diff mbox series

question: malta doesn't use a1 for fdt from bootloader

Message ID 87plyyrupy.fsf@telent.net (mailing list archive)
State RFC
Headers show
Series question: malta doesn't use a1 for fdt from bootloader | expand

Commit Message

dan@telent.net Dec. 22, 2023, 5:51 p.m. UTC
Newbie question:

I noticed while writing some test automation with QEMU and U-Boot that
the Malta kernel appears to ignore any DTB provided by the bootloader, in
favour of using the one embedded into the kernel image. Is there a
reason behind this, or is it just not a thing that's been changed since
the board was converted to use device tree?

I patched my own kernel to check $a1 as the "mips generic" kernel does,
and it seems to work, but maybe I'm missing something? I did have to
rebuild U-Boot with different CONFIG options to make use of it (I don't
khow what real hardware Malta uses for a bootloader) so perhaps it has
limited use.

If the change is acceptable in principle I'll tidy it up and submit
a patch with all the proper procedure.

Opinions welcome

-dan

Comments

Jiaxun Yang Dec. 25, 2023, 1:33 p.m. UTC | #1
在 2023/12/22 17:51, dan@telent.net 写道:
> 
> Newbie question:
> 
> I noticed while writing some test automation with QEMU and U-Boot that
> the Malta kernel appears to ignore any DTB provided by the bootloader, in
> favour of using the one embedded into the kernel image. Is there a
> reason behind this, or is it just not a thing that's been changed since
> the board was converted to use device tree?

Hi,

It's a bit of history that DeviceTree passed by YAMON bootloader to 
kernel does not compatible with upstream kernel's devicetree bindings,
so we decided to just omit it.

> 
> I patched my own kernel to check $a1 as the "mips generic" kernel does,
> and it seems to work, but maybe I'm missing something? I did have to
> rebuild U-Boot with different CONFIG options to make use of it (I don't
> khow what real hardware Malta uses for a bootloader) so perhaps it has
> limited use.
> 
> If the change is acceptable in principle I'll tidy it up and submit
> a patch with all the proper procedure.

I'm not sure if this is ok for YAMON-DT systems, will check on those 
systems.

Thanks
- Jiaxun
> 
> Opinions welcome
> 
> -dan
> 
> 
> diff --git a/arch/mips/mti-malta/malta-setup.c b/arch/mips/mti-malta/malta-setup.c
> index 21cb3ac1237b..52e731f9b4e2 100644
> --- a/arch/mips/mti-malta/malta-setup.c
> +++ b/arch/mips/mti-malta/malta-setup.c
> @@ -192,7 +192,9 @@ static void __init bonito_quirks_setup(void)
>   
>   void __init *plat_get_fdt(void)
>   {
> -       return (void *)__dtb_start;
> +       return (fw_arg0 == -2) ?
> +               (void *) (KSEG1ADDR(fw_arg1)) :
> +               (void *) __dtb_start;
>   }
>   
>   void __init plat_mem_setup(void)
>
Jiaxun Yang Jan. 3, 2024, 1:50 p.m. UTC | #2
在2023年12月25日十二月 下午1:33,Jiaxun Yang写道:
> 在 2023/12/22 17:51, dan@telent.net 写道:
>> 
>> Newbie question:
>> 
>> I noticed while writing some test automation with QEMU and U-Boot that
>> the Malta kernel appears to ignore any DTB provided by the bootloader, in
>> favour of using the one embedded into the kernel image. Is there a
>> reason behind this, or is it just not a thing that's been changed since
>> the board was converted to use device tree?
>
> Hi,
>
> It's a bit of history that DeviceTree passed by YAMON bootloader to 
> kernel does not compatible with upstream kernel's devicetree bindings,
> so we decided to just omit it.

Hi Dan,

I checked malta YAMON firmware release and can confirm that it is safe
on to do so. These firmwares won't set fw_arg0 to -2.

I think you can make it a patch.

Thanks
- Jiaxun


>
>> 
>> I patched my own kernel to check $a1 as the "mips generic" kernel does,
>> and it seems to work, but maybe I'm missing something? I did have to
>> rebuild U-Boot with different CONFIG options to make use of it (I don't
>> khow what real hardware Malta uses for a bootloader) so perhaps it has
>> limited use.
>> 
>> If the change is acceptable in principle I'll tidy it up and submit
>> a patch with all the proper procedure.
>
> I'm not sure if this is ok for YAMON-DT systems, will check on those 
> systems.
>
> Thanks
> - Jiaxun
>> 
>> Opinions welcome
>> 
>> -dan
>> 
>> 
>> diff --git a/arch/mips/mti-malta/malta-setup.c b/arch/mips/mti-malta/malta-setup.c
>> index 21cb3ac1237b..52e731f9b4e2 100644
>> --- a/arch/mips/mti-malta/malta-setup.c
>> +++ b/arch/mips/mti-malta/malta-setup.c
>> @@ -192,7 +192,9 @@ static void __init bonito_quirks_setup(void)
>>   
>>   void __init *plat_get_fdt(void)
>>   {
>> -       return (void *)__dtb_start;
>> +       return (fw_arg0 == -2) ?
>> +               (void *) (KSEG1ADDR(fw_arg1)) :
>> +               (void *) __dtb_start;
>>   }
>>   
>>   void __init plat_mem_setup(void)
>>
diff mbox series

Patch

diff --git a/arch/mips/mti-malta/malta-setup.c b/arch/mips/mti-malta/malta-setup.c
index 21cb3ac1237b..52e731f9b4e2 100644
--- a/arch/mips/mti-malta/malta-setup.c
+++ b/arch/mips/mti-malta/malta-setup.c
@@ -192,7 +192,9 @@  static void __init bonito_quirks_setup(void)
 
 void __init *plat_get_fdt(void)
 {
-       return (void *)__dtb_start;
+       return (fw_arg0 == -2) ?
+               (void *) (KSEG1ADDR(fw_arg1)) :
+               (void *) __dtb_start;
 }
 
 void __init plat_mem_setup(void)