On Thu, Apr 24, 2025 at 08:20:32AM +0000, Philippe Schenker wrote:
On Thu, 2025-04-24 at 09:21 +0200, Francesco Dolcini wrote:
On Thu, Apr 24, 2025 at 08:16:11AM +0200, Wojciech Dubowik wrote:
On Wed, Apr 23, 2025 at 11:23:09AM +0000, Philippe Schenker wrote:
On Wed, 2025-04-23 at 12:21 +0200, Francesco Dolcini wrote:
> > I would backport this to also older kernel, so to me > > Fixes: 6a57f224f734 ("arm64: dts: freescale: add initial > support > for > verdin imx8m mini")
NACK for the proposed Fixes, this introduces a new Kconfig which could have side-effects in users of current stable kernels.
The driver for "regulator-gpio" compatible? I do not agree with your argument, sorry.
The previous description was not correct. There was an unused regulator in the DT that was not switched off just by chance.
Francesco
My previous reasoning about the driver is one point. The other is that the initial implementation in 6a57f224f734 ("arm64: dts: freescale: add initial support for verdin imx8m mini") was not wrong at all it was just different.
My concern is for existing users of stable kernels that you change the underlying implementation of how the SD voltage gets switched. And adding the tag
Fixes: 6a57f224f734 ("arm64: dts: freescale: add initial support for verdin imx8m mini")
to this patch would get this new implementation also to stable kernels not affected by the issue introduced in f5aab0438ef1 ("regulator: pca9450: Fix enable register for LDO5")
I will wait a day and send V4 with what I beleive was result of this discussion.
My opinion is that we should backport this fix. The DT description was wrong, that change on the PMIC driver just made the issue visible, the DT is about the HW description.
From what I understand it was not wrong but it became wrong with the PMIC driver changes.
It was wrong. You had a regulator described as not used, and therefore the OS was free to switch it off. For a bug in the driver this regulator was not switched off by the old kernels.
The DT description was never correct.
With that said, hopefully it makes no difference in pratice, f5aab0438ef17f01c5ecd25e61ae6a03f82a4586 will be backported, and therefore also any fix to it should.
If for some unfortunate reason we do not backport also this one, you'll get a non-working SD card on your LTS kernel.
Francesco