Re: [PATCH 5/5] tests/acceptance: Test PMON on the Fuloong 2E machine

[BALATON Zoltan <balaton@eik.bme.hu>]

[-- Attachment #1: Type: text/plain, Size: 10004 bytes --]

On Tue, 29 Jun 2021, Philippe Mathieu-Daudé wrote:
> On 6/29/21 12:47 PM, BALATON Zoltan wrote:
>> On Tue, 29 Jun 2021, Philippe Mathieu-Daudé wrote:
>>> On 6/24/21 10:43 PM, BALATON Zoltan wrote:
>>>> On Thu, 24 Jun 2021, Philippe Mathieu-Daudé wrote:
>>>>> Test the PMON firmware. As the firmware is not redistributable,
>>>>> it has to be downloaded manually first. Then it can be used by
>>>>> providing its path via the PMON_BIN_PATH environment variable:
>>>>>
>>>>>  $ PMON2E_BIN_PATH=~/images/fuloong2e/pmon_2e.bin \
>>>>>    AVOCADO_ALLOW_UNTRUSTED_CODE=1 \
>>>>>    avocado --show=app,console run
>>>>> tests/acceptance/machine_mips_fuloong2e.py
>>>>>  Fetching asset from
>>>>> tests/acceptance/machine_mips_fuloong2e.py:MipsFuloong2e.test_linux_kernel_isa_serial
>>>>>
>>>>>
>>>>>   (1/3)
>>>>> tests/acceptance/machine_mips_fuloong2e.py:MipsFuloong2e.test_pmon_serial_console:
>>>>>
>>>>>
>>>>>  console: PMON2000 MIPS Initializing. Standby...
>>>>>  console: ERRORPC=00000000 CONFIG=00030932
>>>>>  console: PRID=00006302
>>>>>  console: Init SDRAM Done!
>>>>>  console: Sizing caches...
>>>>>  console: Init caches...
>>>>>  console: godson2 caches found
>>>>>  console: Init caches done, cfg = 00030932
>>>>>  console: Copy PMON to execute location...
>>>>>  console: copy text section done.
>>>>>  console: Copy PMON to execute location done.
>>>>>  Uncompressing Bios........................OK,Booting Bios
>>>>>  PASS (0.25 s)
>>>>>   (2/3)
>>>>> tests/acceptance/machine_mips_fuloong2e.py:MipsFuloong2e.test_pmon_framebuffer_console:
>>>>>
>>>>>
>>>>>  [...]
>>>>>  Uncompressing Bios........................OK,Booting Bios
>>>>>  console: FREQ
>>>>>  console: FREI
>>>>>  console: DONE
>>>>>  console: TTYI
>>>>>  console: TTYD
>>>>>  console: ENVI
>>>>>  console: MAPV
>>>>>  console: Mfg  0, Id 60
>>>>>  console: STDV
>>>>>  console: SBDD
>>>>>  console: PPCIH
>>>>>  console: PCIS
>>>>>  console: PCIR
>>>>>  console: PCIW
>>>>>  console: NETI
>>>>>  console: RTCL
>>>>>  console: PCID
>>>>>  console: VGAI
>>>>>  console: Default MODE_ID 2
>>>>>  console: starting radeon init...
>>>>>  console: radeon init done
>>>>>  console: FRBI
>>>>>  console: cfb_console init,fb=b4000000
>>>>>  console: Video: Drawing the logo ...
>>>>>  console: CONSOLE_SIZE 450560HSTI
>>>>>  PASS (4.10 s)
>>>>>   (3/3)
>>>>> tests/acceptance/machine_mips_fuloong2e.py:MipsFuloong2e.test_linux_kernel_isa_serial:
>>>>>
>>>>>
>>>>>  console: Linux version 2.6.27.7lemote (root@debian) (gcc version
>>>>> 4.1.3 20080623 (prerelease) (Debian 4.1.2-23)) #6 Fri Dec 12 00:11:25
>>>>> CST 2008
>>>>>  console: busclock=33000000,
>>>>> cpuclock=-2145008360,memsize=256,highmemsize=0
>>>>>  console: console [early0] enabled
>>>>>  console: CPU revision is: 00006302 (ICT Loongson-2)
>>>>>  PASS (0.19 s)
>>>>>  RESULTS    : PASS 3 | ERROR 0 | FAIL 0 | SKIP 0 | WARN 0 | INTERRUPT
>>>>> 0 | CANCEL 0
>>>>>  JOB TIME   : 5.10 s
>>>>>
>>>>> Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
>>>>> ---
>>>>> tests/acceptance/machine_mips_fuloong2e.py | 62 ++++++++++++++++++++++
>>>>> 1 file changed, 62 insertions(+)
>>>>>
>>>>> diff --git a/tests/acceptance/machine_mips_fuloong2e.py
>>>>> b/tests/acceptance/machine_mips_fuloong2e.py
>>>>> index 0ac285e2af1..4854ba98560 100644
>>>>> --- a/tests/acceptance/machine_mips_fuloong2e.py
>>>>> +++ b/tests/acceptance/machine_mips_fuloong2e.py
>>>>> @@ -8,15 +8,77 @@
>>>>> # SPDX-License-Identifier: GPL-2.0-or-later
>>>>>
>>>>> import os
>>>>> +import time
>>>>>
>>>>> from avocado import skipUnless
>>>>> from avocado_qemu import Test
>>>>> from avocado_qemu import wait_for_console_pattern
>>>>>
>>>>> +from tesseract_utils import tesseract_available, tesseract_ocr
>>>>> +
>>>>> class MipsFuloong2e(Test):
>>>>>
>>>>>     timeout = 60
>>>>>
>>>>> +    @skipUnless(os.getenv('AVOCADO_ALLOW_UNTRUSTED_CODE'), 'untrusted
>>>>> code')
>>>>> +    @skipUnless(os.getenv('PMON2E_BIN_PATH'), 'PMON2E_BIN_PATH not
>>>>> available')
>>>>> +    def test_pmon_serial_console(self):
>>>>> +        """
>>>>> +        :avocado: tags=arch:mips64el
>>>>> +        :avocado: tags=machine:fuloong2e
>>>>> +        :avocado: tags=endian:little
>>>>> +        :avocado: tags=device:bonito64
>>>>> +        :avocado: tags=device:via686b
>>>>> +        """
>>>>> +        pmon_hash = 'c812e1695d7b2320036f3ef494976969' # v1.1.2
>>>>> +        pmon_path = self.fetch_asset('file://' +
>>>>> os.getenv('PMON2E_BIN_PATH'),
>>>>> +                                     asset_hash=pmon_hash,
>>>>> algorithm='md5')
>>>>> +
>>>>> +        self.vm.set_console()
>>>>> +        self.vm.add_args('-bios', pmon_path)
>>>>> +        self.vm.launch()
>>>>> +        wait_for_console_pattern(self, 'PMON2000 MIPS Initializing.
>>>>> Standby...')
>>>>> +        wait_for_console_pattern(self, 'Booting Bios')
>>>>> +
>>>>> +    @skipUnless(os.getenv('AVOCADO_ALLOW_UNTRUSTED_CODE'), 'untrusted
>>>>> code')
>>>>> +    # Tesseract 4 adds a new OCR engine based on LSTM neural
>>>>> networks. The
>>>>> +    # new version is faster and more accurate than version 3. The
>>>>> drawback is
>>>>> +    # that it is still alpha-level software.
>>>>> +    @skipUnless(tesseract_available(4), 'tesseract v4 OCR tool not
>>>>> available')
>>>>> +    @skipUnless(os.getenv('PMON2E_BIN_PATH'), 'PMON2E_BIN_PATH not
>>>>> available')
>>>>> +    def test_pmon_framebuffer_console(self):
>>>>> +        """
>>>>> +        :avocado: tags=arch:mips64el
>>>>> +        :avocado: tags=machine:fuloong2e
>>>>> +        :avocado: tags=endian:little
>>>>> +        :avocado: tags=device:bonito64
>>>>> +        :avocado: tags=device:ati-vga
>>>>> +        """
>>>>> +        screenshot_path = os.path.join(self.workdir, 'dump.ppm')
>>>>> +
>>>>> +        pmon_hash = 'c812e1695d7b2320036f3ef494976969' # v1.1.2
>>>>> +        pmon_path = self.fetch_asset('file://' +
>>>>> os.getenv('PMON2E_BIN_PATH'),
>>>>> +                                     asset_hash=pmon_hash,
>>>>> algorithm='md5')
>>>>> +
>>>>> +        self.vm.set_console()
>>>>> +        self.vm.add_args('-bios', pmon_path,
>>>>> +                         '-vga', 'std',
>>>>> +                         '-device', 'ati-vga,model=rv100')
>>>>
>>>> I think this is the default if you just drop -vga std so I don't know
>>>> why you have that in the first place but then you should not need to add
>>>> ati-vga explicitely.
>>>
>>> I thought this is what you asked me here:
>>> https://www.mail-archive.com/qemu-devel@nongnu.org/msg753832.html
>>
>> No, I've said the same in that message that fuloong2e has an ati-vga as
>> default so no -vga option should be needed at all.
>>
>>> Currently:
>>>
>>> console: VGAI
>>> console: Default MODE_ID 2
>>> console: starting radeon init...
>>> console: iobase=bfd0a200,mmbase=b6064000
>>> console: mc_status=5
>>> console: mc_status=5
>>> console: mc_status=5
>>> console: mc_status=5
>>> console: ppll_div_3 = 301f4
>>> console: Wrote: 0x00000043 0x000301f4 0x00000000 (0x00000000)
>>> console: Wrote: rd=67, fd=500, pd=3
>>> console: VCLK_ECP_CNTL = 000000C3
>>> console: radeon init done
>>> console: FRBI
>>> console: cfb_console init,fb=b5000000
>>> console: Video: Drawing the logo ...
>>> console: CONSOLE_SIZE 450560HSTI
>>> PASS (4.53 s)
>>>
>>> Without '-vga std -device ati-vga,model=rv100':
>>>
>>> console: VGAI
>>> console: Default MODE_ID 2
>>> console: starting radeon init...
>>> INTERRUPTED: Test interrupted by SIGTERM
>>> Runner error occurred: Timeout reached... (60.29 s)
>>
>> That's strange (with the REG_MASK fixed to 0xff in bonito.c) I get:
>>
>> $ qemu-system-mips64el -M fuloong2e -bios pmon_2e.bin -serial stdio
>> [...]
>> VGAI
>> Default MODE_ID 2
>> starting radeon init...
>> iobase=bfd0a100,mmbase=b5050000
>> mc_status=5
>> mc_status=5
>> mc_status=5
>> mc_status=5
>> ppll_div_3 = 301f4
>> Wrote: 0x00000043 0x000301f4 0x00000000 (0x00000000)
>> Wrote: rd=67, fd=500, pd=3
>> VCLK_ECP_CNTL = 000000C3
>> radeon init done
>> FRBI
>> cfb_console init,fb=b4000000
>> Video: Drawing the logo ...
>> CONSOLE_SIZE 450560HSTI
>>
>> So I think you should not need either -vga std nor -device ati-vga as
>> those would add another VGA card to the one already on the board.
>
> So after looking in git history, Avocado tests inherited iotests,
> and since commit 0fd05e8dd1e ("qemu-iotests: start vms in qtest mode")
> the VMs are started with -display none -vga none.
> See in python/qemu/machine/machine.py:
>
>    def _base_args(self) -> List[str]:
>        args = ['-display', 'none', '-vga', 'none']
>
> Which is why we have to use '-vga std' here.
>
> Except if you disagree, I'll queue the patch using:
>
>      self.vm.add_args('-bios', pmon_path,
>                       '-vga', 'std')

I think it's better modelling the original machine if you use the ati-vga 
that the machine has instead of std vga that it doesn't. Also it tests 
ati-vga that way while it doesn't with -vga std. I also don't see the 
point having both. With this command I get:

$ qemu-system-mips64el -M fuloong2e -bios pmon_2e.bin -serial stdio 
-display none -vga none -device ati-vga,model=rv100

VGAI
Default MODE_ID 2
starting radeon init...
iobase=bfd0a100,mmbase=b5050000
mc_status=5
mc_status=5
mc_status=5
mc_status=5
ppll_div_3 = 301f4
Wrote: 0x00000043 0x000301f4 0x00000000 (0x00000000)
Wrote: rd=67, fd=500, pd=3
VCLK_ECP_CNTL = 000000C3
radeon init done
FRBI
cfb_console init,fb=b4000000
Video: Drawing the logo ...
CONSOLE_SIZE 450560HSTI

and I think that matches the default config of the machine that -vga none 
turns off. What do you need std vga for?

Regards,
BALATON Zoltan