Subject: Re: Evbarm questions
To: McGranaghan, Sean <SMcGranaghan@vanteon.com>
From: Richard Earnshaw <email@example.com>
Date: 02/16/2004 15:11:10
> Hello all,
> I have recently started working with the evbarm port of NetBSD. Please feel
> free to re-direct me to an appropriate list for followup if needed.
Port-arm is normally the correct place for evbarm questions.
> I have two questions:
> 1.) I have configured a kernel to boot from a ramdisk. The default size is
> 8192 blocks (4M). If I try and increase the size of the ramdisk my new
> kernel panics with the following error:
> "panic: pmap_map_chunk: no L2 table for VA 0xc0800000
> Undefined instruction exception!!!"
> I have set MEMORY_DISK_ROOT_SIZE=16384 in my config file and set the image
> size to 8192k for the makefs command. Am I missing something else?
This has come up before, I can't remember the precise details, but it's
something to do with the number of static L2 page tables that are loaded
into the kernel pmap when the system starts. You might be able to find an
answer by searching back through the port-arm archives.
> 2.) I have been trying to build a simple LKM under evbarm. When I use
> modload to load a skeleton driver ld segfaults:
> "ld -R /dev/ksyms -e fibo_lkmentry -o /tmp/fibo -Ttext 0x0 /tmp/fibo.o
>  Segmentation fault (core dumped) ld -R /dev/ksyms...
> modload: can't prelink 'fibo.o' creating 'fibo'
> After searching the archives I found several emails and a bug report that
> mention issues with LKM's for evbarm. Issue # port-arm:22015 specifically.
> What is the current status of LKM's and evbarm? Is this a bug in ld?
I don't think these are working yet.