linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
* [PATCH] mm/memory.c: Introduce non-atomic __{Set,Clear}PageSwapCache
@ 2020-10-19 10:15 Muchun Song
  2020-10-19 12:31 ` Michal Hocko
  0 siblings, 1 reply; 6+ messages in thread
From: Muchun Song @ 2020-10-19 10:15 UTC (permalink / raw)
  To: akpm, sfr, osalvador, alexander.h.duyck, mhocko, yang.shi, david,
	hannes, hughd
  Cc: linux-kernel, linux-mm, Muchun Song

For the exclusive reference page, the non-atomic operations is enough,
so replace them to non-atomic operations.

Signed-off-by: Muchun Song <songmuchun@bytedance.com>
---
 include/linux/page-flags.h | 2 ++
 mm/memory.c                | 4 ++--
 2 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/include/linux/page-flags.h b/include/linux/page-flags.h
index fbbb841a9346..ec039dde5e4b 100644
--- a/include/linux/page-flags.h
+++ b/include/linux/page-flags.h
@@ -401,6 +401,8 @@ static __always_inline int PageSwapCache(struct page *page)
 }
 SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
 CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
+__SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
+__CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
 #else
 PAGEFLAG_FALSE(SwapCache)
 #endif
diff --git a/mm/memory.c b/mm/memory.c
index 2d267ef6621a..02dd62da26e0 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -3128,10 +3128,10 @@ vm_fault_t do_swap_page(struct vm_fault *vmf)
 				set_page_private(page, entry.val);
 
 				/* Tell memcg to use swap ownership records */
-				SetPageSwapCache(page);
+				__SetPageSwapCache(page);
 				err = mem_cgroup_charge(page, vma->vm_mm,
 							GFP_KERNEL);
-				ClearPageSwapCache(page);
+				__ClearPageSwapCache(page);
 				if (err) {
 					ret = VM_FAULT_OOM;
 					goto out_page;
-- 
2.20.1



^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [PATCH] mm/memory.c: Introduce non-atomic __{Set,Clear}PageSwapCache
  2020-10-19 10:15 [PATCH] mm/memory.c: Introduce non-atomic __{Set,Clear}PageSwapCache Muchun Song
@ 2020-10-19 12:31 ` Michal Hocko
  2020-10-19 14:58   ` [External] " Muchun Song
  0 siblings, 1 reply; 6+ messages in thread
From: Michal Hocko @ 2020-10-19 12:31 UTC (permalink / raw)
  To: Muchun Song
  Cc: akpm, sfr, osalvador, alexander.h.duyck, yang.shi, david, hannes,
	hughd, linux-kernel, linux-mm

On Mon 19-10-20 18:15:20, Muchun Song wrote:
> For the exclusive reference page, the non-atomic operations is enough,
> so replace them to non-atomic operations.

I do expect you do not see any difference in runtime and this is mostly
driven by the code reading, right? Being explicit about this in the code
would be preferred.

No objection to the change.

> Signed-off-by: Muchun Song <songmuchun@bytedance.com>

With an improved changelog
Acked-by: Michal Hocko <mhocko@suse.com>

> ---
>  include/linux/page-flags.h | 2 ++
>  mm/memory.c                | 4 ++--
>  2 files changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/include/linux/page-flags.h b/include/linux/page-flags.h
> index fbbb841a9346..ec039dde5e4b 100644
> --- a/include/linux/page-flags.h
> +++ b/include/linux/page-flags.h
> @@ -401,6 +401,8 @@ static __always_inline int PageSwapCache(struct page *page)
>  }
>  SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
>  CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
> +__SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
> +__CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
>  #else
>  PAGEFLAG_FALSE(SwapCache)
>  #endif
> diff --git a/mm/memory.c b/mm/memory.c
> index 2d267ef6621a..02dd62da26e0 100644
> --- a/mm/memory.c
> +++ b/mm/memory.c
> @@ -3128,10 +3128,10 @@ vm_fault_t do_swap_page(struct vm_fault *vmf)
>  				set_page_private(page, entry.val);
>  
>  				/* Tell memcg to use swap ownership records */
> -				SetPageSwapCache(page);
> +				__SetPageSwapCache(page);
>  				err = mem_cgroup_charge(page, vma->vm_mm,
>  							GFP_KERNEL);
> -				ClearPageSwapCache(page);
> +				__ClearPageSwapCache(page);
>  				if (err) {
>  					ret = VM_FAULT_OOM;
>  					goto out_page;
> -- 
> 2.20.1
> 

-- 
Michal Hocko
SUSE Labs


^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [External] Re: [PATCH] mm/memory.c: Introduce non-atomic __{Set,Clear}PageSwapCache
  2020-10-19 12:31 ` Michal Hocko
@ 2020-10-19 14:58   ` Muchun Song
  2020-10-20 11:50     ` Xu, Yanfei
  0 siblings, 1 reply; 6+ messages in thread
From: Muchun Song @ 2020-10-19 14:58 UTC (permalink / raw)
  To: Michal Hocko
  Cc: Andrew Morton, Stephen Rothwell, osalvador, alexander.h.duyck,
	yang.shi, David Hildenbrand, Johannes Weiner, Hugh Dickins, LKML,
	Linux Memory Management List

On Mon, Oct 19, 2020 at 8:31 PM Michal Hocko <mhocko@suse.com> wrote:
>
> On Mon 19-10-20 18:15:20, Muchun Song wrote:
> > For the exclusive reference page, the non-atomic operations is enough,
> > so replace them to non-atomic operations.
>
> I do expect you do not see any difference in runtime and this is mostly
> driven by the code reading, right? Being explicit about this in the code
> would be preferred.

Yeah, just code reading.And the set_bit and __set_bit is actually different
on some architectures. Thanks.

>
> No objection to the change.
>
> > Signed-off-by: Muchun Song <songmuchun@bytedance.com>
>
> With an improved changelog
> Acked-by: Michal Hocko <mhocko@suse.com>
>
> > ---
> >  include/linux/page-flags.h | 2 ++
> >  mm/memory.c                | 4 ++--
> >  2 files changed, 4 insertions(+), 2 deletions(-)
> >
> > diff --git a/include/linux/page-flags.h b/include/linux/page-flags.h
> > index fbbb841a9346..ec039dde5e4b 100644
> > --- a/include/linux/page-flags.h
> > +++ b/include/linux/page-flags.h
> > @@ -401,6 +401,8 @@ static __always_inline int PageSwapCache(struct page *page)
> >  }
> >  SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
> >  CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
> > +__SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
> > +__CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
> >  #else
> >  PAGEFLAG_FALSE(SwapCache)
> >  #endif
> > diff --git a/mm/memory.c b/mm/memory.c
> > index 2d267ef6621a..02dd62da26e0 100644
> > --- a/mm/memory.c
> > +++ b/mm/memory.c
> > @@ -3128,10 +3128,10 @@ vm_fault_t do_swap_page(struct vm_fault *vmf)
> >                               set_page_private(page, entry.val);
> >
> >                               /* Tell memcg to use swap ownership records */
> > -                             SetPageSwapCache(page);
> > +                             __SetPageSwapCache(page);
> >                               err = mem_cgroup_charge(page, vma->vm_mm,
> >                                                       GFP_KERNEL);
> > -                             ClearPageSwapCache(page);
> > +                             __ClearPageSwapCache(page);
> >                               if (err) {
> >                                       ret = VM_FAULT_OOM;
> >                                       goto out_page;
> > --
> > 2.20.1
> >
>
> --
> Michal Hocko
> SUSE Labs



-- 
Yours,
Muchun


^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [External] Re: [PATCH] mm/memory.c: Introduce non-atomic __{Set,Clear}PageSwapCache
  2020-10-19 14:58   ` [External] " Muchun Song
@ 2020-10-20 11:50     ` Xu, Yanfei
  2020-10-20 13:08       ` Muchun Song
  0 siblings, 1 reply; 6+ messages in thread
From: Xu, Yanfei @ 2020-10-20 11:50 UTC (permalink / raw)
  To: Muchun Song, Michal Hocko
  Cc: Andrew Morton, Stephen Rothwell, osalvador, alexander.h.duyck,
	yang.shi, David Hildenbrand, Johannes Weiner, Hugh Dickins, LKML,
	Linux Memory Management List



On 10/19/20 10:58 PM, Muchun Song wrote:
> On Mon, Oct 19, 2020 at 8:31 PM Michal Hocko <mhocko@suse.com> wrote:
>>
>> On Mon 19-10-20 18:15:20, Muchun Song wrote:
>>> For the exclusive reference page, the non-atomic operations is enough,
>>> so replace them to non-atomic operations.
>>
>> I do expect you do not see any difference in runtime and this is mostly
>> driven by the code reading, right? Being explicit about this in the code
>> would be preferred.
> 
> Yeah, just code reading.And the set_bit and __set_bit is actually different
> on some architectures. Thanks.
> 
>>
>> No objection to the change.
>>
>>> Signed-off-by: Muchun Song <songmuchun@bytedance.com>
>>
>> With an improved changelog
>> Acked-by: Michal Hocko <mhocko@suse.com>
>>
>>> ---
>>>   include/linux/page-flags.h | 2 ++
>>>   mm/memory.c                | 4 ++--
>>>   2 files changed, 4 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/include/linux/page-flags.h b/include/linux/page-flags.h
>>> index fbbb841a9346..ec039dde5e4b 100644
>>> --- a/include/linux/page-flags.h
>>> +++ b/include/linux/page-flags.h
>>> @@ -401,6 +401,8 @@ static __always_inline int PageSwapCache(struct page *page)
>>>   }
>>>   SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
>>>   CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
>>> +__SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
>>> +__CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
>>>   #else
>>>   PAGEFLAG_FALSE(SwapCache)
>>>   #endif
>>> diff --git a/mm/memory.c b/mm/memory.c
>>> index 2d267ef6621a..02dd62da26e0 100644
>>> --- a/mm/memory.c
>>> +++ b/mm/memory.c
>>> @@ -3128,10 +3128,10 @@ vm_fault_t do_swap_page(struct vm_fault *vmf)
>>>                                set_page_private(page, entry.val);
>>>
>>>                                /* Tell memcg to use swap ownership records */
>>> -                             SetPageSwapCache(page);
>>> +                             __SetPageSwapCache(page);

Good evening, Muchun. I found there are still some places could be 
replaced with __SetPageSwapCache(). Such as shmem_replace_page(), why 
PG_locked has been set before SetPageSwapCache() is involved.

Would you please to check the rest places? :)

Thanks

Acked-by: Yanfei Xu <yanfei.xu@windriver.com>

>>>                                err = mem_cgroup_charge(page, vma->vm_mm,
>>>                                                        GFP_KERNEL);
>>> -                             ClearPageSwapCache(page);
>>> +                             __ClearPageSwapCache(page);
>>>                                if (err) {
>>>                                        ret = VM_FAULT_OOM;
>>>                                        goto out_page;
>>> --
>>> 2.20.1
>>>
>>
>> --
>> Michal Hocko
>> SUSE Labs
> 
> 
> 


^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [External] Re: [PATCH] mm/memory.c: Introduce non-atomic __{Set,Clear}PageSwapCache
  2020-10-20 11:50     ` Xu, Yanfei
@ 2020-10-20 13:08       ` Muchun Song
  2020-10-21  2:52         ` Xu, Yanfei
  0 siblings, 1 reply; 6+ messages in thread
From: Muchun Song @ 2020-10-20 13:08 UTC (permalink / raw)
  To: Xu, Yanfei
  Cc: Michal Hocko, Andrew Morton, Stephen Rothwell, osalvador,
	alexander.h.duyck, yang.shi, David Hildenbrand, Johannes Weiner,
	Hugh Dickins, LKML, Linux Memory Management List

On Tue, Oct 20, 2020 at 7:51 PM Xu, Yanfei <yanfei.xu@windriver.com> wrote:
>
>
>
> On 10/19/20 10:58 PM, Muchun Song wrote:
> > On Mon, Oct 19, 2020 at 8:31 PM Michal Hocko <mhocko@suse.com> wrote:
> >>
> >> On Mon 19-10-20 18:15:20, Muchun Song wrote:
> >>> For the exclusive reference page, the non-atomic operations is enough,
> >>> so replace them to non-atomic operations.
> >>
> >> I do expect you do not see any difference in runtime and this is mostly
> >> driven by the code reading, right? Being explicit about this in the code
> >> would be preferred.
> >
> > Yeah, just code reading.And the set_bit and __set_bit is actually different
> > on some architectures. Thanks.
> >
> >>
> >> No objection to the change.
> >>
> >>> Signed-off-by: Muchun Song <songmuchun@bytedance.com>
> >>
> >> With an improved changelog
> >> Acked-by: Michal Hocko <mhocko@suse.com>
> >>
> >>> ---
> >>>   include/linux/page-flags.h | 2 ++
> >>>   mm/memory.c                | 4 ++--
> >>>   2 files changed, 4 insertions(+), 2 deletions(-)
> >>>
> >>> diff --git a/include/linux/page-flags.h b/include/linux/page-flags.h
> >>> index fbbb841a9346..ec039dde5e4b 100644
> >>> --- a/include/linux/page-flags.h
> >>> +++ b/include/linux/page-flags.h
> >>> @@ -401,6 +401,8 @@ static __always_inline int PageSwapCache(struct page *page)
> >>>   }
> >>>   SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
> >>>   CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
> >>> +__SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
> >>> +__CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
> >>>   #else
> >>>   PAGEFLAG_FALSE(SwapCache)
> >>>   #endif
> >>> diff --git a/mm/memory.c b/mm/memory.c
> >>> index 2d267ef6621a..02dd62da26e0 100644
> >>> --- a/mm/memory.c
> >>> +++ b/mm/memory.c
> >>> @@ -3128,10 +3128,10 @@ vm_fault_t do_swap_page(struct vm_fault *vmf)
> >>>                                set_page_private(page, entry.val);
> >>>
> >>>                                /* Tell memcg to use swap ownership records */
> >>> -                             SetPageSwapCache(page);
> >>> +                             __SetPageSwapCache(page);
>
> Good evening, Muchun. I found there are still some places could be
> replaced with __SetPageSwapCache(). Such as shmem_replace_page(), why

Yeah, thanks for your suggestion.

> PG_locked has been set before SetPageSwapCache() is involved.

In this case, It doesn't matter whether PG_locked is set before
SetPageSwapCache.

>
> Would you please to check the rest places? :)

Ok, I'll take a look. Thanks.

>
> Thanks
>
> Acked-by: Yanfei Xu <yanfei.xu@windriver.com>
>
> >>>                                err = mem_cgroup_charge(page, vma->vm_mm,
> >>>                                                        GFP_KERNEL);
> >>> -                             ClearPageSwapCache(page);
> >>> +                             __ClearPageSwapCache(page);
> >>>                                if (err) {
> >>>                                        ret = VM_FAULT_OOM;
> >>>                                        goto out_page;
> >>> --
> >>> 2.20.1
> >>>
> >>
> >> --
> >> Michal Hocko
> >> SUSE Labs
> >
> >
> >



--
Yours,
Muchun


^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [External] Re: [PATCH] mm/memory.c: Introduce non-atomic __{Set,Clear}PageSwapCache
  2020-10-20 13:08       ` Muchun Song
@ 2020-10-21  2:52         ` Xu, Yanfei
  0 siblings, 0 replies; 6+ messages in thread
From: Xu, Yanfei @ 2020-10-21  2:52 UTC (permalink / raw)
  To: Muchun Song
  Cc: Michal Hocko, Andrew Morton, Stephen Rothwell, osalvador,
	alexander.h.duyck, yang.shi, David Hildenbrand, Johannes Weiner,
	Hugh Dickins, LKML, Linux Memory Management List



On 10/20/20 9:08 PM, Muchun Song wrote:
> On Tue, Oct 20, 2020 at 7:51 PM Xu, Yanfei <yanfei.xu@windriver.com> wrote:
>>
>>
>>
>> On 10/19/20 10:58 PM, Muchun Song wrote:
>>> On Mon, Oct 19, 2020 at 8:31 PM Michal Hocko <mhocko@suse.com> wrote:
>>>>
>>>> On Mon 19-10-20 18:15:20, Muchun Song wrote:
>>>>> For the exclusive reference page, the non-atomic operations is enough,
>>>>> so replace them to non-atomic operations.
>>>>
>>>> I do expect you do not see any difference in runtime and this is mostly
>>>> driven by the code reading, right? Being explicit about this in the code
>>>> would be preferred.
>>>
>>> Yeah, just code reading.And the set_bit and __set_bit is actually different
>>> on some architectures. Thanks.
>>>
>>>>
>>>> No objection to the change.
>>>>
>>>>> Signed-off-by: Muchun Song <songmuchun@bytedance.com>
>>>>
>>>> With an improved changelog
>>>> Acked-by: Michal Hocko <mhocko@suse.com>
>>>>
>>>>> ---
>>>>>    include/linux/page-flags.h | 2 ++
>>>>>    mm/memory.c                | 4 ++--
>>>>>    2 files changed, 4 insertions(+), 2 deletions(-)
>>>>>
>>>>> diff --git a/include/linux/page-flags.h b/include/linux/page-flags.h
>>>>> index fbbb841a9346..ec039dde5e4b 100644
>>>>> --- a/include/linux/page-flags.h
>>>>> +++ b/include/linux/page-flags.h
>>>>> @@ -401,6 +401,8 @@ static __always_inline int PageSwapCache(struct page *page)
>>>>>    }
>>>>>    SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
>>>>>    CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
>>>>> +__SETPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
>>>>> +__CLEARPAGEFLAG(SwapCache, swapcache, PF_NO_TAIL)
>>>>>    #else
>>>>>    PAGEFLAG_FALSE(SwapCache)
>>>>>    #endif
>>>>> diff --git a/mm/memory.c b/mm/memory.c
>>>>> index 2d267ef6621a..02dd62da26e0 100644
>>>>> --- a/mm/memory.c
>>>>> +++ b/mm/memory.c
>>>>> @@ -3128,10 +3128,10 @@ vm_fault_t do_swap_page(struct vm_fault *vmf)
>>>>>                                 set_page_private(page, entry.val);
>>>>>
>>>>>                                 /* Tell memcg to use swap ownership records */
>>>>> -                             SetPageSwapCache(page);
>>>>> +                             __SetPageSwapCache(page);
>>
>> Good evening, Muchun. I found there are still some places could be
>> replaced with __SetPageSwapCache(). Such as shmem_replace_page(), why
> 
> Yeah, thanks for your suggestion.
> 
>> PG_locked has been set before SetPageSwapCache() is involved.
> 
> In this case, It doesn't matter whether PG_locked is set before
> SetPageSwapCache.
> 
Sorry for this mistake. PG_locked is used for disk I/O.
>>
>> Would you please to check the rest places? :)
> 
> Ok, I'll take a look. Thanks.
> 
>>
>> Thanks
>>
>> Acked-by: Yanfei Xu <yanfei.xu@windriver.com>
>>
>>>>>                                 err = mem_cgroup_charge(page, vma->vm_mm,
>>>>>                                                         GFP_KERNEL);
>>>>> -                             ClearPageSwapCache(page);
>>>>> +                             __ClearPageSwapCache(page);
>>>>>                                 if (err) {
>>>>>                                         ret = VM_FAULT_OOM;
>>>>>                                         goto out_page;
>>>>> --
>>>>> 2.20.1
>>>>>
>>>>
>>>> --
>>>> Michal Hocko
>>>> SUSE Labs
>>>
>>>
>>>
> 
> 
> 
> --
> Yours,
> Muchun
> 


^ permalink raw reply	[flat|nested] 6+ messages in thread

end of thread, other threads:[~2020-10-21  2:52 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-10-19 10:15 [PATCH] mm/memory.c: Introduce non-atomic __{Set,Clear}PageSwapCache Muchun Song
2020-10-19 12:31 ` Michal Hocko
2020-10-19 14:58   ` [External] " Muchun Song
2020-10-20 11:50     ` Xu, Yanfei
2020-10-20 13:08       ` Muchun Song
2020-10-21  2:52         ` Xu, Yanfei

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).