vc4: Detect working driver from cmdline

vc4: Detect working driver from cmdline

[Stefan Wahren]

Hi,
in order to improve kernelci.org results and avoid false positive cases like this [1], i suggested to also test for a working VC4 driver. In order to keep it simple, we should do it from userspace.

My first idea was:

test -d /sys/kernel/debug/dri/0 || echo "Error case"

But maybe you have a better / more reliable solution?

[1] - https://storage.kernelci.org/mainline/master/v4.17-rc5-69-g3acf4e395260/arm64/defconfig/lab-mhart/boot-bcm2837-rpi-3-b.html
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

Re: vc4: Detect working driver from cmdline

[Eric Anholt]

[-- Attachment #1.1: Type: text/plain, Size: 891 bytes --]

Stefan Wahren <stefan.wahren@i2se.com> writes:

> Hi,
> in order to improve kernelci.org results and avoid false positive cases like this [1], i suggested to also test for a working VC4 driver. In order to keep it simple, we should do it from userspace.
>
> My first idea was:
>
> test -d /sys/kernel/debug/dri/0 || echo "Error case"
>
> But maybe you have a better / more reliable solution?
>
> [1] - https://storage.kernelci.org/mainline/master/v4.17-rc5-69-g3acf4e395260/arm64/defconfig/lab-mhart/boot-bcm2837-rpi-3-b.html

GL testing involves running piglit and comparing to a previous run to
confirm no regressions, and we're building out the intel-gpu-tools
support for display testing now.  That's what you'd need to make sure
the driver actually works.  If you just want to make sure module probe
doesn't fail, I guess that test is fine (it would catch bugs we've seen
in the past).

[-- Attachment #1.2: signature.asc --]
[-- Type: application/pgp-signature, Size: 832 bytes --]

[-- Attachment #2: Type: text/plain, Size: 160 bytes --]

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

Re: vc4: Detect working driver from cmdline

[Stefan Wahren]

> Eric Anholt <eric@anholt.net> hat am 21. Mai 2018 um 19:55 geschrieben:
> 
> 
> Stefan Wahren <stefan.wahren@i2se.com> writes:
> 
> > Hi,
> > in order to improve kernelci.org results and avoid false positive cases like this [1], i suggested to also test for a working VC4 driver. In order to keep it simple, we should do it from userspace.
> >
> > My first idea was:
> >
> > test -d /sys/kernel/debug/dri/0 || echo "Error case"
> >
> > But maybe you have a better / more reliable solution?
> >
> > [1] - https://storage.kernelci.org/mainline/master/v4.17-rc5-69-g3acf4e395260/arm64/defconfig/lab-mhart/boot-bcm2837-rpi-3-b.html
> 
> GL testing involves running piglit and comparing to a previous run to
> confirm no regressions, and we're building out the intel-gpu-tools
> support for display testing now.  That's what you'd need to make sure
> the driver actually works.  If you just want to make sure module probe
> doesn't fail, I guess that test is fine (it would catch bugs we've seen
> in the past).

AFAIK kernelci uses a initramfs, so adding much complexity isn't desired. I think module probe is sufficient and an improvement.
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

Re: vc4: Detect working driver from cmdline

[Tomeu Vizoso]

On 21 May 2018 at 21:02, Stefan Wahren <stefan.wahren@i2se.com> wrote:
>
>> Eric Anholt <eric@anholt.net> hat am 21. Mai 2018 um 19:55 geschrieben:
>>
>>
>> Stefan Wahren <stefan.wahren@i2se.com> writes:
>>
>> > Hi,
>> > in order to improve kernelci.org results and avoid false positive cases like this [1], i suggested to also test for a working VC4 driver. In order to keep it simple, we should do it from userspace.
>> >
>> > My first idea was:
>> >
>> > test -d /sys/kernel/debug/dri/0 || echo "Error case"
>> >
>> > But maybe you have a better / more reliable solution?
>> >
>> > [1] - https://storage.kernelci.org/mainline/master/v4.17-rc5-69-g3acf4e395260/arm64/defconfig/lab-mhart/boot-bcm2837-rpi-3-b.html
>>
>> GL testing involves running piglit and comparing to a previous run to
>> confirm no regressions, and we're building out the intel-gpu-tools
>> support for display testing now.  That's what you'd need to make sure
>> the driver actually works.  If you just want to make sure module probe
>> doesn't fail, I guess that test is fine (it would catch bugs we've seen
>> in the past).
>
> AFAIK kernelci uses a initramfs, so adding much complexity isn't desired. I think module probe is sufficient and an improvement.

Yeah, but would be good to have something a bit more generic that
doesn't need for people to write and maintain tests specific to
particular drivers and boards. There has been talk in #kernelci of
storing the list of devices that probed in a run, and consider a
regression when a device stops probing (or even being registered at
all) in a subsequent run. If anybody wants to work on this, I can give
pointers.

Regarding igt-gpu-tools, at Collabora we are working on running these
and other tests in kernelci, by building ramdisks with all the
required dependencies.

Separately, we are also working on running piglit and deqp on virgl in
fdo's gitlab, and hopefully that work will facilitate running piglit
on devices such as the rpis. I don't think that should belong in
kernelci, though.

Regards,

Tomeu
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel