[U-Boot] Fwd: [PATCH] Initial fw_cfg support for qemu-x86 targets.

[Saket Sinha]

Hi Mian,

This was the initial attempt to add fw_cfg interface support for
qemu-x86 targets to mainly support ACPI tables.

The patches were dropped due to the issues discussed in the thread.

If we can collaborate the patches in the sense that we are able to
load ACPI tables(and other tables) successfully, it would be great.


Regards,
Saket Sinha



---------- Forwarded message ----------
From: Bin Meng <bmeng.cn at gmail.com>
Date: Fri, Jun 5, 2015 at 2:36 PM
Subject: Re: [PATCH] Initial fw_cfg support for qemu-x86 targets.
To: Saket Sinha <saket.sinha89 at gmail.com>
Cc: Simon Glass <sjg at chromium.org>


Hi Saket,

On Fri, Jun 5, 2015 at 4:50 PM, Saket Sinha <saket.sinha89 at gmail.com> wrote:
> Hi Bin,
>
>>>  Error: Hardware did not enter ACPI mode
>>>  Unable to enable ACPI
>>>  Interpreter disabled.
>>>
>>> What can be the reason for these errors ?
>>
>> Sorry I have no idea. I think you have to read the whole ACPI
>> specification to understand how the ACPI is used to be the interface
>> between OS and the underlying hardware, and that's your primary GSoC
>> job I believe :)
>>
>
> Sure. I have been going through the ACPI specification.
> Hope to find how to debug this error soon.
>
> Would it be reasonable to send this patch to mainline now, with the
> present fw_cfg support that we have got working so far?
>

I would recommend we investigate the kernel ACPI error first and see
what's going on there. Also I noticed that the QEMU fw_cfg ACPI table
is huge (exactly 128KB). But lots of its memory space is just zero. I
think this needs to be investigated. Also before sending patches to
mainline, you need do some house keeping like fixing up all these
camel case variables/structure names.

> The next thing I am looking forward to is adding builtin ACPI support
> for qemu targets in u-boot.
>
> If getting tables from fw_cfg fails, then all these ACPI tables would
> be written.
> Since it would fail for all non-qemu targets, this code snipet would
> also apply for minnowboard and other x86 targets.
>
> Something like this -
>         fw = fw_cfg_acpi_tables(current);
>         if (fw)
>                 return fw;
>         else
>                 write_acpi_tables(current);

Please follow the same way we write PIRQ/SFI tables. See
arch/x86/Kconfig for a Kconfig option control.

> I would deliberatley get it failed for qemu targets to see if builtin
> ACPI table works.
> Once these builtin table start working on qemu, it should work on
> minnowmax too with little or no changes.
>

Regards,
Bin