On Thu, Apr 11, 2024 at 03:57:16AM -0400, Sasha Levin wrote:
On Thu, Apr 11, 2024 at 09:34:39AM +0200, Greg KH wrote:
On Thu, Apr 11, 2024 at 09:27:28AM +0200, Krzysztof Kozlowski wrote:
On 11/04/2024 09:22, Greg KH wrote:
On Wed, Apr 10, 2024 at 08:24:49PM +0200, Krzysztof Kozlowski wrote:
On 10/04/2024 20:02, Greg KH wrote:
On Wed, Apr 10, 2024 at 07:58:40PM +0200, Konrad Dybcio wrote: > > > On 4/10/24 17:57, Sasha Levin wrote: >> This is a note to let you know that I've just added the patch titled >> >> arm64: dts: qcom: Add support for Xiaomi Redmi Note 9S > > autosel has been reeaaaaaly going over the top lately, particularly > with dts patches.. I'm not sure adding support for a device is > something that should go to stable
Simple device ids and quirks have always been stable material.
That's true, but maybe DTS should have an exception. I guess you think this is trivial device ID, because the patch contents is small. But it is or it can be misleading. The patch adds new small DTS file which includes another file:
#include "sm7125-xiaomi-common.dtsi"
Which includes another 7 files:
#include <dt-bindings/arm/qcom,ids.h> #include <dt-bindings/firmware/qcom,scm.h> #include <dt-bindings/gpio/gpio.h> #include <dt-bindings/regulator/qcom,rpmh-regulator.h> #include "sm7125.dtsi" #include "pm6150.dtsi" #include "pm6150l.dtsi"
Out of which last three are likely to be changing as well.
This means that following workflow is reasonable and likely:
- Add sm7125.dtsi (or pm6150.dtsi or pm6150l.dtsi)
- Add some sm7125 board (out of scope here).
- Release new kernel, e.g. v6.7.
- Make more changes to sm7125.dtsi
- The patch discussed here, so one adding sm7125-xiaomi-curtana.dts.
Now if you backport only (5) above, without (4), it won't work. Might compile, might not. Even if it compiles, might not work.
The step (4) here might be small, but might be big as well.
Fair enough. So should we drop this change?
I vote for dropping. Also, I think such DTS patches should not be picked automatically via AUTOSEL. Manual backports or targetted Cc-stable, assuming that backporter investigated it, seem ok.
Sasha now dropped this, thanks.
Sasha, want to add dts changes to the AUTOSEL "deny-list"?
Sure, this makes sense.
Does it? Seems like a rather big hammer to me. I totally understand blocking the addition of new dts files to stable, but there's a whole load of different people maintaining dts files with differing levels of remembering to cc stable explicitly.
That said, often a dts backport depends on a driver (or binding) change too, so backporting one without the other may have no effect. I have no idea whether or not AUTOSEL is capable of picking out those sort of dependencies.