Subject: Re: APM and halt
To: Wolfgang Rupprecht <wolfgang@wsrcc.com>
From: Martin Husemann <martin@rumolt.teuto.de>
List: port-i386
Date: 04/02/1998 21:44:25
> I've noticed that "shutdown -h XXX" appears to power down my Intel
> Venus PPro board in such a way that the IDE disk's final writes don't
> seem to ever make it onto the platters.  The next power-up invokes the
> full dirty-shutdown type fsck.

I've seen several variants of this:

 - I have one machine running 1.3 (but it has happened with various 1.2x
   versions before), IDE, no APM. A "shutdown -r now" will sometimes reboot this
   box too fast - the final superblock write won't hit the disk. After the
   reboot it will run fsck and find an unexpected inconsistency: primary
   superblock and first copy differ. I need to run fsck manualy and tell it
   to use the first copy.

 - On another box with SCSI, no APM either, after doing "shutdown -h now" 
   and waiting for the "Press any key to reboot", then powering down, the
   system is running fsck on boot the next time - and doesn't find anything,
   mark the disks clean and continue to boot. This started happening after
   I upgraded the kernel to 1.3.1 (userland still untouched 1.3).


Martin