Subject: Re: A Boot!
To: None <Chris.Smith@raytheon.co.uk>
From: Ben Harris <bjh21@cam.ac.uk>
List: port-arm26
Date: 05/02/2000 11:11:07
On Tue, 2 May 2000 Chris.Smith@raytheon.co.uk wrote:

> Does BBBB's page memory map output have a key?

Erm, RTFS?  ALternatively, I suppose I could do so...

Each character in the map represents one page of (physical) RAM.  The
ticks along the top are to stop you getting lost.  The characters in the
map indicate what the memory's being used for (actually where it's
logically mapped):

0 -> zero-page
+ -> BBBB workspace
* -> Free space (BBBB wants to put the kernel here)
d -> RAM disc
s -> System sprite area
m -> RMA
h -> System heap/stack
f -> Font cache
S -> Screen memory

[ This section is from memory ]

On a machine with 32k pages (which is all NetBSD/arm26 supports), the left
half of the first line is potential screen memory, and hence not used by
BBBB.  The next page is usually zero page under RISC OS, and is used for
zero page under NetBSD.  The next is usually system heap under RISC OS,
and is used for process 0's kernel stack under NetBSD.  The next is used
for the message buffer under NetBSD.  Pages from there on are used to load
the kernel, and must be free if BBBB is to do so successfully.  Future
bootloaders should load the kernel into whatever pages are free, then kick
out RISC OS and shuffle them into the right shape.  This is left as an
exercise for the enthusiatic reader.

> There is only one problem (mine), is that my monitor loses sync
> after about 2 seconds of it booting (sync too low).  I have a
> proper Acorn monitor AKF60 in the post so I'l have a proper boot
> soon :)  I'll even see things past the copyright message before
> it loses sync then :)

What monitortype is the machine configured for?  VGA or SVGA (sorry, can't
remember the numbers) should drive most PC monitors happily.

> [for Ben >> Can we have a serial console for development? :) ]

Well, I can probably do it for Archimedes-class machines (which use
6551s), but I haven't got an IOEB machine (which use 16450-clones), so
testing it there is a little harder.

-- 
Ben Harris
Unix Support, University of Cambridge Computing Service.