* pppd segfaults on AMD64 with ms-chap
@ 2004-06-08 7:16 mole
2004-06-08 7:20 ` James Cameron
` (2 more replies)
0 siblings, 3 replies; 4+ messages in thread
From: mole @ 2004-06-08 7:16 UTC (permalink / raw)
To: linux-ppp
[-- Attachment #1: Type: text/plain, Size: 688 bytes --]
I am new to this list, sorry if this problem is already solved.
I have tried to use pptp on fedora Core 2 for AMD64 and found that pppd
segfaults on my machine when ms-chap is in use.
If pppd is compiled with openssl's sha then pppd doesn't segfault on
authentication but mppe-enabled kernel then gives me oops in mppe sha code.
The same kernel/pppd work fine in 32 bit mode.
The problem proved to be in the broken sha1 implementation that assumes
that unsigned long is 32-bit wide.
The quick/minimal change to make it all work in 64-bit mode is in the
attached patch. But it looks like the code needs more cleanups to make
it obviously 64-bit safe.
Best,
Oleg Makarenko
[-- Attachment #2: pppd-x86_64.patch --]
[-- Type: text/plain, Size: 1463 bytes --]
diff -urN ppp-2.4.2_cvs_20030610.orig/linux/mppe/sha1.c ppp-2.4.2_cvs_20030610/linux/mppe/sha1.c
--- ppp-2.4.2_cvs_20030610.orig/linux/mppe/sha1.c 2002-04-02 18:01:37.000000000 +0400
+++ ppp-2.4.2_cvs_20030610/linux/mppe/sha1.c 2004-06-08 14:55:46.000000000 +0400
@@ -19,6 +19,7 @@
#if defined(__linux__)
#include <asm/byteorder.h>
#include <linux/string.h>
+#include <linux/types.h>
#else if defined(__solaris__)
#include <sys/isa_defs.h>
#include <sys/ddi.h>
@@ -59,10 +60,10 @@
static void
SHA1_Transform(unsigned long state[5], const unsigned char buffer[64])
{
- unsigned long a, b, c, d, e;
+ u32 a, b, c, d, e;
typedef union {
unsigned char c[64];
- unsigned long l[16];
+ u32 l[16];
} CHAR64LONG16;
CHAR64LONG16 *block;
diff -urN ppp-2.4.2_cvs_20030610.orig/pppd/sha1.c ppp-2.4.2_cvs_20030610/pppd/sha1.c
--- ppp-2.4.2_cvs_20030610.orig/pppd/sha1.c 2002-04-02 17:54:59.000000000 +0400
+++ ppp-2.4.2_cvs_20030610/pppd/sha1.c 2004-06-08 14:54:44.000000000 +0400
@@ -18,6 +18,7 @@
#include <string.h>
#include <netinet/in.h> /* htonl() */
+#include <sys/types.h> /* u_int32_t */
#include "sha1.h"
static void
@@ -44,10 +45,10 @@
static void
SHA1_Transform(unsigned long state[5], const unsigned char buffer[64])
{
- unsigned long a, b, c, d, e;
+ u_int32_t a, b, c, d, e;
typedef union {
unsigned char c[64];
- unsigned long l[16];
+ u_int32_t l[16];
} CHAR64LONG16;
CHAR64LONG16 *block;
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: pppd segfaults on AMD64 with ms-chap
2004-06-08 7:16 pppd segfaults on AMD64 with ms-chap mole
@ 2004-06-08 7:20 ` James Cameron
2004-06-09 5:38 ` Guy Rouillier
2004-06-11 7:59 ` mole
2 siblings, 0 replies; 4+ messages in thread
From: James Cameron @ 2004-06-08 7:20 UTC (permalink / raw)
To: linux-ppp
Matt Domsch posted a patch to one of the PPTP mailing lists today that
causes the MPPE module to use the in-kernel implementations of SHA1 and
ARC4. You might find that the existing kernel code is 64-bit clean.
I'm sure we'd love to hear how it goes.
--
James Cameron http://quozl.netrek.org/
HP Open Source, Volunteer http://opensource.hp.com/
PPTP Client Project, Release Engineer http://pptpclient.sourceforge.net/
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: pppd segfaults on AMD64 with ms-chap
2004-06-08 7:16 pppd segfaults on AMD64 with ms-chap mole
2004-06-08 7:20 ` James Cameron
@ 2004-06-09 5:38 ` Guy Rouillier
2004-06-11 7:59 ` mole
2 siblings, 0 replies; 4+ messages in thread
From: Guy Rouillier @ 2004-06-09 5:38 UTC (permalink / raw)
To: linux-ppp
On Tue, 08 Jun 2004 15:15:13 +0400
mole <mole@quadra.ru> wrote:
> I am new to this list, sorry if this problem is already solved.
>
> I have tried to use pptp on fedora Core 2 for AMD64 and found that
> pppd segfaults on my machine when ms-chap is in use.
I just reported this same bug to the list two days ago. See "2.4.2 on
AMD64 Linux: assumes 4-byte long". I haven't gotten any replies yet, so
I'm going to try making the changes I found at the link I provided.
Unfortunately, the version of sha1.c in the kernel has the same bug in
it, so I'm going to have to rebuild the kernel. I tried just rebuilding
pppd, and that resulted in a kernel panic. More changes are required
than your patch includes; for example, the SHA1_CTX structure itself
must be modified.
>
> If pppd is compiled with openssl's sha then pppd doesn't segfault on
> authentication but mppe-enabled kernel then gives me oops in mppe sha
> code.
>
> The same kernel/pppd work fine in 32 bit mode.
>
> The problem proved to be in the broken sha1 implementation that
> assumes that unsigned long is 32-bit wide.
>
> The quick/minimal change to make it all work in 64-bit mode is in the
> attached patch. But it looks like the code needs more cleanups to make
>
> it obviously 64-bit safe.
>
> Best,
> Oleg Makarenko
>
>
>
>
>
>
>
--
Guy Rouillier
--
Guy Rouillier
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: pppd segfaults on AMD64 with ms-chap
2004-06-08 7:16 pppd segfaults on AMD64 with ms-chap mole
2004-06-08 7:20 ` James Cameron
2004-06-09 5:38 ` Guy Rouillier
@ 2004-06-11 7:59 ` mole
2 siblings, 0 replies; 4+ messages in thread
From: mole @ 2004-06-11 7:59 UTC (permalink / raw)
To: linux-ppp
James Cameron wrote:
>Matt Domsch posted a patch to one of the PPTP mailing lists today that
>causes the MPPE module to use the in-kernel implementations of SHA1 and
>ARC4. You might find that the existing kernel code is 64-bit clean.
>
>
>
I have tried it (original patch) without any success. I get kernel oops
on both 32 and 64 bit kernels.
>I'm sure we'd love to hear how it goes.
>
>
Going to try his second patch version and let you know of the results.
Oleg Makarenko
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2004-06-11 7:59 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2004-06-08 7:16 pppd segfaults on AMD64 with ms-chap mole
2004-06-08 7:20 ` James Cameron
2004-06-09 5:38 ` Guy Rouillier
2004-06-11 7:59 ` mole
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).