Re: [PATCH v3 5/7] riscv: Kconfig.socs: Allow SOC_CANAAN with MMU for K230

From: Damien Le Moal <dlemoal@kernel.org>
To: Yangyu Chen <cyy@cyyself.name>
Cc: Conor Dooley <conor@kernel.org>,
	linux-riscv@lists.infradead.org, Rob Herring <robh+dt@kernel.org>,
	Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>,
	Paul Walmsley <paul.walmsley@sifive.com>,
	Palmer Dabbelt <palmer@dabbelt.com>,
	Albert Ou <aou@eecs.berkeley.edu>, Guo Ren <guoren@kernel.org>,
	devicetree@vger.kernel.org, linux-kernel@vger.kernel.org,
	Conor Dooley <conor.dooley@microchip.com>
Subject: Re: [PATCH v3 5/7] riscv: Kconfig.socs: Allow SOC_CANAAN with MMU for K230
Date: Fri, 15 Mar 2024 17:19:04 +0900	[thread overview]
Message-ID: <2b0511af-1b5b-4c90-a673-c9113bb58142@kernel.org> (raw)
In-Reply-To: <tencent_9543398AE3BAC424842837CEF01508D6680A@qq.com>

On 3/14/24 01:56, Yangyu Chen wrote:
>> Hi,
>>
>> Thanks for the review comments. After thinking about it for a while,
>> I think we don't need to change it as we have changed the help
>> message which deleted the "K210". And the dts on k210.dtsi shows
>> mmu-type is riscv.none, I think if someone noticed this would know
>> why it fails to boot on the S-Mode MMU Kernel on K210. The only
>> special thing for ARCH_CANAAN is that a loader.bin will be built
>> when M-Mode is on arch/riscv/Makefile. However, Canaan has no other
>> M-Mode chips except for K210. So I think we don't need to change
>> it.

You completely lost me here. I do not understand what you are trying to say.

>> Another reason is that SOC_CANAAN for K210 is somehow hard to change.
>> If we continue using SOC_CANAAN for K210 but not for other Canaan
>> SoCs such as K230, it will cause some confusion to users. If we
>> rename SOC_CANAAN to SOC_CANAAN_K210, it will change many drivers
>> in many subsystems like my patch v5 [1]. So I don't think we need
>> to fix it.
>>
>>
>> If we don't change it, A concern for this is that some drivers for
>> K210 will be built when SOC_CANAAN=y and if we add this to defconfig,
>> all riscv builds will also build some K210 drivers even on MMU. But
>> I think this will not be a problem just need some memory/storage
>> for a slightly bigger kernel. Also, we will enable some new configs
>> in defconfig when a new soc gets supported, it's normal for K210
>> SoC drivers.
>>
>> Thus, I think we don't need to change it. If you have some other
>> opinions, please let me know.

1) Rename SOC_CANAAN to SOC_CANAAN_K210 and use that for any conditional code or
driver selection that is specific to the K210, which is what's done now.
2) Create a "new" SOC_CANAAN config and make SOC_CANAAN_K210 depend on it and on
!MMU

You could also add SOC_CANAAN_K230 if needed and make it depend on SOC_CANAAN &&
MMU.

With that, dirvers common to both the K210 and K230 can easilly be selected and
selecting SOC_CANAAN will end up either building for the K230 or the K210,
depending on MMU being set or not.

That's my 2 cents. Feel free to ignore. I am not involved that much with riscv
these days and I am far too busy with other kernel areas to be of any help. But
I really think that allowing building the K210 when MMU is enabled and "hoping
that the user understand his mistake" is not a great approach.

-- 
Damien Le Moal
Western Digital Research

_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv