Subject: kern/31477: Heavy disk IO to raid0 locks up system
To: None <kern-bug-people@netbsd.org, gnats-admin@netbsd.org,>
From: None <steven_grunza@ieee.org>
List: netbsd-bugs
Date: 10/05/2005 03:21:00
>Number:         31477
>Category:       kern
>Synopsis:       Heavy disk IO to raid0 locks up system
>Confidential:   no
>Severity:       critical
>Priority:       high
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Wed Oct 05 03:21:00 +0000 2005
>Originator:     Steven Grunza
>Release:        1.6.2 of port-i386
>Organization:
>Environment:
NetBSD barc 1.6.2 NetBSD 1.6.2 (BARC) #0: Thu Nov 4 15:18:00 EST 2004 toor@barc:/usr/src/sys/arch/i386/compile/BARC i386
>Description:
I have a Promise Ultra100TX2/ATA controller on the PCI bus with two 74GB (ST380011A) drives.  Each drive is on it's own bus as a master.  The motherboard (Dell XPS T500) has a DVD+RW drive as a secondary master and a 13 GB IBM drive as a primary master.

Networking is with a 3C905C 10/100 Ethernet card running at 100 Mbps full-duplex.

The system is normally stable and is used as a file server.  If a client connects (samba 3.0.14a) to a samba share and transfers large amounts of data for several minutes then the file server (barc) stops responding to the client and doesn't respond to pings.  The console (vga text mode /w/ PS2 keyboard) also stops responding to the keyboard.  The only way to recover is to reboot using the PC's front panel reset button.

The data transfers are to a file system on raid0 which is a RAID 1 (mirrored) set of disks.

I often see lines in dmesg about "stray interrupt 7".  I haven't found anything that suggests a connection between "stray interrupt 7" and the crashes.

Any help would be welcome.  I would consider moving to a NetBSD 2.x or even NetBSD 3.x version if that would bring some stability.  I would like to stick with using RAID level 1 for increased data-loss prevention.
>How-To-Repeat:
Use FTP or Samba to transfer a large amount (4GB in one case, installing Cygwin from the Samba share in another) of data.
>Fix: