On Sat, 27 Feb 2016, Borislav Petkov wrote: > * sum is computed here: > > sum = orig_sum > - (mc_header->sig + mc_header->pf + mc_header->cksum) > + (ext_sig->sig + ext_sig->pf + ext_sig->cksum);
Well, orig_sum is always zero, because we exit early when it isn't. sum really ought to be u32 in the first place (the only reason the code works is because gcc isn't that insane), and the other fields should have been cast to u32 as well before being added, for clarity if nothing else. > and it adds a bunch of integers which can overflow, sure, but if it > overflows, we exit early. There is no reason why "sig + pf + cksum" wouldn't overflow 32 bits... in fact, they often (always?) do on valid microcode. And we need all of the intel-microcode-checksum-related arithmetic to be done modulo-2^32, because it is expected to overflow *and* warp around when it does so. -- "One disk to rule them all, One disk to find them. One disk to bring them all and in the darkness grind them. In the Land of Redmond where the shadows lie." -- The Silicon Valley Tarot Henrique Holschuh
