[PATCH v2 2/2] sandbox: add bootmethod EFI boot-manager

[Heinrich Schuchardt]

On 10/18/24 22:36, Tom Rini wrote:
> On Fri, Oct 18, 2024 at 02:17:36PM -0600, Simon Glass wrote:
>> Hi Heinrich,
>>
>> On Fri, 18 Oct 2024 at 11:26, Heinrich Schuchardt
>> <heinrich.schuchardt at canonical.com> wrote:
>>>
>>> On 10/18/24 19:21, Simon Glass wrote:
>>>> Hi Heinrich,
>>>>
>>>> On Thu, 17 Oct 2024 at 19:18, Heinrich Schuchardt
>>>> <heinrich.schuchardt at canonical.com> wrote:
>>>>>
>>>>> The EFI boot-manager is the default method to boot EFI binaries.
>>>>> We should be able to use it on the Sandbox.
>>>>>
>>>>> * Enable EFI boot-manager bootmethod on the sandbox.
>>>>> * Adjust unit tests to reflect the additional boot method.
>>>>>
>>>>> Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt at canonical.com>
>>>>> Reviewed-by: Simon Glass <sjg at chromium.org>
>>>>> Reviewed-by: Mattijs Korpershoek <mkorpershoek at baylibre.com>
>>>>> ---
>>>>> v2:
>>>>>           no changes
>>>>> ---
>>>>>    arch/sandbox/dts/test.dts |  4 ++++
>>>>>    test/boot/bootmeth.c      | 28 +++++++++++++++++-----------
>>>>>    2 files changed, 21 insertions(+), 11 deletions(-)
>>>>
>>>> This is better, but I still see this.
>>>>
>>>> => bootflow scan
>>>>        efi_var_to_file() Cannot persist EFI variables without system partition
>>>>
>>>> Please find a way to get rid of it. I can give some suggestions if you like.
>>>>
>>>
>>> Would you, please, mind providing a rationale why you want this correct
>>> message not to be written.
>>>
>>> I expect the sandbox to behave as any other board and provide warnings
>>> when they are adequate.
>>
>> Well, what I am supposed to do to fix this warning? I just don't want
>> unactionable messages coming up in bootstd.

The user runs a system that tries to boot via EFI. He does not have an 
block device with an ESP. For sure that is an actionable warning:

The user should insert properly formatted media.

And your test should put an ESP on the image.

>>
>> If it happens all the time and cannot be fixed, you could add an
>> option to the driver to ignore it, which can be controlled from the
>> devicetree node.

As said it can be easily fixed. Create an ESP on one of the block devices.

Do you really expect a user to change the device-tree to suppress a warning?

Wouldn't it be more plausible to remove the EFI boot method from the 
bootflow instead if that device is never to be booted via EFI?

> 
> So, you're scanning a disk image to boot from, and it lacks a system
> partition. Can you describe how you're creating this disk image? And are
> you testing something EFI related, or not? And then Heinrich, can we be
> checking for variables any later than we do today?
> 

The EFI specification requires us to create non-volatile BOOT#### 
variables for each block device. Without an ESP we cannot persist these. 
We are already at the latest possible moment when we reach the boot manager.

Best regards

Heinrich