Debian messed OpenSSL (again)

[Vincenzo Ciaschini <vincenzo.ciaschini@cnaf.infn.it>]

Hi everyone,

   The version of OpenSSL distributed with Debian 6 applies this patch:

Index: openssl-0.9.8k/crypto/rand/md_rand.c
===================================================================
--- openssl-0.9.8k.orig/crypto/rand/md_rand.c   2008-09-16 
13:50:05.000000000 +0200
+++ openssl-0.9.8k/crypto/rand/md_rand.c        2009-07-19 
11:36:05.000000000 +0200
@@ -477,8 +477,10 @@
                MD_Update(&m,local_md,MD_DIGEST_LENGTH);
                MD_Update(&m,(unsigned char *)&(md_c[0]),sizeof(md_c));
 #ifndef PURIFY
+#if 0 /* Don't add uninitialised data. */
                MD_Update(&m,buf,j); /* purify complains */
 #endif
+#endif
                k=(st_idx+MD_DIGEST_LENGTH/2)-st_num;
                if (k > 0)
                        {
(valgrind.patch)

This essentially cripples the ssleay_rand_bytes() and 
ssleay_rand_pseudo_bytes() functions, which in turn are used extensively 
throughout the code for salt generation, initialization vectors, writing 
private keys, etc... under the names of RAND_bytes() and 
RAND_pseudo_bytes().

Please note that this is essentially the second half of the same patch 
that was already retracted for crippling OpenSSL a few years ago, and 
while this is less damaging, since at least RAND_add() still works, 
there are still function where only RAND_bytes() or RAND_pseudo_bytes() 
are used, like for example EVP_CIPHER_CTX_rand_key().  And in those who 
also use RAND_add() it still weakens the algorithm by discarding one 
source of random data.

I believe that this patch should be reverted.

Ciao,
   Vincenzo