Subject: Re: 1.3.2 panics when creating big files?
To: Jukka Marin <jmarin@pyy.jmp.fi>
From: Chuck Silvers <chuq@chuq.com>
List: port-i386
Date: 04/08/1999 15:15:40
Jukka Marin writes:
> I'm trying to create two 4 GB and one 8 GB file using dd (or cat which
> produces the same result) under NetBSD 1.3.2.  The machine has 80 MB of
> RAM, 16 MB of which has been configured as buffer cache.  There are six
> SCSI disks with an Adaptec UW controller.
> 
> The system panics after writing about 3.8 to 4.2 GB of data.  The panics
> look like this:
> 
> vm_fault(0xf09ee200, 0, 3, 0) -> 1
> fatal page fault in supervisor mode
> trap type 6 code 2 eip f0139e4e cs 8 eflags 10246 cr2 8 cpl c000c840
> panic: trap
> syncing disks... 13 13 6 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 giving up
> 
> or this:
> 
> vm_fault(0xf098a100, 0, 3, 0) -> 1
> fatal page fault in supervisor mode
> trap type 6 code 2 eip f0139e4e cs 8 eflags 10246 cr2 8 cpl c000c840
> panic: trap
> syncing disks... 3 3 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 giving up
> 
> I'm compiling a 1.3.3 kernel now to see if that fixes the problem, but
> I doubt it.
> 
> I guess 1.4 fixes this (thanks to the new VM system), but it may also
> introduce new problems like this, so I don't want to install 1.4ALPHA
> on this production system yet.

I wouldn't be so sure of that...  all that "vm_fault(...) -> 1" means
is that something in the kernel touched an unmapped address.
could be anything, most likely some filesystem code in this case.
maybe some problem crossing the 32bit filesize boundary...


> Any hints?  I should really get this system back up, but every time I
> write those files it crashes..  (I'm trying to put one file per disk,
> I'm not using ccd on these disks, just plain FFS.)

if you have a crash dump from this panic, could you put it
somewhere I could could ftp it?

-Chuck


> As always, thanks a bunch..
> 
>   -jm