[Devel] [ openser-Bugs-1614192 ] Socket str description is malformed on big-endian platforms

[SourceForge.net]

Bugs item #1614192, was opened at 2006-12-12 20:43
Message generated for change (Comment added) made by bogdan_iancu
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=743020&aid=1614192&group_id=139143

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: core
Group: ver devel
Status: Open
Resolution: Invalid
Priority: 5
Private: No
Submitted By: Ovidiu Sas (osas)
Assigned to: Bogdan (bogdan_iancu)
Summary: Socket str description is malformed on big-endian platforms

Initial Comment:
The way that socket string description is constructed is not compatible with big endian platforms.
Here's a patch that will fix the problem:

# cvs diff -Nu socket_info.h
Index: socket_info.h
===================================================================
RCS file: /cvsroot/openser/sip-server/socket_info.h,v
retrieving revision 1.4
diff -u -r1.4 socket_info.h
--- socket_info.h	13 Dec 2005 13:38:30 -0000	1.4
+++ socket_info.h	12 Dec 2006 18:42:12 -0000
@@ -241,11 +241,6 @@
 
 static inline char* socket2str(struct socket_info *sock, char *s, int* len)
 {
-#define PROTOC2UINT(a, b, c, d) htonl((	(((unsigned int)(a))<<24)+ \
-								(((unsigned int)(b))<<16)+  \
-								(((unsigned int)(c))<<8)+  \
-								((unsigned int)(d)) ) | 0x20202020)
-
 	static char buf[MAX_SOCKET_STR];
 	char *p,*p1;
 
@@ -262,26 +257,31 @@
 
 	switch (sock->proto) {
 		case PROTO_UDP:
-			*((unsigned int*)p) = PROTOC2UINT('u', 'd', 'p', ':');
-			p += 4;
+			*(p++) = 'u';
+			*(p++) = 'd';
+			*(p++) = 'p';
 			break;
 		case PROTO_TCP:
-			*((unsigned int*)p) = PROTOC2UINT('t', 'c', 'p', ':');
-			p += 4;
+			*(p++) = 't';
+			*(p++) = 'c';
+			*(p++) = 'p';
 			break;
 		case PROTO_TLS:
-			*((unsigned int*)p) = PROTOC2UINT('t', 'l', 's', ':');
-			p += 4;
+			*(p++) = 't';
+			*(p++) = 'l';
+			*(p++) = 's';
 			break;
 		case PROTO_SCTP:
-			*((unsigned int*)p) = PROTOC2UINT('s', 'c', 't', 'p');
-			p += 4;
-			*(p++) = ':';
+			*(p++) = 's';
+			*(p++) = 'c';
+			*(p++) = 't';
+			*(p++) = 'p';
 			break;
 		default:
 			LOG(L_CRIT,"BUG:socket2str: unsupported proto %d\n", sock->proto);
 			return 0;
 	}
+	*(p++) = ':';
 	memcpy( p, sock->address_str.s, sock->address_str.len);
 	p += sock->address_str.len;
 	*(p++) = ':';

----------------------------------------------------------------------

>Comment By: Bogdan (bogdan_iancu)
Date: 2006-12-14 17:05

Message:
Logged In: YES 
user_id=1275325
Originator: NO

Hi Ovidiu,

I agree that the patch is correct, but actually I fail to see the bug :(.
During some tests on sparc64, nothing wrong came up. Could you please run
on the arm the attached test program and send me the output?

thanks and regards,
Bogdan
File Added: test.c

----------------------------------------------------------------------

Comment By: Ovidiu Sas (osas)
Date: 2006-12-13 16:42

Message:
Logged In: YES 
user_id=1395524
Originator: YES

Hi Bogdan,

I was compiling on arm big endian and It didn't come up properly for me. 
I noticed when I was trying to use dbtext for usrloc (this is how I noticed
and that's why I submitted the other patch for the output of the fifo
usrloc show command).  With this patch, I was able to use the dbtext.

Regards,
Ovidiu Sas

----------------------------------------------------------------------

Comment By: Bogdan (bogdan_iancu)
Date: 2006-12-13 16:12

Message:
Logged In: YES 
user_id=1275325
Originator: NO

Ovidiu,

I am not so sure if it a bug. First of all the htonl function is used,
which gives you arch. independency. Secondly, I just tried on a sparc
machine (big endian) and the string was properly generated.

regards,
bogdan

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=743020&aid=1614192&group_id=139143