Re: [RFC PATCH v3] mm: use READ_ONCE in page_cpupid_xchg_last()

From	Vlastimil Babka <vbabka@suse.cz>
Newsgroups	linux.kernel
Subject	Re: [RFC PATCH v3] mm: use READ_ONCE in page_cpupid_xchg_last()
Date	2016-12-07 10:40 +0100
Message-ID	<sLGRI-6H9-19@gated-at.bofh.it> (permalink)
References	<sKWOR-2aU-5@gated-at.bofh.it> <sLdd0-3RU-9@gated-at.bofh.it> <sLG5j-6bA-3@gated-at.bofh.it> <sLG5k-6bA-11@gated-at.bofh.it> <sLGf0-6eS-19@gated-at.bofh.it>
Organization	linux.* mail to news gateway

Show all headers | View raw

On 12/07/2016 09:58 AM, Michal Hocko wrote:
> On Wed 07-12-16 09:48:52, Vlastimil Babka wrote:
>> On 12/07/2016 09:43 AM, Michal Hocko wrote:
>>> On Tue 06-12-16 09:53:14, Xishi Qiu wrote:
>>>> A compiler could re-read "old_flags" from the memory location after reading
>>>> and calculation "flags" and passes a newer value into the cmpxchg making 
>>>> the comparison succeed while it should actually fail.
>>>>
>>>> Signed-off-by: Xishi Qiu <qiuxishi@huawei.com>
>>>> Suggested-by: Christian Borntraeger <borntraeger@de.ibm.com>
>>>> ---
>>>>  mm/mmzone.c | 2 +-
>>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>>
>>>> diff --git a/mm/mmzone.c b/mm/mmzone.c
>>>> index 5652be8..e0b698e 100644
>>>> --- a/mm/mmzone.c
>>>> +++ b/mm/mmzone.c
>>>> @@ -102,7 +102,7 @@ int page_cpupid_xchg_last(struct page *page, int cpupid)
>>>>  	int last_cpupid;
>>>>  
>>>>  	do {
>>>> -		old_flags = flags = page->flags;
>>>> +		old_flags = flags = READ_ONCE(page->flags);
>>>>  		last_cpupid = page_cpupid_last(page);
>>>
>>> what prevents compiler from doing?
>>> 		old_flags = READ_ONCE(page->flags);
>>> 		flags = READ_ONCE(page->flags);
>>
>> AFAIK, READ_ONCE tells the compiler that page->flags is volatile. It
>> can't read from volatile location more times than being told?
> 
> But those are two different variables which we assign to so what
> prevents the compiler from applying READ_ONCE on each of them
> separately?

I would naively expect that it's assigned to flags first, and then from
flags to old_flags. But I don't know exactly the C standard evaluation
rules that apply here.

> Anyway, this could be addressed easily by

Yes, that way there should be no doubt.

> diff --git a/mm/mmzone.c b/mm/mmzone.c
> index 5652be858e5e..b4e093dd24c1 100644
> --- a/mm/mmzone.c
> +++ b/mm/mmzone.c
> @@ -102,10 +102,10 @@ int page_cpupid_xchg_last(struct page *page, int cpupid)
>  	int last_cpupid;
>  
>  	do {
> -		old_flags = flags = page->flags;
> +		old_flags = READ_ONCE(page->flags);
>  		last_cpupid = page_cpupid_last(page);
>  
> -		flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
> +		flags = old_flags & ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
>  		flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
>  	} while (unlikely(cmpxchg(&page->flags, old_flags, flags) != old_flags));
>  
> 
>>> Or this doesn't matter?
>>
>> I think it would matter.
>>
>>>>  
>>>>  		flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
>>>> -- 
>>>> 1.8.3.1
>>>>
>>>
>

Thread

[RFC PATCH] mm: use ACCESS_ONCE in page_cpupid_xchg_last() Xishi Qiu <qiuxishi@huawei.com> - 2016-12-05 09:30 +0100
  Re: [RFC PATCH] mm: use ACCESS_ONCE in page_cpupid_xchg_last() Christian Borntraeger <borntraeger@de.ibm.com> - 2016-12-05 09:40 +0100
    Re: [RFC PATCH] mm: use ACCESS_ONCE in page_cpupid_xchg_last() Christian Borntraeger <borntraeger@de.ibm.com> - 2016-12-05 10:00 +0100
      Re: [RFC PATCH] mm: use ACCESS_ONCE in page_cpupid_xchg_last() Xishi Qiu <qiuxishi@huawei.com> - 2016-12-05 10:30 +0100
      [RFC PATCH v2] mm: use ACCESS_ONCE in page_cpupid_xchg_last() Xishi Qiu <qiuxishi@huawei.com> - 2016-12-05 10:30 +0100
        Re: [RFC PATCH v2] mm: use ACCESS_ONCE in page_cpupid_xchg_last() Christian Borntraeger <borntraeger@de.ibm.com> - 2016-12-05 10:50 +0100
  [RFC PATCH v3] mm: use READ_ONCE in page_cpupid_xchg_last() Xishi Qiu <qiuxishi@huawei.com> - 2016-12-06 03:00 +0100
    Re: [RFC PATCH v3] mm: use READ_ONCE in page_cpupid_xchg_last() Vlastimil Babka <vbabka@suse.cz> - 2016-12-07 09:40 +0100
    Re: [RFC PATCH v3] mm: use READ_ONCE in page_cpupid_xchg_last() Michal Hocko <mhocko@kernel.org> - 2016-12-07 09:50 +0100
      Re: [RFC PATCH v3] mm: use READ_ONCE in page_cpupid_xchg_last() Vlastimil Babka <vbabka@suse.cz> - 2016-12-07 09:50 +0100
        Re: [RFC PATCH v3] mm: use READ_ONCE in page_cpupid_xchg_last() Michal Hocko <mhocko@kernel.org> - 2016-12-07 10:00 +0100
          Re: [RFC PATCH v3] mm: use READ_ONCE in page_cpupid_xchg_last() Vlastimil Babka <vbabka@suse.cz> - 2016-12-07 10:40 +0100
            Re: [RFC PATCH v3] mm: use READ_ONCE in page_cpupid_xchg_last() Christian Borntraeger <borntraeger@de.ibm.com> - 2016-12-07 10:50 +0100
              Re: [RFC PATCH v3] mm: use READ_ONCE in page_cpupid_xchg_last() Christian Borntraeger <borntraeger@de.ibm.com> - 2016-12-07 11:10 +0100
                Re: [RFC PATCH v3] mm: use READ_ONCE in page_cpupid_xchg_last() Rasmus Villemoes <linux@rasmusvillemoes.dk> - 2016-12-07 23:20 +0100
              Re: [RFC PATCH v3] mm: use READ_ONCE in page_cpupid_xchg_last() Michal Hocko <mhocko@kernel.org> - 2016-12-07 11:10 +0100

csiph-web