[PATCH 0/12] EFI: drop arch efi

[Jean-Christophe PLAGNIOL-VILLARD]

> On 13 Mar 2017, at 6:27 PM, Michael Olbrich <m.olbrich at pengutronix.de> wrote:
> 
> On Sun, Mar 12, 2017 at 01:05:38PM +0100, Jean-Christophe PLAGNIOL-VILLARD wrote:
>> On 17:26 Sat 11 Mar     , Michael Olbrich wrote:
>>> On Wed, Feb 15, 2017 at 08:29:32PM +0100, Jean-Christophe PLAGNIOL-VILLARD wrote:
>>>> Today the EFI is build as an other ARCH when in fact it's just a boot mode
>>>> 
>>>> so move it back to arch/x86 for the spicific x86 part and the common
>>>> to common and driver
>>>> 
>>>> The last 3 patches allow to debug EFI and prepare for more efi support
>>>> 
>>>> The following changes since commit b225bbf295c92263adbcec2c385c5e8d83542c72:
>>>> 
>>>>  i.MX: esdhc: fix imx-esdhc driver for non-OF boards (2017-02-14 08:48:34 +0100)
>>>> 
>>>> are available in the git repository at:
>>>> 
>>>>  git://git.jcrosoft.org/barebox.git
>>>> 
>>>> for you to fetch changes up to df12536f83b485c9a4ee28f78637356c6abab6f6:
>>>> 
>>>>  efi: add efi handle dump command (2017-02-15 07:32:24 +0800)
>>> 
>>> While testing the serial stuff I noticed that the
>>> CONFIG_ARCH_EFI_REGISTER_COM1 option got lost during this rework.
>> I did not merge it as it's a hack more than a feature as this by pass the EFI
>> API and access the hw directly instead of using Serial IO
>> 
>> Which was add after
>> 
>> And I plan to drop this hack
> 
> You cannot assume a sane EFI. I have a hardware with a serial port that
> works just fine with the ns16550 driver but the EFI does not implement the
> serial IO protocol. So this hack is still needed in some way.
> 

I agree with you on that EFI implementation are erradict at best.

> Maybe we could register the ns16550 console if no efi serial IO protocol is
> found?

I’d prefer we use a DT to add more device at build time or runtime that hack in the 
general code.

Or Create a specific board code for you hardware

So we keep the generic code as generic as possible

Best Regards,
J.