lkml.org 
[lkml]   [2023]   [Mar]   [20]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectRe: [PATCH 11/15] arm64: dts: nuvoton: Add initial ma35d1 device tree
From
Dear Arnd,


On 2023/3/18 下午 10:04, Arnd Bergmann wrote:
> On Sat, Mar 18, 2023, at 14:17, Jacky Huang wrote:
>> On 2023/3/16 下午 10:17, Arnd Bergmann wrote:
>>> On Wed, Mar 15, 2023, at 08:28, Jacky Huang wrote:
>>>> + mem: memory@80000000 {
>>>> + device_type = "memory";
>>>> + reg = <0x00000000 0x80000000 0 0x20000000>; /* 512M DRAM */
>>>> + };
>>>> +};
>>> In most machines, the memory size is detected by the boot loader
>>> and filled in the dtb in memory before starting the kernel, so
>>> you should not need two separate files here for the two common
>>> memory configurations.
>>
>> On ma35d1, memory size is determined early before uboot.
>>
>> BL1 (MaskROM boot code) -> BL2 (arm-trust-firmware) -> BL32 (op-tee) &
>> BL33 (uboot).
>> The DDR was initialized in BL2 stage with a selected DDR setting, which
>> is hard coded, including DDR size.
>>
>> We searched the arm64 dts and found that almost all vendors claimed
>> memory size in board level dtsi/dts. This seems to be common.
>>
>> So, can we have it unchanged?
> I see the memory size encoded in about one out of three .dts files,
> which is more than I expected. It's clearly not harmful to have it
> listed in the dts, it just shouldn't be necessary.
>
> If it helps you with your current u-boot, then leave it in, but
> consider adding detection logic into u-boot so it can override
> the value in the dtb file at boot time.


Thank you for your understanding. As more drivers are added, I think
this memory

size encoded will look less conspicuous. In fact, in the previous arm9
project, we

did detect the memory size by uboot, and then passed it to the kernel.
If there is

a need in the future, we will consider to support it in ma35d1.

>>> Since the machine is called 'som', I would assume that this is a
>>> module that is integrated on another board, so more commonly one
>>> would have a dtsi file for the som in addition to the one for the
>>> soc, and have all the components of the module listed in this
>>> file, while the dts file that includes the som.dtsi lists the
>>> devices on the carrier board and enables the on-chip devices
>>> that are connected to the outside.
>>>
>> You are right, ma35d1 som have a base board, and a cpu board on it.
>>
>> It is a good suggestion that we should have a dtsi for som base board.
>>
>> Consider that we are in the initial submit, and such a dtsi will be an empty
>> file at this stage. So, I would like to do it when peripheral drivers
>> upstream started. Is it ok?
> It's not a big deal either way. I if you want to keep it only with
> one dts file and one dtsi file, that's fine, but maybe rename the dts
> file based on the name of the carrier rather than the SoM in this
> case.
>
> Arnd


Thank you. As the dts names are consistent with the ma35d1 BSP on
linux-5.10.y,

we would like to keep the consistence still.


Best regards,

Jacky Huang


\
 
 \ /
  Last update: 2023-03-27 01:09    [W:0.405 / U:0.744 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site