On Tue, Aug 07, 2007 at 03:41:29AM +0900, KIYOHARA Takashi wrote:
> Hi! dyoung,
> 
> 
> From: David Young <dyoung@pobox.com>
> Date: Sun, 5 Aug 2007 15:07:27 -0500
> 
> > On Sat, Aug 04, 2007 at 03:25:13AM -0500, David Young wrote:
> > > On Sat, Aug 04, 2007 at 03:47:12PM +0900, KIYOHARA Takashi wrote:
> > > > Hi! dyoung,
> > > > 
> > > > 
> > > > The rt_tables was broken since 07/19/2007 20:49:00 perhaps.  X-<
> > > > 
> > > >   http://mail-index.netbsd.org/source-changes/2007/07/19/0032.html
> > > 
> > > The bug is not new since 19 July 2007.  The kernel has been sloppy
> > > about copying link-layer addresses for a long time.  fwip(4) suffers
> > > the buffer overflow because its link-layer addresses are the longest.
> > > I have attached a patch that shows an approach to a fix for the bugs.
> > > I leave it to you to make the patch compile and run. :-)
> > 
> > Here is a new patch, also untested.  Lengthening sockaddr_dl is still
> > left as an exercise for the reader. :-)
> 
> I tried your patch.  However repeated the message.  X-<
> 
>   sockaddr_dl_setaddr: no room for 16-byte media address
> 
> 
> I think that the return value of sockaddr_dup() that is called in
> rt_setgate() is a problem.  This is a value of pool(9).
> Can you resize it?
> 
> Do you keep it to a more longer size beforehand?

Try again with a longer sockaddr_dl.  For the size of sdl_data, I
recommend IFNAMSIZ+16 bytes

Dave
 
-- 
David Young             OJC Technologies
dyoung@ojctech.com      Urbana, IL * (217) 278-3933 ext 24