Subject: uvm_fault kernel: page fault trap while un-tar-ing a large file
To: None <>
From: =?ISO-8859-1?Q?Edgar_Fu=DF?= <>
List: port-amd64
Date: 06/20/2007 21:24:38
I already posted this to tech-kern, but maybe it's amd64 specific.

About a week ago, I got the following on our file sever while un-tar-ing
a huge (8G) archive:
(this is all typed from handwritten notes, so probably slightly wrong)

uvm_fault(0xffff80004e263b10, 0x10, 2)-> e
kernel: page fault trap, code=0
Stopped in pid 8186.1 (tar) at netbsd:dqget+0x118 movq %rdx,0(%rax)

[writing on the data RAID for about 20 seconds]
panic: locking against myself
Stopped in pid 8186.1 (tar) at netbsd:cpu_Debugger+0x1 ret
dump device bad

Looks like I have my dumps misconfigured.

Unfortunately, bt didn't show any more information.

The system runs 4.0 beta2 as of 03Jun plus fvdl's IOAPIC patch.

Today, I got nearly the same, but in the meantime, I had my dump device
set up correctly, so after ~40 minutes I had a core file.

Could someone please help me investigate this?
Kernel panics on a file server can really spoil your day.
Approx. 40 mins downtime (dump, savecore, fsck. quotacheck) are no  
fun either.

Thanks for any help.