linux-kernel - Re: [PATCH 2/2] lib: kstrtox: _parse_integer: use hex_to_bin instead local conversion, and reduce branches

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <79ee7e49-3780-5e00-528e-9814f467501a@163.com>
Date:	Thu, 30 Jun 2016 22:45:03 +0800
From:	Zhaoxiu Zeng <zengzhaoxiu@....com>
To:	Alexey Dobriyan <adobriyan@...il.com>
Cc:	linux-kernel@...r.kernel.org,
	Zhaoxiu Zeng <zhaoxiu.zeng@...il.com>,
	Andrew Morton <akpm@...ux-foundation.org>,
	Kees Cook <keescook@...omium.org>
Subject: Re: [PATCH 2/2] lib: kstrtox: _parse_integer: use hex_to_bin instead
 local conversion, and reduce branches

On 2016/6/30 6:06, Alexey Dobriyan wrote:
> On Thu, Jun 30, 2016 at 12:22:13AM +0800, zengzhaoxiu@....com wrote:
>> --- a/lib/kstrtox.c
>> +++ b/lib/kstrtox.c
>> @@ -48,38 +48,26 @@ unsigned int _parse_integer(const char *s, unsigned int base, unsigned long long
>>  {
>>  	unsigned long long res;
>>  	unsigned int rv;
>> -	int overflow;
>> +	unsigned int overflow;
>> +	unsigned int val;
>>  
>>  	res = 0;
>>  	rv = 0;
>>  	overflow = 0;
>> -	while (*s) {
>> -		unsigned int val;
>> -
>> -		if ('0' <= *s && *s <= '9')
>> -			val = *s - '0';
>> -		else if ('a' <= _tolower(*s) && _tolower(*s) <= 'f')
>> -			val = _tolower(*s) - 'a' + 10;
>> -		else
>> -			break;
>> -
>> -		if (val >= base)
>> -			break;
>> +	while ((val = hex_to_bin(*s++)) < base) {
> I hate this function. And it has a branch if your table patch doesn't
> go it. And it is beartrap (unsigned int = -1 < base).

How about this?

for (;;) {
    unsigned int val = hex_to_bin(*s++);
    if (val >= base)
        break;

> ACK *s++ bit, though. Should make code smaller in my experience.
> Please, change to "unsigned char c; while ((c = *s++)".
> This is about maximum code compression I can understand.

The previous tests are useless until reach the end of s.
The '\0' will be caught by "if (val >= base)" too.

>>  		/*
>>  		 * Check for overflow only if we are within range of
>>  		 * it in the max base we support (16)
>>  		 */
>>  		if (unlikely(res & (~0ull << 60))) {
>>  			if (res > div_u64(ULLONG_MAX - val, base))
>> -				overflow = 1;
>> +				overflow = KSTRTOX_OVERFLOW;
> Just do |= KSTRTOX_OVERFLOW here directly, it is the leftmost bit.

I thought so too at first， but finally I decided to reserve the varaible "overflow",
because this hack depend on the definition of KSTRTOX_OVERFLOW.