[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-mm
Subject:    Re: [PATCH v2] mm,writeback: fix divide by zero in pos_ratio_polynom
From:       Rik van Riel <riel () redhat ! com>
Date:       2014-04-30 14:52:08
Message-ID: 53610E18.8050809 () redhat ! com
[Download RAW message or body]

On 04/30/2014 10:49 AM, Michal Hocko wrote:
> On Wed 30-04-14 10:31:29, Rik van Riel wrote:
>> On 04/30/2014 09:48 AM, Michal Hocko wrote:
>>> On Wed 30-04-14 09:30:35, Rik van Riel wrote:
>>> [...]
>>>> Subject: mm,writeback: fix divide by zero in pos_ratio_polynom
>>>>
>>>> It is possible for "limit - setpoint + 1" to equal zero, leading to a
>>>> divide by zero error. Blindly adding 1 to "limit - setpoint" is not
>>>> working, so we need to actually test the divisor before calling div64.
>>>>
>>>> Signed-off-by: Rik van Riel <riel@redhat.com>
>>>> ---
>>>>   mm/page-writeback.c | 13 +++++++++++--
>>>>   1 file changed, 11 insertions(+), 2 deletions(-)
>>>>
>>>> diff --git a/mm/page-writeback.c b/mm/page-writeback.c
>>>> index ef41349..f98a297 100644
>>>> --- a/mm/page-writeback.c
>>>> +++ b/mm/page-writeback.c
>>>> @@ -597,11 +597,16 @@ static inline long long pos_ratio_polynom(unsigned long setpoint,
>>>>   					  unsigned long dirty,
>>>>   					  unsigned long limit)
>>>>   {
>>>> +	unsigned long divisor;
>>>>   	long long pos_ratio;
>>>>   	long x;
>>>>
>>>> +	divisor = limit - setpoint;
>>>> +	if (!divisor)
>>>> +		divisor = 1;	/* Avoid div-by-zero */
>>>> +
>>>
>>> This is still prone to u64 -> s32 issue, isn't it?
>>> What was the original problem anyway? Was it really setpoint > limit or
>>> rather the overflow?
>>
>> Thinking about it some more, is it possible that
>> limit and/or setpoint are larger than 32 bits, but
>> the difference between them is not?
>>
>> In that case, truncating both to 32 bits before
>> doing the subtraction would be troublesome, and
>> it would be better to do a cast in the comparison:
>>
>> if (!(s32)divisor)
>> 	divisor = 1;
>
> How is that any different than defining divisor as 32b directly?

For unsigned, it probably doesn't make a difference.

For signed int vs unsigned long, I wonder if there is a
corner case where casting the second to the first before
doing the "limit - setpoint" calculation can lead to a
different outcome...

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
[prev in list] [next in list] [prev in thread] [next in thread]