[parisc-linux] nmap -O -> kernel panic on 712

[hgrothe@mathematik.tu-darmstadt.de]

--EVF5PPMfhYS0aIcm
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline

On Tue, Sep 18, 2001 at 07:38:55PM +0200, thunder7@xs4all.nl wrote:
> At this point, newbies like you and me can only hope one of the real
> kernel hackers sees this and says 'A-ha!'.
> 
> If I look at that code, I see a lot of (__u16 *)ptr and the like.
> 
> Am I correct in assuming those are all suspects and this is just another
> example of the missing unaligned access trap haunting us?

Far away from being a kernel hacker I tracked down the problem a little bit,
because I hate it if my favourite 'playing around' machine can be easily
crashed down by network. The following patch is not a solution in sense
of missing unaligned access trap(s) (I have much too few knowledge especially
of parisc assembler). It's a quick'n (really) dirty workaround which works
for me. The patch (for linux-2.4.9-pa24) breaks (__u32 *)ptr into two 
(__u16 *)ptr.

Comments are welcome.
   Holger
-- 
Holger Grothe  (Email: hgrothe@mathematik.tu-darmstadt.de)
Fachbereich Mathematik, TU Darmstadt

--EVF5PPMfhYS0aIcm
Content-Type: text/plain; charset=us-ascii
Content-Disposition: attachment; filename="linux-2.4.9-pa24.diff"

*** net/ipv4/tcp_input.c.dist	Fri Aug 17 12:04:25 2001
--- net/ipv4/tcp_input.c	Wed Sep 19 19:25:18 2001
***************
*** 2051,2058 ****
--- 2051,2063 ----
  						if ((estab && tp->tstamp_ok) ||
  						    (!estab && sysctl_tcp_timestamps)) {
  							tp->saw_tstamp = 1;
+ #if defined (__hppa__)							
+ 							tp->rcv_tsval = (((__u32)ntohs(*(__u16 *)ptr))<<16) | ((__u32)ntohs(*(__u16 *)(ptr+2)));
+ 							tp->rcv_tsecr = (((__u32)ntohs(*(__u16 *)(ptr+4)))<<16) | ((__u32)ntohs(*(__u16 *)(ptr+6)));
+ #else
  							tp->rcv_tsval = ntohl(*(__u32 *)ptr);
  							tp->rcv_tsecr = ntohl(*(__u32 *)(ptr+4));
+ #endif
  						}
  					}
  					break;

--EVF5PPMfhYS0aIcm--