Chipset docs

<ul>

Since I didn't get a reply I'll try to be more specific.

Where can I for example find the Schizo documentation?

I'd love to get my hands on a copy so that we can make OpenBSD run on these machines.

</ul>

And we'd be delighted to see OpenBSD running on these machines!"The more, the merrier".

Apparently the right people weren't monitoring the forum when you posted it the first time... So I'll offer an initial response plus have forwarded a message to others who might be able to directly respond to the chipset docs question.

My initial response is, if I was porting an OS to run on existing

platforms (e.g. Sun's T1000 and T2000 boxes) and future platforms along

the same line,

I would port the OS to the

<a HREF="http://opensparc.sunsource.net/specs/Hypervisor-api-current-draft.pdf"> ;]

Hypervisor API spec </a>, not to the bare hardware.

The

<a HREF="http://opensparc.sunsource.net/specs/Hypervisor-api-current-draft.pdf"> ;]

Hypervisor API spec </a>

is downloadable from the

<a HREF="http://opensparc.sunsource.net/nonav/opensparct1.html"> OpenSPARC T1 home page</a> via the "UltraSPARC T1

Hypervisor API Specification" link).

If you port to the Hypervisor API, the OS port won't be any

more difficult than porting to bare hardware. In fact, might well be easier** -- for one thing, you won't have to read up on all the mind-numbing details of low-level devices and handling hardware errors. The effort of doing an OS port to run on top of the Hypervisor will only have to be done once to cover many generations of systems. The OS port

will continue to run on many later-generation hardware platforms -- much as application software can run across multiple generations of OS versions if it follows a specified API for the OS.

Once anyone understands about the Hypervisor, I'm not sure why they

would want to port to "bare hardware" on a T1000/T2000 platform, given

that all the low-level work has already been done for them (in the

Hypervisor firmware). On the T1000/T2000, porting an OS to bare hardware

when the Hypervisor API exists is akin to porting an appplication

software package to run on bare hardware when a fully-functional OS with

a robust API is already available on that harware to port the

application to.

...Note that this response is different than if the same question was

asked even 6 months ago. Back then, it would have been about porting an OS to

older UltraSPARC(up through UltraSPARC IV+)-based systems. Those

processors do not support hyperprivileged mode or hypervisor software,

so doing a port to a Hypervisor API was not an option back then -- a messier "bare hardware port" was the only option.

If there are strong motivations to do a "bare hardware" OS port (perhaps

to an OpenSPARC derivative other than UltraSPARC T1?), then yes you

would need chipset info to go with that particular processor.

Re Sun chipset info, I've sent a message to some folks who might be able to respond to that ...

= Dave

** you might have heard that one open-source OS already has a functional kernel, running on top of Hypervisor, after less than 2 months of effort ... there's more work to be done, but the core OS appears to be working well, from what I've read. So there's some initial anecdotal evidence that porting to Hypervisor isn't too painful :-)

dweaver at 2007-7-6 13:34:48 >

I don't understand your reply. You start off by saying that you would love to see OpenBSD running on machines using schizo, and then you go on to say you don't see the point.

I would like to point out that Sun is selling more than just T1000 and T2000's. All the other UltraSparc systems for sale from Sun are based around chips that lack the hypervisor. To run on those we need to talk to the bare metal. This is especially true of the UltraSparc workstations (man, I'd love one of those), and of all the currently deployed Sun systems.

The range of Suns systems seems to reflect the reality that one solution doesn't fulfil everyones problems. To use a metaphor, square pegs wont necessarily fit in a round hole. So Sun sells both round and square pegs.

However, hardware is only part of a solution, the software you run is part of it too. By giving the community the chipset docs you're giving them the opportunity to have a wider choice in software to use. To continue the metaphor: you'll have more pegs to fit into all those different holes.

dlg at 2007-7-6 13:34:48 >

Quick update -- I've received confirmation that Hypervisor code resides in flash memory on the T1000/T2000 platforms.An update can be downloaded (e.g. via FTP) and installed on the system via a special command (presumably a boot PROM command).

So yes, Hypervisor code can indeed be updated ... but such an event is expected to be rare (if ever) on most systems.

dweaver at 2007-7-6 13:34:48 >