[U-Boot] [PATCH 1/1] sandbox: dt: sandbox.dts set skip-localhost = <1>

[Heinrich Schuchardt]

On 10/16/2018 06:42 AM, Joe Hershberger wrote:
> Hi Heinrich,
> 
> On Mon, Oct 15, 2018 at 7:20 PM Heinrich Schuchardt <xypron.glpk at gmx.de> wrote:
>>
>> On 10/15/2018 11:52 PM, Joe Hershberger wrote:
>>> On Sun, Oct 14, 2018 at 2:27 PM Heinrich Schuchardt <xypron.glpk at gmx.de> wrote:
>>>>
>>>> The local interface is not usable for many network operations. It has been
>>>> disabled in all sandbox device trees except sandbox.dts. Let's disable it
>>>> here too.
>>>>
>>>> 'bootefi selftest' tries to do execute DHCP.  This fails on the lo device.
>>>
>>> Shouldn't the tests be using test.dts?
>>
>> bootefi selftest is a command inside U-Boot. It does not choose a device
>> tree.
>>
>> This is for hand testing, where
>>> it can be available, and people can use it or not by specifying
>>> ethact.
>>>
>>
>> make sandbox_defconfig
>> make
>> ./u-boot -D
>>
>> uses sandbox.dts.
> 
> Sure... and ./u-boot -d test/dm/test.dtb uses that device tree. Or any
> other you choose to build.
>>
>> For all other sandbox*.dts we also set skip-localhost=1. Why should
>> sandbox_dts be inconsistent with those?
> 
> The only one that I may consider inconsistent is sandbox64.dts (I
> assume that is what you mean when you say "all other", you are talking
> about this one file, right?), and I'm not sure where it's used. I've
> never used it that I remember. Unless it's magically picked somehow
> based on something environmental.
> 
> You seem to think that if you don't force the local interface to not
> show up that you have to use it. Instead you can just set the ethact
> U-Boot environment variable to the appropriate interface you want to
> use, and off you go.
> 
>> Why would I want to use the cloned local interface?
> 
> So that you could test against, for instance, a TFTP server on your
> host machine.
> 
>> board/sandbox/README.sandbox does not even mention test.dts. The only
>> reference is in a Python test, see tools/binman/ftest.py.
> 
> You'll notice the message in test/dm/test-main.c that claims the
> requirement to use test.dts.
> 
>> The network devices defined in test.dts do not allow to access the real
>> world. So why should I use it?
> 
> Again, that is for the dm unit tests, not hand testing.
> 
> Cheers,
> -Joe

Thanks Joe for your valuable feedback.

I have set the patch to rejected.

Best regards

Heinrich