Re: [RFC 1/2] compiler: use compiler to detect integer overflows

[Sasha Levin <sasha.levin@oracle.com>]

On 11/26/2014 12:55 PM, Linus Torvalds wrote:
> On Nov 26, 2014 6:00 AM, "Sasha Levin" <sasha.levin@oracle.com <mailto:sasha.levin@oracle.com>> wrote:
>>
>> We've used to detect integer overflows by causing an overflow and testing the
>> result. For example, to test for addition overflow we would:
>>
>>         if (a + b < a)
>>                 /* Overflow detected */
>>
>> While it works, this is actually an undefined behaviour and we're not
>> guaranteed to have integers overflowing this way.
> 
> Bullshit.
> 
> Integer overflow is completely well defined in unsigned types.
> 
> Don't make up things like this.

Yes, I messed up and picked case where both types are unsigned in my example
patch. Apologies.

The kernel still has it's share of *signed* integer overflows. Example? fadvise64_64():

	loff_t offset, len;
	[...]
	loff_t endbyte;
	[...]
        /* Careful about overflows. Len == 0 means "as much as possible" */
        endbyte = offset + len;
        if (!len || endbyte < len)
                endbyte = -1;
        else
                endbyte--;              /* inclusive */


In essence, it's checking (offset + len < len), all of which are signed.


Thanks,
Sasha