Subject: softdep crash
To: None <port-i386@netbsd.org>
From: Brian Gregor <bgregor@BUPHY.bu.edu>
List: port-i386
Date: 09/28/2001 08:33:47
I've just updated my system to 1.5Y, with code CVS'd two days ago.

Initially my source code was on a RAID 0 partition between a
2 gig SCSI drive and a 2 gig partition on my main IDE drive (had
to do something with the SCSI drive ;-) )

When this was mounted via NFS by my Sparc Classic that does
NAT off my cable modem so it could update from 1.5 to 1.5.2,
the kernel on my PC dropped into the debugger with a softdep
error.  This is a hard crash - a reboot is the only way to
fix it.

Figuring it might be the SCSI disk or the old controller (some
NCR PCI card, only does SCSI-2), I switched to just the IDE
partition and tried again.  The PC successfully did a 'make build'
and rebuilt the kernel overnight with no problems.  Once again
I started a 'make build' on the 1.5.2 sources with the Sparc,
and 15 min later got a crash.

I recall the debugger saying it was in the softdep code. I don't
have the exact text, but I'll wager I can recreate this bug tonight
and post the exact error.

System is a Celeron 366, Intel i810 motherboard, 256 meg RAM.
Boot disk is a 6 gig Maxtor IDE, main disk is a 8.4 gig Quantum
IDE drive.  Other hardware - NCR SCSI controller, Realtek RTL8139
ethernet, 10 Mbs hub.

Besides noting the text from the debugger and a real dmesg output, is
there any other info I should post or gather?


Thanks,

Brian