Subject: Re: kcore on sun3x (Re: CVS commit: syssrc)
To: Izumi Tsutsui <tsutsui@ceres.dti.ne.jp>
From: Chuck Silvers <chuq@chuq.com>
List: port-sun3
Date: 05/29/2001 08:49:48
actually I tried that when the machine was up multi-user and the device
it should have dumped to was enabled as a swap device.  I just tested it
and sdopen() is called in the swap_on() path.

-Chuck


On Tue, May 29, 2001 at 11:13:18PM +0900, Izumi Tsutsui wrote:
> In <20010528180421.A23935@spathi.chuq.com>
> chuq@chuq.com wrote:
> 
> > I didn't try actually taking a dump, I just wanted to make it boot.
> > I tried "reboot -d" just now and it crashed in sddump().
> > after fixing the bug with uninitialized variables in sddump(),
> > now it fails to dump because this check fails:
> > 
> > 	/* Make sure it was initialized. */
> > 	if ((periph->periph_flags & PERIPH_MEDIA_LOADED) != 0)
> > 		return (ENXIO);
> 
> > anyone have any ideas why this check would fail for a normal scsi disk?
> 
> I guess you tried on the first single user prompt.
> I saw the similar problem before import of thorpej-scsipi.
> PERIPH_MEDIA_LOADED (which was SDEV_MEDIA_LOADED) is
> set in sdopen(), so we have to open any sd device files
> at least once before getting dumps.
> 
> But I notice my news68k (which has ncr5380) hangs on dumps:
> ---
> syncing disks... done
> 
> dumping to dev 0,1 offset 68061
> dump 
> ---
> I'll check news68k/machdep.c and ncr5380sbc.c, too.
> ---
> Izumi Tsutsui
> tsutsui@ceres.dti.ne.jp