Peter Howkins wrote:
On Thu, Oct 9, 2008 at 10:52 PM, Chris Gilbert <chris%dokein.co.uk@localhost> wrote:
Hmm. There was a window where current kernels have a serious interrupt handling bug (now fixed) that usually manifested just after boot as soon as you start to do a reasonable amount of disk activity. Wonder if the kernel you are using suffers from that?Status: Running Command: /sbin/newfs -V2 -O 1 -b 8192 -f 1024 /dev/rwd0a ----------------------------------------------------------------------------- /dev/rwd0a: 804.7MB (1648080 sectors) block size 8192, fragment size 1024 using 18 cylinder groups of 44.71MB, 5723 blks, 11136 inodes. Stopped in pid 10.1 (newfs) at 0x1000: address 0x1000 is invalid andeq r0, r0, r0 db>
Any 4.99.x kernel built after 21/8 should have the fix. Current kernels built between 28/4 and 21/8 will be broken.
Even with that fix I'm still seeing crashes that look a lot like this on my fully installed system when running the overnight scripts. Thats probably a similar load to that done by newfs. I've got some filesystem creation/updating to do on my acorn32 system so I'll see if I can reproduce your newfs on my current build.
If there isn't a recent current install kernel on ftp.netbsd.org let me know I can soon get one built and uploaded where you can download it.
Mike