Hi,
I spent most of the evening reading and programming on ssh/hmac.c. While the
stuff I tried to do didn't work, here is something I believe will make the
security better in any possible corefiles. We conceal the contents of the
secret hmac key from being dumped. Also an update on a comment on how to
compile the -DTEST of hmac.c:
Best Regards,
-peter
Index: hmac.c
===================================================================
RCS file: /cvs/src/usr.bin/ssh/hmac.c,v
retrieving revision 1.14
diff -u -p -u -r1.14 hmac.c
--- hmac.c 26 Feb 2020 13:40:09 -0000 1.14
+++ hmac.c 14 Aug 2023 19:50:49 -0000
@@ -51,7 +51,7 @@ ssh_hmac_start(int alg)
(ret->digest = ssh_digest_start(alg)) == NULL)
goto fail;
ret->buf_len = ssh_digest_blocksize(ret->ictx);
- if ((ret->buf = calloc(1, ret->buf_len)) == NULL)
+ if ((ret->buf = calloc_conceal(1, ret->buf_len)) == NULL)
goto fail;
return ret;
fail:
@@ -133,8 +133,12 @@ ssh_hmac_free(struct ssh_hmac_ctx *ctx)
}
#ifdef TEST
+/*
+ cc -DTEST digest-openssl.c hmac.c cleanup.c fatal.c log.c \
+ xmalloc.c sshbuf.c sshbuf-misc.c match.c misc.c \
+ ssherr.c addrmatch.c addr.c sshbuf-getput-basic.c -lcrypto
+ */
-/* cc -DTEST hmac.c digest.c buffer.c cleanup.c fatal.c log.c xmalloc.c
-lcrypto */
static void
hmac_test(void *key, size_t klen, void *m, size_t mlen, u_char *e, size_t elen)
{
--
Over thirty years experience on Unix-like Operating Systems starting with QNX.
