Hi, I am running openssl 0.9.7 on an MP-RAS Unix box with 32 bit Intel hardware. Very occassionally (four times in eight months) I have seen a segmentation fault in the big number code, when CRYTPO_malloc calls malloc. I understand that heap corruption can arise from just about anywhere, but it is interesting that the backtrace is so similar each time. The back trace from the dump is pretty garbled, and actual parameter values were not available, but it looks like this:
_real_malloc(?) malloc(num) default_malloc_ex(num,file,line) CRYPTO_malloc((int)sizeof(BN_ULONG)*(words+1), __FILE__,__LINE__) bn_expand_internal(b, words) bn_expand2(r, max+1) BN_uadd(X, X, n) BN_mod_inverse(&Ri,R,&tmod,ctx) BN_MONT_CTX_set((BN_MONT_CTX *)dh->method_mont_p, dh->p,ctx) dh->meth->generate_key(dh) DH_generate_key I know it's a long shot, but since I cannot recreate it, and since the dump is pretty much worthless, I am wondering if anyone has had a similar problem, and if so, is there a later version of openssl that fixes it? Thanks, Tom
