Subject: Re: Deskstation Tyne booting ALPHA2
To: Maki Kato <mk2s@digitalcommute.com>
From: Noriyuki Soda <soda@sra.co.jp>
List: port-arc
Date: 10/17/2000 02:54:22
>>>>> On Sat, 14 Oct 2000 13:10:09 -0400,
	Maki Kato <mk2s@digitalcommute.com> said:

> I had to recompile ARCDIAG to start at a0600000.  I got that number from 
> a ARCINST.EXE which was the only utility that loaded(but didn't run too 
> well).  The following is the arc diag for my machine.

Thanks.

> magic:      SCRA

Mmmm, not "ARCS", but "SCRA".

>        class: Processor, type: CPU
		:
>         id: 38:[Tyne v4634x  33/134  ARCSBIOS 1.3.026\x00]

Hmmm, R4634...

>        class: Adapter, type: MultiFunctionAdapter
>         flags:
>         version: 0, revision: 1
>         key: 0
>         affinity_mask: 0xffffffff
>         config_data_len: 0
>         id: 4:[ISA\x00]

Hmmm, interesting.
No [Jazz-Internal Bus], but [ISA] bus as expected.

>        class: Adapter, type: SCSIAdapter
		:
>         id: 8:[AHA154X\x00]

Hmm, the firmware knows AHA-154x.

>>>>> On Mon, 16 Oct 2000 01:12:57 -0400,
	Maki Kato <mk2s@digitalcommute.com> said:

> I got my Deskstation to boot into single user mode after some messing 
> around.

Great!

> The following are a few notes.

> - this box seems to not like to boot the ecoff output from objcopy, 
> instead I used elf2ecoff.

Hmmm.

> - I had some spooky exceptions and panics, that would go away when a 
> printf was inserted above the problem code.  I then modified locore.h to 
> enable mips3_CacheFlush_2way and other 2way cache code, and the problems 
> went away.

Yeah, this is correct fix, since R46xx has two way cache.
I seem to recall that (at least early) R46xx chip has a bug which
requires to disable interrupt while flushing cache.
Perhaps your machine may need the workaround. But I'm not sure.
--
soda