Re: [PATCH] mm: Add PageLayzyFree() helper functions for MADV_FREE

["Huang\, Ying" <ying.huang@intel.com>]

Hi, Andrew,

Andrew Morton <akpm@linux-foundation.org> writes:

> On Tue,  3 Mar 2020 11:37:38 +0800 "Huang, Ying" <ying.huang@intel.com> wrote:
>
>> From: Huang Ying <ying.huang@intel.com>
>> 
>> Now PageSwapBacked() is used as the helper function to check whether
>> pages have been freed lazily via MADV_FREE.  This isn't very obvious.
>> So Dave suggested to add PageLazyFree() family helper functions to
>> improve the code readability.
>> 
>> --- a/include/linux/page-flags.h
>> +++ b/include/linux/page-flags.h
>> @@ -498,6 +498,31 @@ static __always_inline int PageKsm(struct page *page)
>>  TESTPAGEFLAG_FALSE(Ksm)
>>  #endif
>>  
>> +/*
>> + * For pages freed lazily via MADV_FREE.  lazyfree pages are clean
>> + * anonymous pages.  They have SwapBacked flag cleared to distinguish
>> + * with normal anonymous pages
>> + */
>> +static __always_inline int PageLazyFree(struct page *page)
>> +{
>> +	page = compound_head(page);
>> +	return PageAnon(page) && !PageSwapBacked(page);
>> +}
>> +
>> +static __always_inline void SetPageLazyFree(struct page *page)
>> +{
>> +	VM_BUG_ON_PAGE(PageTail(page), page);
>> +	VM_BUG_ON_PAGE(!PageAnon(page), page);
>> +	ClearPageSwapBacked(page);
>> +}
>> +
>> +static __always_inline void ClearPageLazyFree(struct page *page)
>> +{
>> +	VM_BUG_ON_PAGE(PageTail(page), page);
>> +	VM_BUG_ON_PAGE(!PageAnon(page), page);
>> +	SetPageSwapBacked(page);
>> +}
>
> These BUG_ONs aren't present in the current code and are
> unchangelogged.
>
> And they aren't free!
>
> before:
>
> q:/usr/src/25> size -t mm/rmap.o mm/swap.o mm/vmscan.o 
>    text    data     bss     dec     hex filename
>   41580    4211     192   45983    b39f mm/rmap.o
>   50684    9259    1184   61127    eec7 mm/swap.o
>  117728   46775     128  164631   28317 mm/vmscan.o
>  209992   60245    1504  271741   4257d (TOTALS)
>
> after:
>
>    text    data     bss     dec     hex filename
>   42035    4299     192   46526    b5be mm/rmap.o
>   51091    9347    1184   61622    f0b6 mm/swap.o
>  118132   46775     128  165035   284ab mm/vmscan.o
>  211258   60421    1504  273183   42b1f (TOTALS)
>
> ALso, if they actually trigger, Linus goes berserk.
>
> If you have some special reason to add these assertions, please do it
> as a (changelogged!) followup patch and I'll keep it as a linux-next only thing.

Sorry, originally I thought that VM_BUG_ON_PAGE() will not be used in
the production kernel, so it's cheap.  But it seems that people do care
about its usage.  I will send v2 to fix this.

Best Regards,
Huang, Ying