Subject: Re: unusual 1.4.1 behaviour with 'halt' and 'shutdown' commands on
To: Craig Dewick <cdewick@lios.apana.org.au>
From: David Brownlee <abs@netbsd.org>
List: port-sparc
Date: 06/21/2000 08:15:05
On 21 Jun 2000, Craig Dewick wrote:

> 
> Hi everyone,
> 
> My FTP server is now a Sparc 1+ after the previous box (Sparc Classic) died.
> 
> I've spent a day replacing the SCSI drive with a bigger one (and copied all
> the filesystems over to the new drive, plus installed a bootblock), and I
> noticed some unusual behaviour...
> 
> If I log in as root when the machine is running in multi-user mode, and I
> type 'halt' to shut the system down, the kernel hangs and I have to manually
> abort to get back to the ROM monitor prompt.
> 
> If I log in as root when the machin is running in multi-user mode, and I
> type 'shutdown now' to get into single-user mode, then type 'halt' from the
> single-user prompt, the kernel aborts properly and the machine drops back to
> the ROM monitor as expected.
> 
> If I boot the kernel in single-user mode, 'halt' works as above.
> 
> So, what is going on to prevent 'halt' (as root) in multi-user mode from
> aborting the kernel properly?
> 
> The machine has a *really* old v1.3 bootROM btw... Not sure if that is part
> of the problem or not. The machine is running the generic 1.4.1 kernel from
> the binary release.
> 
> Is there some problem with processes refusing to die in multi-user mode
> which is causing the kernel to get stuck in a loop waiting for all the
> processes to exit?
> 
	Some random ideas:
	    You could build a kernel with DDB, then when it hangs STOP+A
	    will drop into DDB and you can type ps to find out what is
	    running.

	    Try a 1.4.2 kernel

	    Are you running an mfs tmp or similar?

		David/absolute
				       -- www.netbsd.org: No hype required --