[PATCH] mm: avoid unnecessary page table walk for __get_user_pages

[PATCH] mm: avoid unnecessary page table walk for __get_user_pages

[Chen Wandun]

It is no need to walk page table and find pages if faultin_page success
and __get_user_pages does't care pages at all, so directly handle next
page.

Signed-off-by: Chen Wandun <chenwandun@huawei.com>
---
 mm/gup.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/mm/gup.c b/mm/gup.c
index 983e24fd4b70..c8de33cc37af 100644
--- a/mm/gup.c
+++ b/mm/gup.c
@@ -1198,7 +1198,10 @@ static long __get_user_pages(struct mm_struct *mm,
 					   PTR_ERR(page) == -EMLINK, locked);
 			switch (ret) {
 			case 0:
-				goto retry;
+				if (pages)
+					goto retry;
+				else
+					goto next_page;
 			case -EBUSY:
 			case -EAGAIN:
 				ret = 0;
-- 
2.25.1

Re: [PATCH] mm: avoid unnecessary page table walk for __get_user_pages

[Andrew Morton]

On Tue, 6 Sep 2022 10:44:01 +0800 Chen Wandun <chenwandun@huawei.com> wrote:

> It is no need to walk page table and find pages if faultin_page success
> and __get_user_pages does't care pages at all, so directly handle next
> page.
>
> ...
>
> --- a/mm/gup.c
> +++ b/mm/gup.c
> @@ -1198,7 +1198,10 @@ static long __get_user_pages(struct mm_struct *mm,
>  					   PTR_ERR(page) == -EMLINK, locked);
>  			switch (ret) {
>  			case 0:
> -				goto retry;
> +				if (pages)
> +					goto retry;
> +				else
> +					goto next_page;
>  			case -EBUSY:
>  			case -EAGAIN:
>  				ret = 0;

Looks right to me.  I redid the changelog somewhat:

: There is no need to walk page tables and find pages if faultin_page)_
: succeeds and __get_user_pages)_ doesn't care about populating the pages[]
: array.  So directly go on to handle the next page.

Re: [PATCH] mm: avoid unnecessary page table walk for __get_user_pages

[John Hubbard]

On 9/7/22 16:34, Andrew Morton wrote:
> On Tue, 6 Sep 2022 10:44:01 +0800 Chen Wandun <chenwandun@huawei.com> wrote:
> 
>> It is no need to walk page table and find pages if faultin_page success
>> and __get_user_pages does't care pages at all, so directly handle next
>> page.
>>
>> ...
>>
>> --- a/mm/gup.c
>> +++ b/mm/gup.c
>> @@ -1198,7 +1198,10 @@ static long __get_user_pages(struct mm_struct *mm,
>>   					   PTR_ERR(page) == -EMLINK, locked);
>>   			switch (ret) {
>>   			case 0:
>> -				goto retry;
>> +				if (pages)
>> +					goto retry;
>> +				else
>> +					goto next_page;

Wow, good point. And it's been doing that for a *long* time!

>>   			case -EBUSY:
>>   			case -EAGAIN:
>>   				ret = 0;
> 
> Looks right to me.  I redid the changelog somewhat:
> 
> : There is no need to walk page tables and find pages if faultin_page)_
> : succeeds and __get_user_pages)_ doesn't care about populating the pages[]
> : array.  So directly go on to handle the next page.
> 
> 

With that re-worded commit description, please feel to add

Reviewed-by: John Hubbard <jhubbard@nvidia.com>


thanks,
-- 
John Hubbard
NVIDIA