Subject: Re: one more: stop at haltwo0
To: None <port-sgimips@netbsd.org>
From: Florian Stoehr <netbsd@wolfnode.de>
List: port-sgimips
Date: 12/24/2004 02:04:16
On Fri, 24 Dec 2004, Ilpo Ruotsalainen wrote:

> On Thu Dec 23 2004 at 17:53:39 +0100, Dennis Grevenstein wrote:
>> haltwo0 at hpc0 offset 0x58000
>>
>>
>> It's dead after that. It won't even come back when I
>> send a break via serial console.
>
> The haltwo probe is anything but robust, since the docs (and the
> hardware, far as I could figure out) give no real way of doing totally
> safe probe for device existence.
>
> The driver code could be better though, it should use bus_space_peek()
> and bus_space_poke() before it has confirmed the presence of the audio
> hardware rather than just expecting something to be mapped there
> (actually just checking the address range has *something* mapped would
> probably be a good start for the whole detection business, IIRC nothing
> else can end up mapped at that HPC3 offset anyway).
>
> And the whole damn detection business should be in haltwo_match(), not
> in haltwo_attach() aswell.
>
> -- 
> Ilpo Ruotsalainen - <lonewolf@iki.fi> - http://www.iki.fi/lonewolf/
>

Hm ... I wonder what has changed ... the haltwo probe didn't stop the 
system with 1.6.2...

-Florian