[PATCH RESEND 21/62] init: remove all mentions of root=/dev/ram*

[Krzysztof Kozlowski]

On 14/09/2025 22:13, Andrew Morton wrote:
> On Sun, 14 Sep 2025 12:06:24 +0200 Krzysztof Kozlowski <krzk at kernel.org> wrote:
> 
>>>  Documentation/admin-guide/kernel-parameters.txt          | 3 +--
>>>  Documentation/arch/m68k/kernel-options.rst               | 9 ++-------
>>>  arch/arm/boot/dts/arm/integratorap.dts                   | 2 +-
>>>  arch/arm/boot/dts/arm/integratorcp.dts                   | 2 +-
>>>  arch/arm/boot/dts/aspeed/aspeed-bmc-facebook-cmm.dts     | 2 +-
>>>  .../boot/dts/aspeed/aspeed-bmc-facebook-galaxy100.dts    | 2 +-
>>>  .../arm/boot/dts/aspeed/aspeed-bmc-facebook-minipack.dts | 2 +-
>>>  .../arm/boot/dts/aspeed/aspeed-bmc-facebook-wedge100.dts | 2 +-
>>>  arch/arm/boot/dts/aspeed/aspeed-bmc-facebook-wedge40.dts | 2 +-
>>>  arch/arm/boot/dts/aspeed/aspeed-bmc-facebook-yamp.dts    | 2 +-
>>>  .../boot/dts/aspeed/ast2600-facebook-netbmc-common.dtsi  | 2 +-
>>
>> No, don't do that. DTS is always separate.
> 
> Why can't DTS changes be carried in a different tree?


It must be carried in a different kernel tree and it must be ALWAYS a
separate commit. Embedding it in the middle of this patchset and in the
middle of some other commit breaks these two rules.

If you asked why it cannot be carried by VFS (or by any non-SoC tree in
general), it is because DTS is completely independent hardware
description, so by keeping it on separate tree we enforce that rule of
lack of dependency between DTS and any driver or core code.

If there is a dependency here, then it would be a NAK, because there
cannot be such - it would be a breach of contract for outside users (DTS
is shared with other, non-Linux projects).


Best regards,
Krzysztof