linux-kernel - Re: [PATCH 2/2] dt-bindings: clock: intel,cgu-lgm: add mxl,control-gate option

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Tue, 01 Aug 2023 10:09:50 +0200
From:   Florian Eckert <fe@....tdt.de>
To:     Krzysztof Kozlowski <krzysztof.kozlowski@...aro.org>
Cc:     mturquette@...libre.com, sboyd@...nel.org, yzhu@...linear.com,
        rtanwar@...linear.com, robh+dt@...nel.org,
        krzysztof.kozlowski+dt@...aro.org, conor+dt@...nel.org,
        linux-clk@...r.kernel.org, linux-kernel@...r.kernel.org,
        devicetree@...r.kernel.org, Eckert.Florian@...glemail.com
Subject: Re: [PATCH 2/2] dt-bindings: clock: intel,cgu-lgm: add
 mxl,control-gate option

Hello Krzysztof,

>>> You described the desired Linux feature or behavior, not the actual
>>> hardware. The bindings are about the latter, so instead you need to
>>> rephrase the property and its description to match actual hardware
>>> capabilities/features/configuration etc.
>> 
>> You have correctly identified that this is not a hardware 
>> configuration,
>> but a driver configuration. Currently, the driver is configured so 
>> that
>> the gates cannot be switched via the clk subsystem callbacks. When
>> registering the data structures from the driver, I have to pass a flag
>> GATE_CLK_HW so that the gate is managed by the driver.
>> 
>> I didn't want to always change the source of the driver when it has to
>> take
>> care of the GATE, so I wanted to map this via the dts.
>> 
>> I have a board support package from Maxlinear for the Lightning 
>> Mountain
>> Soc
>> with other drivers that are not upstream now. Some of them use the
>> clock framework some of them does not.
>> 
>> Due to missing documents it is not possible to send these drivers
>> upstream.
> 
> So when you upstream them, the binding becomes wrong or not needed?
> Sorry, bindings are entirely independent of OS, so using this as an
> argument is clear no-go.

Yes, that would probably be the case, as the maxlinear drivers are at
an early stage and are not yet upstreamable in my opinion. If I had the
documents, I would take a closer look. But they are developing behind
closed doors. Nothing can be contributed. Not until the drivers are
hopefully upstream at some point as the cgu-lgm.

>> Strictly speaking, this is about the gptc and the watchdog.
>> 
>> Since it is a buildin_platform driver, it can also not work via
>> module parameters.
> 
> None of this explains any hardware related part of this binding. You
> created now policy for one specific OS. Devicetree, which is OS
> independent, is not for such purposes.

Yes this would be the case. Maybe I need to patch the cgu-lgm.c [1]
and send it upstream to restore the old behavior.
Because the following commit has changed the behaviour [2].
Unfortunately, it is also included in 5.15 stable branch.
Which in my opinion should not have happened!

Best regards,
Florian

[1] 
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/clk/x86/clk-lgm.c?h=v6.5-rc4
[2] 
https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/drivers/clk/x86/clk-cgu.c?h=v5.15.123&id=a0583edea4fdb7b5b87a077263dddab476e9f138