On Mon, 2007-04-30 at 14:08 +0200, Stefan Wenk wrote:

On Sunday 29 April 2007 23:20, Andrew Morton wrote:

quoted

(switch to email - please retain all ccs)

On Sun, 29 Apr 2007 13:01:25 -0700 bugme-daemon@bugzilla.kernel.org wrote:

quoted

http://bugzilla.kernel.org/show_bug.cgi?id=8405

Here is the log vor /var/log/messages:

Apr 29 16:53:34 linux kernel: PPP Deflate Compression module registered
Apr 29 16:53:47 linux pppd[9090]: pppd 2.4.4 started by root, uid 0
Apr 29 16:53:47 linux pppd[9090]: Using interface ppp0
Apr 29 16:53:47 linux pppd[9090]: Connect: ppp0 <--> /dev/pts/9
Apr 29 16:53:47 linux pppd[9090]: kernel does not support PPP filtering
Apr 29 16:53:47 linux pppd[9090]: Deflate (15) compression enabled
Apr 29 16:53:47 linux pppd[9090]: Cannot determine ethernet address for
proxy ARP Apr 29 16:53:47 linux pppd[9090]: local  IP address 192.168.3.2
Apr 29 16:53:47 linux pppd[9090]: remote IP address 192.168.3.1
Apr 29 16:54:03 linux kernel: z_decompress0: inflate returned -5 ()
Apr 29 16:54:03 linux pppd[9090]: PPPIOCGFLAGS flags: c030c0
Apr 29 16:54:03 linux pppd[9090]: Lost compression sync: disabling
compression

I had a look at the kernel patch of 2.6.18 - the lib/zlib/zlib_inflate
was changed a lot and for me it looks to be the reason of the problem.

If you're referring to 4f3865fb57a04db7cca068fed1c15badc064a302,
"zlib_inflate: Upgrade library code to a recent version" then that was
mainly a zlib_inflate change, not a zlib_deflate change.

But yes, I'd say that this patch might well have been the cause.

http://userweb.kernel.org/~akpm/zlib-backout.patch is a backout patch
against 2.6.21.  If you apply that to 2.6.21 does it make ppp work again?

thanks for this backout patch. I have applied it to 2.6.21 and now pppd is 
working with deflate again.

The key line of that log is 
"Apr 29 16:54:03 linux kernel: z_decompress0: inflate returned -5 ()"

linux/zlib.h says:

#define Z_BUF_ERROR    (-5)

so somehow we're triggering a buffer error...

Looking at zlib_inflate() in lib/zlib_inflate/inflate.c, it seems that
error only occurs if we had success (Z_OK) but no remaining input/output
space. I suspect you can just have ppp_deflate ignore that "error" code
and it might just work.

Can you try the following patch:

Index: linux-2.6.20/drivers/net/ppp_deflate.c
===================================================================

--- linux-2.6.20.orig/drivers/net/ppp_deflate.c	2007-02-04 18:44:54.000000000 +0000
+++ linux-2.6.20/drivers/net/ppp_deflate.c	2007-04-30 14:27:08.000000000 +0100

@@ -487,12 +487,16 @@ int z_decompress(void *arg, unsigned cha
 	 */
 	for (;;) {
 		r = zlib_inflate(&state->strm, Z_PACKET_FLUSH);
-		if (r != Z_OK) {
+		if ((r != Z_OK) && (r != Z_BUF_ERROR)) {
 			if (state->debug)
 				printk(KERN_DEBUG "z_decompress%d: inflate returned %d (%s)\n",
 				       state->unit, r, (state->strm.msg? state->strm.msg: ""));
 			return DECOMP_FATALERROR;
 		}
+		if (r == Z_BUF_ERROR) {
+			printk(KERN_DEBUG "z_decompress%d: Would have triggered an error as inflate returned %d (%s)\n",
+				       state->unit, r, (state->strm.msg? state->strm.msg: ""));
+		}
 		if (state->strm.avail_out != 0)
 			break;		/* all done */
 		if (decode_proto) {

This should continue but print some debug output when the problem
occurs.

Richard