) || is_free_buddy_page(p))" check is not used, so the check
itself is unneeded.

> ---
>  mm/memory-failure.c | 9 ++++-----
>  1 file changed, 4 insertions(+), 5 deletions(-)
> 
> diff --git a/mm/memory-failure.c b/mm/memory-failure.c
> index 5b663eca1f29..92f951df3e87 100644
> --- a/mm/memory-failure.c
> +++ b/mm/memory-failure.c
> @@ -373,11 +373,10 @@ void shake_page(struct page *p)
>  	if (PageHuge(p))
>  		return;
>  
> -	if (!PageSlab(p)) {
> -		lru_add_drain_all();
> -		if (PageLRU(p) || is_free_buddy_page(p))
> -			return;
> -	}
> +	if (PageSlab(p))
> +		return;
> +
> +	lru_add_drain_all();
>  
>  	/*
>  	 * TODO: Could shrink slab caches here if a lightweight range-based

I think that this TODO comment can be put together with "if (PageSlab)" block.

Thanks,
Naoya Horiguchi

Re: [PATCH] mm: memory-failure: remove unneeded page state check in shake_page()

[Miaohe Lin]

On 2023/6/26 8:52, Naoya Horiguchi wrote:
> On Sun, Jun 25, 2023 at 07:34:30PM +0800, Miaohe Lin wrote:
>> Remove unneeded PageLRU(p) and is_free_buddy_page(p) check as slab caches
>> are not shrunk now. This check can be added back when a lightweight range
>> based shrinker is available.
>>
>> Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
> 
> This looks to me a good cleanup because the result of
> "if (PageLRU(p) || is_free_buddy_page(p))" check is not used, so the check
> itself is unneeded.
> 
>> ---
>>  mm/memory-failure.c | 9 ++++-----
>>  1 file changed, 4 insertions(+), 5 deletions(-)
>>
>> diff --git a/mm/memory-failure.c b/mm/memory-failure.c
>> index 5b663eca1f29..92f951df3e87 100644
>> --- a/mm/memory-failure.c
>> +++ b/mm/memory-failure.c
>> @@ -373,11 +373,10 @@ void shake_page(struct page *p)
>>  	if (PageHuge(p))
>>  		return;
>>  
>> -	if (!PageSlab(p)) {
>> -		lru_add_drain_all();
>> -		if (PageLRU(p) || is_free_buddy_page(p))
>> -			return;
>> -	}
>> +	if (PageSlab(p))
>> +		return;
>> +
>> +	lru_add_drain_all();
>>  
>>  	/*
>>  	 * TODO: Could shrink slab caches here if a lightweight range-based
> 
> I think that this TODO comment can be put together with "if (PageSlab)" block.

Thanks for your comment and advice. Do you mean something like below?

diff --git a/mm/memory-failure.c b/mm/memory-failure.c
index 5b663eca1f29..66e7b3ceaf2d 100644
--- a/mm/memory-failure.c
+++ b/mm/memory-failure.c
@@ -372,17 +372,14 @@ void shake_page(struct page *p)
 {
        if (PageHuge(p))
                return;
-
-       if (!PageSlab(p)) {
-               lru_add_drain_all();
-               if (PageLRU(p) || is_free_buddy_page(p))
-                       return;
-       }
-
        /*
         * TODO: Could shrink slab caches here if a lightweight range-based
         * shrinker will be available.
         */
+       if (PageSlab(p))
+               return;
+
+       lru_add_drain_all();
 }
 EXPORT_SYMBOL_GPL(shake_page);

Thanks.

Re: [PATCH] mm: memory-failure: remove unneeded page state check in shake_page()

[Naoya Horiguchi]

On Mon, Jun 26, 2023 at 09:43:44AM +0800, Miaohe Lin wrote:
> On 2023/6/26 8:52, Naoya Horiguchi wrote:
> > On Sun, Jun 25, 2023 at 07:34:30PM +0800, Miaohe Lin wrote:
> >> Remove unneeded PageLRU(p) and is_free_buddy_page(p) check as slab caches
> >> are not shrunk now. This check can be added back when a lightweight range
> >> based shrinker is available.
> >>
> >> Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
> > 
> > This looks to me a good cleanup because the result of
> > "if (PageLRU(p) || is_free_buddy_page(p))" check is not used, so the check
> > itself is unneeded.
> > 
> >> ---
> >>  mm/memory-failure.c | 9 ++++-----
> >>  1 file changed, 4 insertions(+), 5 deletions(-)
> >>
> >> diff --git a/mm/memory-failure.c b/mm/memory-failure.c
> >> index 5b663eca1f29..92f951df3e87 100644
> >> --- a/mm/memory-failure.c
> >> +++ b/mm/memory-failure.c
> >> @@ -373,11 +373,10 @@ void shake_page(struct page *p)
> >>  	if (PageHuge(p))
> >>  		return;
> >>  
> >> -	if (!PageSlab(p)) {
> >> -		lru_add_drain_all();
> >> -		if (PageLRU(p) || is_free_buddy_page(p))
> >> -			return;
> >> -	}
> >> +	if (PageSlab(p))
> >> +		return;
> >> +
> >> +	lru_add_drain_all();
> >>  
> >>  	/*
> >>  	 * TODO: Could shrink slab caches here if a lightweight range-based
> > 
> > I think that this TODO comment can be put together with "if (PageSlab)" block.
> 
> Thanks for your comment and advice. Do you mean something like below?

Yes, this is what I mean.  Feel free to add my ack when posting v2.

Acked-by: Naoya Horiguchi <naoya.horiguchi@nec.com>

Thank you.

> 
> diff --git a/mm/memory-failure.c b/mm/memory-failure.c
> index 5b663eca1f29..66e7b3ceaf2d 100644
> --- a/mm/memory-failure.c
> +++ b/mm/memory-failure.c
> @@ -372,17 +372,14 @@ void shake_page(struct page *p)
>  {
>         if (PageHuge(p))
>                 return;
> -
> -       if (!PageSlab(p)) {
> -               lru_add_drain_all();
> -               if (PageLRU(p) || is_free_buddy_page(p))
> -                       return;
> -       }
> -
>         /*
>          * TODO: Could shrink slab caches here if a lightweight range-based
>          * shrinker will be available.
>          */
> +       if (PageSlab(p))
> +               return;
> +
> +       lru_add_drain_all();
>  }
>  EXPORT_SYMBOL_GPL(shake_page);
> 
> Thanks.
> 
>