Subject: AlphaServer 4000 not dropping to DDB with serial BREAK and not restarting from halt
To: NetBSD/alpha Discussion List <port-alpha@NetBSD.ORG>
From: Greg A. Woods <woods@weird.com>
List: port-alpha
Date: 07/30/2003 02:47:39
I'm having trouble getting my "new" AS4000 to recognize a BREAK on its
serial console.

It's running 1.6.1 installed from the official binary distribution, with
the default "hw.cnmagic" setting:

	$ /sbin/sysctl  hw.cnmagic | od -c 
	0000000    h   w   .   c   n   m   a   g   i   c       =       ' 001  \n
	0000020

The console is indeed on com0:

	com0 at isa0 port 0x3f8-0x3ff irq 4: ns16550a, working fifo
	com0: console

I'm running the GENERIC kernel from the binary distribution, which does
appear to have the debugger in it, and ddb.fromconsole is enabled:

	$ /sbin/sysctl ddb 
	ddb.radix = 16
	ddb.maxoff = 268435456
	ddb.maxwidth = 80
	ddb.lines = 24
	ddb.tabstops = 8
	ddb.onpanic = 1
	ddb.fromconsole = 1

However nothing happens when I send a BREAK (via conserver):

[Tue Jul 29 23:42:06 2003]^M
[Tue Jul 29 23:42:06 2003]NetBSD/alpha (building) (console)^M
[Tue Jul 29 23:42:06 2003]^M
[Tue Jul 29 23:42:07 2003]login: [-- break #1 sent -- `\z' -- Tue Jul 29 23:42:11 2003]
<C0>^Mlogin: 
[Tue Jul 29 23:42:12 2003]^M
[Tue Jul 29 23:42:12 2003]NetBSD/alpha (building) (console)^M
[Tue Jul 29 23:42:12 2003]^M
[Tue Jul 29 23:42:12 2003]login:


(yes, my conserver setup works fine for sending BREAK to my sparcstations)


I can jump into DDB by first entering the SRM via the RCM (by sending
"^]^]rcm" to the console port) and running a "halt" command in the RCM
(and presumably I could also get to the SRM by pressing the halt
button), and then I get to DDB by running a "continue" command from the
SRM.

However when I do that I can't seem to get the kernel restarted
successfully:

[Wed Jul 30 00:26:46 2003]RCM>halt
[Wed Jul 30 00:27:32 2003]
[Wed Jul 30 00:27:32 2003]Focus returned to COM port
[Wed Jul 30 00:27:32 2003]
[Wed Jul 30 00:27:32 2003]halted CPU 0
[Wed Jul 30 00:27:32 2003]
[Wed Jul 30 00:27:32 2003]halt code = 1
[Wed Jul 30 00:27:32 2003]operator initiated halt
[Wed Jul 30 00:27:32 2003]PC = fffffc0000300648
[Wed Jul 30 00:27:40 2003]P00>>>continue
[Wed Jul 30 00:27:44 2003]
[Wed Jul 30 00:27:44 2003]continuing CPU 0
[Wed Jul 30 00:27:44 2003]CP - RESTORE_TERM routine to be called
[Wed Jul 30 00:27:45 2003]panic: user requested console halt
[Wed Jul 30 00:27:45 2003]^MStopped at      cpu_Debugger+0x4:       ret     zero,(ra)

OK, so far, so good....

[Wed Jul 30 00:27:45 2003]^Mdb> cont
[Wed Jul 30 00:27:55 2003]^Msyncing disks... 

ooops!  it shouldn't have done that!  what's going on here?

[Wed Jul 30 00:27:55 2003]^MCPU 0: fatal kernel trap:
[Wed Jul 30 00:27:55 2003]
[Wed Jul 30 00:27:55 2003]^MCPU 0    trap entry = 0x2 (memory management fault)
[Wed Jul 30 00:27:55 2003]^MCPU 0    a0         = 0x1a4
[Wed Jul 30 00:27:55 2003]^MCPU 0    a1         = 0x1
[Wed Jul 30 00:27:55 2003]^MCPU 0    a2         = 0x0
[Wed Jul 30 00:27:55 2003]^MCPU 0    pc         = 0xfffffc000052763c
[Wed Jul 30 00:27:55 2003]^MCPU 0    ra         = 0xfffffc00005246a4
[Wed Jul 30 00:27:55 2003]^MCPU 0    pv         = 0xfffffc00005272c0
[Wed Jul 30 00:27:55 2003]^MCPU 0    curproc    = 0x0
[Wed Jul 30 00:27:55 2003]
[Wed Jul 30 00:27:55 2003]^Mpanic: trap
[Wed Jul 30 00:27:55 2003]^MStopped at      cpu_Debugger+0x4:       ret     zero,(ra)
[Wed Jul 30 00:27:55 2003]^Mdb> 

-- 
						Greg A. Woods

+1 416 218-0098                  VE3TCP            RoboHack <woods@robohack.ca>
Planix, Inc. <woods@planix.com>          Secrets of the Weird <woods@weird.com>