"device-tree" test failed on vexpress A9 ubuntu.

in linaro/configs/ubuntu.conf: CONFIG_PROC_DEVICETREE=y

The support for device tree in /proc doesn't seems to work on Vexpress.

Ryan/Tixy, opinion?

It this test worth bothering with? Only the A9 CoreTile is capable of booting without device-tree and that ability is probably going away in the Linux 3.9.

As to why the test fails, that is because the vexpress kernels are built with the ubuntu-minimal.conf config fragment, not ubuntu.conf. I.e. with the 25 config options needed to boot Ubuntu, not the 2130 'kitchen sink' config options that upstream Ubuntu ships with to support every X86 PC, server and peripheral on the planet. ;-)

And there doesn't seem much point in adding CONFIG_PROC_DEVICETREE to ubuntu-minimal.conf if it's sole purpose is to check that you've configured your bootloader to boot with device-tree.

On 7 December 2012 12:45, Tixy (Jon Medhurst)
<email address hidden> wrote:
> It this test worth bothering with? Only the A9 CoreTile is capable of
> booting without device-tree and that ability is probably going away in
> the Linux 3.9.
>
> As to why the test fails, that is because the vexpress kernels are built
> with the ubuntu-minimal.conf config fragment, not ubuntu.conf. I.e. with
> the 25 config options needed to boot Ubuntu, not the 2130 'kitchen sink'
> config options that upstream Ubuntu ships with to support every X86 PC,
> server and peripheral on the planet. ;-)
>
> And there doesn't seem much point in adding CONFIG_PROC_DEVICETREE to
> ubuntu-minimal.conf if it's sole purpose is to check that you've
> configured your bootloader to boot with device-tree.

device-tree test is part of the basic requirements tests. I'll prefer
to have a single test to cover all the boards than treat vexpress as
an exception. +1 for CONFIG_PROC_DEVICETREE in ubuntu-minimal.conf.

On Fri, 2012-12-07 at 11:27 +0000, Fathi Boudra wrote:
> device-tree test is part of the basic requirements tests. I'll prefer
> to have a single test to cover all the boards than treat vexpress as
> an exception. +1 for CONFIG_PROC_DEVICETREE in ubuntu-minimal.conf.

Android has similar test case and it would probably be part of any
platform's testing, so I think the proper place is to put the config is
in linaro-base.conf...

http://lists.linaro.org/pipermail/linaro-dev/2012-December/014708.html

On 7 December 2012 15:50, Tixy (Jon Medhurst)
<email address hidden> wrote:
> On Fri, 2012-12-07 at 11:27 +0000, Fathi Boudra wrote:
>> device-tree test is part of the basic requirements tests. I'll prefer
>> to have a single test to cover all the boards than treat vexpress as
>> an exception. +1 for CONFIG_PROC_DEVICETREE in ubuntu-minimal.conf.
>
> Android has similar test case and it would probably be part of any
> platform's testing, so I think the proper place is to put the config is
> in linaro-base.conf...
>
> http://lists.linaro.org/pipermail/linaro-dev/2012-December/014708.html

Make sense. Thanks Tixy!

http://git.linaro.org/gitweb?p=kernel/configs.git;a=commitdiff;h=0bb0ec12da402957b7caf7ed51839189df14aedd

latest test result: http://validation.linaro.org/lava-server/dashboard/attachment/209460/view

has_proc_device_tree: PASS
device_tree_model_not_empty: FAIL - Unable to find /proc/device-tree/model

@Tixy, is it expected to have the device tree model empty?

On Sat, 2012-12-15 at 10:43 +0000, Fathi Boudra wrote:
> latest test result: http://validation.linaro.org/lava-
> server/dashboard/attachment/209460/view
>
> has_proc_device_tree: PASS
> device_tree_model_not_empty: FAIL - Unable to find /proc/device-tree/model
>
> @Tixy, is it expected to have the device tree model empty?

Probably not, but if this is from an automated testing in Lava then I
believe that doesn't boot with device-tree as UEFI still has bugs with
device-tree on A9.

On Android (what I have handy) then cat /proc/device-tree/model
gives V2P-CA9 on A9 and V2P-CA15_CA7 on TC2.

I can confirm that there is still an outstanding bug in UEFI the prevents us from booting with device tree:

https://bugs.launchpad.net/linaro-landing-team-arm/+bug/1067784