Re: [PATCH 0/6] tests/acceptance: Test the Fuloong 2E machine

[BALATON Zoltan via <qemu-devel@nongnu.org>]

[-- Attachment #1: Type: text/plain, Size: 3559 bytes --]

On Sun, 25 Oct 2020, Philippe Mathieu-Daudé wrote:
> On 10/25/20 12:18 PM, Mark Cave-Ayland wrote:
>> On 21/10/2020 11:50, Philippe Mathieu-Daudé wrote:
>> 
>>> This series add tests for the MIPS Fuloong 2E machine.
>>> 
>>> I have been using these tests for over 1 year now.
>>> After recent discussions with Huacai and Mark regarding
>>> how to contribute tests for machines using private blob,
>>> I think it is time to publish this series, to see how
>>> the final part will be discussed, as it tests a blob
>>> which Lemote company doesn't publish the sources. The
>>> binary is know to work and has been used for various
>>> years on real hardware. Its MD5 is also know (I am not
>>> using SHA-1 because it has been published with MD5).
>>> 
>>> The idea is to be able to share tests without infringing
>>> licenses issues.
>>> 
>>> Please comment :)
>>> 
>>> Phil.
>>> 
>>> Philippe Mathieu-Daudé (6):
>>>    tests/acceptance: Remove unused import
>>>    tests/acceptance: Use .ppm extention for Portable PixMap files
>>>    tests/acceptance: Extract tesseract_available() helper in new
>>>      namespace
>>>    tests/acceptance: Introduce tesseract_ocr() helper
>>>    tests/acceptance: Test Linux on the Fuloong 2E machine
>>>    tests/acceptance: Test PMON on the Fuloong 2E machine
>>> 
>>>   MAINTAINERS                                |   1 +
>>>   tests/acceptance/machine_m68k_nextcube.py  |  51 ++---------
>>>   tests/acceptance/machine_mips_fuloong2e.py | 101 +++++++++++++++++++++
>>>   tests/acceptance/tesseract_utils.py        |  46 ++++++++++
>>>   4 files changed, 158 insertions(+), 41 deletions(-)
>>>   create mode 100644 tests/acceptance/machine_mips_fuloong2e.py
>>>   create mode 100644 tests/acceptance/tesseract_utils.py
>> 
>> In general this looks good: certainly I don't see any reason why the 
>> tesseract_ocr helper changes shouldn't be merged, since I expect that this 
>> is something that will be used more in future.
>> 
>> Looking at the fuloong test it seems that it stops fairly early in the 
>> kernel boot: one of the problems that Zoltan's VIA southbridge patches were 
>> trying to solve earlier in the year was because of problems with PCI IRQ 
>> routing, so it would be nice to include a small OS image that can be used 
>> to boot to userspace to ensure that such problems can be detected in 
>> future.
>
> The Linux tests (ab)use of the rescue image prepared for the Yeeloong
> machine which we don't model (Loongson 2F CPU, 1GB of DRAM). The kernel
> triggers DATA ABORT while accessing >256MB because we do not model
> mapping more than 256MB :/ (The Fuloong comes with 512MB extensible to 1GB). 
> I have a branch adding that but had the idea nobody was interested
> in the Fuloong 2E so I didn't insist (I have more interest in embedded
> 32-bit CPUs). I might try to post it after the release, as it makes this
> test complete.

I've noticed that missing high RAM when cleaning up SPD EEPROMs and addded 
a FIXME comment but did't know MIPS enough to attemt to fix it. Anyway I'm 
also only interested in fuloong2e as a test case for the devices also used 
by Pegasos2 (that I'll need to update and clean up to submit eventually) 
but not beyond that.

I haven't seen these patches before, so I've spotted only now that your 
test case uses -vga std. Shouldn't it work with ati-vga,model=rv100 which 
I think is the default for this machine and that's what the pmon firmware 
expects and has a driver for?

Regards,
BALATON Zoltan