Re: CVE request: crypt_blowfish 8-bit character mishandling

[Solar Designer <solar () openwall com>]

On Tue, Jun 21, 2011 at 12:09:16PM -0600, Vincent Danen wrote:
> Ok, so taking a quick look at php-suhosin, we have:
>
> ...
>  61 typedef unsigned int BF_word;
> ...
> 558     BF_word tmp;
> 559 
> 560     for (i = 0; i < BF_N + 2; i++) {
> 561         tmp = 0;
> 562         for (j = 0; j < 4; j++) {
> 563             tmp <<= 8;
> 564             tmp |= *ptr;
>
> I'm assuming the above means it is vulnerable (unsigned int vs unsigned
> char).

No, we can't conclude anything from just the excerpt you quoted above.
If *ptr is signed char, then we have the bug.  If it's unsigned char,
then we don't.  If it's just char, which it was in my original code,
then we have the bug on most platforms, but not on those few where char
defaults to unsigned.  Or rather, the bug is mitigated on those.

Alexander