Re: [Qemu-devel] [PULL 30/30] target-sparc: fix up niagara machine

[Artyom Tarasenko]

On Mon, Jan 23, 2017 at 3:24 PM, Peter Maydell <address@hidden> wrote:
> On 23 January 2017 at 14:10, Artyom Tarasenko <address@hidden> wrote:
>> On Mon, Jan 23, 2017 at 1:40 PM, Peter Maydell <address@hidden> wrote:
>>> I see that 'make check' now warns:
>>>   GTESTER check-qtest-sparc64
>>> Could not open option rom 'nvram1': No such file or directory
>>> Could not open option rom '1up-md.bin': No such file or directory
>>> Could not open option rom '1up-hv.bin': No such file or directory
>>> Could not open option rom 'reset.bin': No such file or directory
>>> Could not open option rom 'q.bin': No such file or directory
>>> Could not open option rom 'openboot.bin': No such file or directory
>>>
>>> (though the tests still pass).
>>>
>>> Could we either ship these images in pc-bios if they're
>>> necessary, or not complain that they don't exist if they're
>>> not necessary, please?
>>
>> I wonder what would be the best option  here. The images are
>> necessary, so the last option - not complaining - can be misleading
>> for a user.
>
> If they're actually necessary then perhaps we should refuse
> to start entirely?

Yes, I think it's a best option. Don't load any images with the
-nodefaults option and fail on missing ones wintout -nodefaults.

Is there a failing variant of rom_add_file_fixed (I guess it's not
uncommon, but I don't find it in include/hw/loader.h) or do I just
check the return status?

>> Concerning shipping them.
>> Pros:
>> - the images are obviously freely distributable (the link above).
>> - the corresponding source code was open-sourced by Sun under various
>> licenses (GPL for hypervisor, BSD for openboot).
>> Cons:
>> - there is no exact tag the the OpenSPARC source tree which would
>> correspond to the binaries.
>> - building them is tricky, because it requires Solaris 9 / SPARC.
>>
>> What do you think would be a better option?
>
> One thing we could do is only warn if !qtest_enabled().
> We do this for some other boards that otherwise fail entirely
> when their BIOS image is not present. This is sufficient for
> the qtest checks which don't actually try to run code on the
> guest, but merely interact with it via the qtest protocol.
>
> We do ship some other ROMs that are only buildable on the
> right host hardware, so it's not impossible, but I don't know
> the details of our rules about what we put in pc-bios/.

Who may know them? I see no general maintainer for the pc-bios
directory as such in our MAINTAINERS file.

-- 
Regards,
Artyom Tarasenko

SPARC and PPC PReP under qemu blog: http://tyom.blogspot.com/search/label/qemu