Re: [PATCH v2 0/4] regulator: spacemit-p1: Fix voltage ranges and support board power tree

From: Vivian Wang <wangruikang@iscas.ac.cn>
To: Yixun Lan <dlan@gentoo.org>, Guodong Xu <guodong@riscstar.com>
Cc: Liam Girdwood <lgirdwood@gmail.com>,
	Mark Brown <broonie@kernel.org>, Alex Elder <elder@riscstar.com>,
	Lee Jones <lee@kernel.org>, Rob Herring <robh@kernel.org>,
	Krzysztof Kozlowski <krzk+dt@kernel.org>,
	Conor Dooley <conor+dt@kernel.org>,
	Troy Mitchell <troy.mitchell@linux.spacemit.com>,
	Paul Walmsley <pjw@kernel.org>,
	Palmer Dabbelt <palmer@dabbelt.com>,
	Albert Ou <aou@eecs.berkeley.edu>,
	Alexandre Ghiti <alex@ghiti.fr>,
	linux-kernel@vger.kernel.org, linux-riscv@lists.infradead.org,
	spacemit@lists.linux.dev, devicetree@vger.kernel.org
Subject: Re: [PATCH v2 0/4] regulator: spacemit-p1: Fix voltage ranges and support board power tree
Date: Sun, 25 Jan 2026 21:02:42 +0800	[thread overview]
Message-ID: <b705ceed-3ee1-49ec-8f8e-e885e87f871e@iscas.ac.cn> (raw)
In-Reply-To: <20260125110333-GYD71302@gentoo.org>

On 1/25/26 19:03, Yixun Lan wrote:
> Hi Guodong,
>
> On 12:27 Sun 25 Jan     , Guodong Xu wrote:
>> On Sun, Jan 25, 2026 at 12:18 PM Guodong Xu <guodong@riscstar.com> wrote:
>>> On Sat, Jan 24, 2026 at 2:25 PM Vivian Wang <wangruikang@iscas.ac.cn> wrote:
>>>>
>>>> On 1/24/26 08:20, Guodong Xu wrote:
>>>>> [...]
>>>>>
>>>>> Note: Patch 3 introduces a bisect breakage by transitioning to
>>>>> pin-specific supply names. Probe failures will occur on existing boards
>>>>> until Patch 4 updates the corresponding DTS file.
>>>> Ouch, that's not a bisect breakage, that's an *ABI breakage*. And AFAICT
>>>> this is still not okay in 2026,
>>>> see Documentation/devicetree/bindings/ABI.rst
>>>>
>>>> So the bindings would need to be changed to accept both the new and old way.
>>> Ideally yes. However, considering this ABI change's actual effect, the two
>>> K1 boards (BPI-F3 and Jupiter) in the kernel get their power settings
>>> from boot firmware as well, and the types of peripherals enabled in the .dts
>>> files are very limited, the probe failure of the pmic regulator doesn't
>>> affect much. So, I think this breakage is acceptable.
>>>
>>>> Driver-wise, at a cursory look from someone not familiar with the
>>>> regulator stuff, maybe we can make it compatible with old DTS by adding
>>>> the new names as aliases ({devm_,}regulator_register_supply_alias?) as
>>>> "vin" or "buck5", if we see the old vin-supply definitions?
>>>>
>>> We can do that of course. My hesitation is, however, it makes the driver take
>>> extra code which may not be needed once all .dts files have been updated. The
>>> driver code will be left there forever.
>>>
>> Mark gave his opinion in v1 review [1], please allow me to partially quote
>> here: "(it's an ABI change so shouldn't really happen, but perhaps there are
>> few enough users for everyone to coordinate and it's what you all prefer)."
>>
>> I do expect to collect more ideas before I decide whether and what to do in
>> v3, or maybe v3 is not required.
>>
> As I checked the dts tree (DT queued for v6.20), although we introduced the
> regulator of P1/PMIC, but there is no consumers so far, so in real life, we
> shouldn't break anything. In this case, I'd suggest we just give up for doing
> the ABI backward compatible work which should simplify our life..

Having checked again, I agree that this is not that big of a problem.
The breakage with old DT is limited to an otherwise harmless error on
boot that doesn't affect functionality since there are no users.

Vivian "dramforever" Wang.

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