It's alive!

It's alive!

[forsyth]

i am fairly sure that a cpu server will automatically choose the
first boot method after a modest time out to allow you time to type another.
if the first method is il, that's what it will use.

It's alive!

[Stephen]

This is a multi-part message in MIME format.

--------------ABD322C31DFF4F5237C228A
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

The fix we used for boot was:
1. if only one boot method, build in kernel don't ask
   question. boot.c/rootserver()

Method*
rootserver(char *arg)
{
	char prompt[256];
	char reply[64];
	Method *mp;
	char *cp, *goodarg;
	int n, j;

	/* don't ask if this is a cpu server and only one method */
	if (cpuflag && method[1].name == 0) {
		print("Root is from %s\n", method[0].name);
		return method;
	}

...

2. if the cpu server is authing itself, then don't have annoying
   timeout.

--------------ABD322C31DFF4F5237C228A
Content-Type: text/plain; charset=us-ascii; name="diff"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline; filename="diff"

*** original/doauthenticate.c	Mon Jul 24 05:52:57 1995
--- doauthenticate.c	Wed Dec 20 10:30:04 1995
***************
*** 5,15 ****
  
  static char *pbmsg = "AS protocol botch";
  static char *ccmsg = "can't connect to AS";
  
--- 5,17 ----
  
  static char *pbmsg = "AS protocol botch";
  static char *ccmsg = "can't connect to AS";
+ static char *selfmsg = "Authorize self";
  
***************
*** 30,37 ****
  		if(mp->auth == 0)
  			fatal("no method for accessing auth server");
  		afd = (*mp->auth)();
! 		if(afd < 0)
! 			return ccmsg;
  	}
  	if(write(afd, trbuf, TICKREQLEN) < 0 || read(afd, &t, 1) != 1){
  		close(afd);
--- 32,43 ----
  		if(mp->auth == 0)
  			fatal("no method for accessing auth server");
  		afd = (*mp->auth)();
! 		if(afd < 0) {
! 			if (afd == -2)
! 				return selfmsg;
! 			else
! 				return ccmsg;
! 		}
  	}
  	if(write(afd, trbuf, TICKREQLEN) < 0 || read(afd, &t, 1) != 1){
  		close(afd);
***************
*** 82,88 ****
  			return;
  
  	/* didn't work, go for the security hole */
! 	fprint(2, "no authentication server (%s), using your key as server key\n", msg);
  }
  
  char*
--- 88,95 ----
  			return;
  
  	/* didn't work, go for the security hole */
! 	if (strcmp(msg, selfmsg) != 0)
! 		fprint(2, "no authentication server (%s), using your key as server key\n", msg);
  }
  
  char*
*** original/ip.c	Mon Jul 24 05:52:57 1995
--- ip.c	Fri Oct 18 23:34:26 1996
***************
*** 75,80 ****
--- 108,115 ----
  {
  	int fd[2]; 
  
+ 	if(memcmp(auip, ipaddr, sizeof(ipaddr)) == 0)
+ 		return -2;
  	if(auip[0] == 0 || ipdial(fd, "#Iil/il", auip, 566) < 0)
  		return -1;
  	close(fd[0]);

--------------ABD322C31DFF4F5237C228A--

It's alive!

[Brandon]

On Thu, 7 Nov 1996 forsyth@plan9.cs.york.ac.uk wrote:

> i am fairly sure that a cpu server will automatically choose the
> first boot method after a modest time out to allow you time to type another.
> if the first method is il, that's what it will use.
> 

I just wanted a clean feel.. no prompt.. no extra seconds of waiting..
I'm just picky... 

Brandon

It's alive!

[Brandon]

Thanks to everyone who has helped me... its alive!

To anyone else who basically knows what they're doing but is having a hard
time installing from the CD using the instructions in the book... here
were my three biggest problems:

1. When the auth server asks for the IP addres of the auth server, put in
0.1.0.0, _not_ 0.0.0.0

2. After you first boot up your auth server (the first time), and the
fileserver is still in allow mode (or you haven't rebooted it since installing
the CD, same thing)... do an "auth/changeuser authid" where authid is your
domain's authid.

3. When you're editing /lib/ndb/local, go down to the services and add:
il=ticket	port=566


Now I'm off to call AssureNet Pathways and get a securenet box.... before
I go, one other hint.... 

Re-typing the IP info on your auth server every time you boot is a bitch...
edit /sys/src/9/boot/ip.c, go down to the bootp section, wipe out the bootp
code, down to where is starts asking for the info manually after bootp failed.
remove the outin() calls and other such stuff, and hardcode the ip address
info for the auth server in.  The go to /sys/src/9/pc, and "mk CONF=pccpu".
(be sure to put the original ip.c back when you're done, and mk clean on
both directories...)
To save a few extra seconds and optimize the kernel a bit, I also took the
datakit stuff out of /sys/src/9/pc/pccpu (the dk lines in the devices list
and boot list, and the sturp line, I think it was), so that IL was the only
boot method, and the datakit code was not present, and modified
/sys/src/9/boot/boot.c to automatically pick the first boot method on the list.
To do that, pretty much wipe out the rootserver() function's internals, and
fill it in with:

mp=method;
return mp;

or something.. I don't have the sources here(my plan 9 setup is a half hour
drive from where I can get email... so... )

Brandon

..............................................
: Brandon Lee Black  : photon@gnu.ai.mit.edu :
:....................: photon@nol.net        :
: "Sanity is the     : photon@vanity.ops.org :
: trademark of a     : vis_blb@unx1.shsu.edu :
: weak mind. . ."    : +1 713 397 3490       :
:....................:.......................: