Re: [PATCH] Changes to support booting NetBSD/alpha

[Philippe Mathieu-Daudé]

On 10/7/20 5:13 PM, Jason Thorpe wrote:
> 
>> On Oct 6, 2020, at 9:42 PM, Philippe Mathieu-Daudé <f4bug@amsat.org> wrote:
>>
>> Hi Jason,
>>
>> Well, this is not the correct way to do that, so this patch
>> is unlikely to be accepted. We don't want Frankenstein models.
>>
>> What is it you miss from the i82378? Why not implement the cy82c693ub?
>>
>> The code you want to modify is in hw/alpha/dp264.c, see clipper_init().
> 
> The Clipper emulation *already* implements a Frankenstein model, a 
> combination of hardware that DEC never shipped.  While some Tsunami-based 
> systems did use the Cypress bridge, none of them, to my knowledge, had the 
> CMD646 IDE controller.  The only system I'm aware of that shipped with a 
> CMD646 IDE controller was the early version of the Miata (EV56 + Pyxis-based 
> Digital Personal Workstation), and that system had the Intel i82378 PCI-ISA 
> bridge.
> 
> The are a lot of differences in the Cypress PCI-ISA bridge, but most 
> importantly, it has a different interrupt controller (the ELCR registers are 
> programmed differently), and it has a different built-in IDE controller.  As 
> far as I can tell, Qemu does not currently have emulation for the Cypress 
> interrupt controller nor the Cypress IDE controller.
> 
> All Alpha systems that had a PCI-ISA bridge[*] shipped with, as far as I'm 
> aware, one of 3 possible chips:
> 
> - Intel i82378 SIO
> - Cypress 82C693
> - Acer Labs M1533
> 
> ...and the emulation provided by the existing Clipper model implements the 
> Intel flavor (including the interrupt controller).  I just want software that 
> expects one of those 3 chips to be there to work.  The existing emulation 
> seems to assume that system software will just go out and probe the 
> peripherals in the ISA space, but some operating systems (NetBSD in 
> particular) support Alpha systems that have PCI without any ISA/LPC bus at 
> all, and thus only go looking for ISA peripherals if a PCI-ISA bridge is 
> present.  That is what my change here is intended to do, without changing any 
> of the rest of the hardware configuration.

Can you provide the information to reproduce your problem
with a NetBSD image please?

> 
> 
> [*] ...ignoring the weirdo early models that had PCI + EISA + oddly-wired-up 
> junk I/O chips.  Also, I don't have any direct experience with post-Titan 
> models, so I'm not sure what they have...
> 
> -- thorpej
> 
>