[2360] in Kerberos
Re: safe message problem: HP to 486
daemon@ATHENA.MIT.EDU (Shawn Mamros)
Mon Nov 9 12:16:36 1992
Date: 9 Nov 92 15:38:39 GMT
From: mamros@athena.lkg.dec.com (Shawn Mamros)
Reply-To: mamros@athena.lkg.dec.com (Shawn Mamros)
To: kerberos@shelby.Stanford.EDU
I noticed the same problem with safe messages between machines with different
byte-ordering a couple months ago (in my case, between a DECstation (little-
endian) and a Sun SPARCstation (big-endian). I changed the checksum
byte-swapping code in src/lib/krb/rd_safe.c to swap one longword (four bytes)
at a time, and that seemed to fix it.
I also sent my fix to kerberos-bugs@athena.mit.edu, but haven't received
any acknowledgement one way or another. I have no doubt that this has already
been noticed long before I did, and odds are good there was probably even
a patch posted to this mailing list a long time ago, but I've never seen
any place where "official" fixes to patchlevel 9 have been kept... For
what it's worth, here's my patchlevel 9-based fix:
*** rd_safe.c Mon Sep 14 10:25:32 1992
--- rd_safe.c.orig Mon Jan 23 15:16:38 1989
***************
*** 165,179 ****
*/
bcopy((char *)p,(char *)big_cksum,sizeof(big_cksum));
! if (swap_bytes) {
! u_long *swapptr;
! int i;
!
! for (i = 0; i < 4; i++) {
! swapptr = (u_long *)(big_cksum) + i;
! swap_u_long(*swapptr);
! }
! }
#ifdef NOENCRYPTION
bzero(calc_cksum, sizeof(calc_cksum));
--- 165,171 ----
*/
bcopy((char *)p,(char *)big_cksum,sizeof(big_cksum));
! if (swap_bytes) swap_u_16(big_cksum);
#ifdef NOENCRYPTION
bzero(calc_cksum, sizeof(calc_cksum));
Private messages don't seem to suffer from the same problem... actually, if
you look at rd_priv.c from patchlevel 9, you'll see that the checksum-checking
clause has been #ifdef'd out... sigh.
-Shawn Mamros
E-mail to: mamros@athena.lkg.dec.com
