| |||||||||||||
Cryptographic weakness on Debian systemsCryptographic weakness on Debian systemsPosted May 14, 2008 22:50 UTC (Wed) by dvdeug (subscriber, #10998)In reply to: Cryptographic weakness on Debian systems by tialaramex Parent article: Cryptographic weakness on Debian systems
> I'd really want to see chapter and verse quoted before I believed that the same applies to
> { char m, x[40]; m = x[20]; }
On real hardware, without any compiler optimizations, both
{float m, x[40]; m = x[20];}
and
{char *m, (*x)[40]; m = x[20];}
can cause the program to crash, as the mere copy of an invalid float or invalid pointer can
generate fatal errors.
(Log in to post comments)
Cryptographic weakness on Debian systems Posted May 15, 2008 6:06 UTC (Thu) by lysse (guest, #3190) [Link] A float I can buy, if the FPU trips over an invalid bit pattern - but an invalid pointer? Educate me, please - cite an instance!
Cryptographic weakness on Debian systems Posted May 15, 2008 21:38 UTC (Thu) by Ross (subscriber, #4065) [Link] Not all memory spaces are flat, so some pointer values as stored in memory might not be loadable into the registers that implement them. (Due to segmentation, typed memory, etc.) Obviously that doesn't apply to most of the systems in use today (with the exception of function pointers).
Cryptographic weakness on Debian systems Posted May 19, 2008 4:38 UTC (Mon) by donwaugaman (subscriber, #4214) [Link] On a segmented architecture, load an invalid descriptor into a segment register. Boom! Not on all architectures, of course - I think the 386 family would be safe with this until you tried to access the segment - but some machines would generate an exception on the load.
|
|||||||||||||