linux-kernel - Re: [PATCH 15/15] arm64: dts: qcom: sm6115: Add VDD_CX to GPU

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]

Message-ID: <8c342c2b-db37-1360-c716-0e4cdf3809a7@linaro.org>
Date:   Tue, 18 Jul 2023 15:36:35 +0200
From:   Konrad Dybcio <konrad.dybcio@...aro.org>
To:     Dmitry Baryshkov <dmitry.baryshkov@...aro.org>
Cc:     Stephan Gerhold <stephan@...hold.net>,
        Bjorn Andersson <andersson@...nel.org>,
        Andy Gross <agross@...nel.org>,
        Michael Turquette <mturquette@...libre.com>,
        Stephen Boyd <sboyd@...nel.org>,
        Rob Herring <robh+dt@...nel.org>,
        Krzysztof Kozlowski <krzysztof.kozlowski+dt@...aro.org>,
        Conor Dooley <conor+dt@...nel.org>,
        Marijn Suijten <marijn.suijten@...ainline.org>,
        linux-arm-msm@...r.kernel.org, linux-clk@...r.kernel.org,
        linux-kernel@...r.kernel.org, devicetree@...r.kernel.org
Subject: Re: [PATCH 15/15] arm64: dts: qcom: sm6115: Add VDD_CX to GPU_CCC

On 18.07.2023 15:08, Dmitry Baryshkov wrote:
> On Tue, 18 Jul 2023 at 15:48, Konrad Dybcio <konrad.dybcio@...aro.org> wrote:
>>
>> On 18.07.2023 13:56, Stephan Gerhold wrote:
>>> On Mon, Jul 17, 2023 at 09:18:21PM +0200, Konrad Dybcio wrote:
>>>> On 17.07.2023 19:23, Stephan Gerhold wrote:
>>>>> On Mon, Jul 17, 2023 at 07:11:33PM +0200, Konrad Dybcio wrote:
>>>>>> On 17.07.2023 18:56, Stephan Gerhold wrote:
>>>>>>> On Mon, Jul 17, 2023 at 06:50:18PM +0200, Konrad Dybcio wrote:
>>>>>>>> On 17.07.2023 18:28, Stephan Gerhold wrote:
>>>>>>>>> On Mon, Jul 17, 2023 at 05:19:22PM +0200, Konrad Dybcio wrote:
>>>>>>>>>> The GPU_CC block is powered by VDD_CX. Describe that.
>>>>>>>>>>
>>>>>>>>>> Signed-off-by: Konrad Dybcio <konrad.dybcio@...aro.org>
>>>>>>>>>> ---
>>>>>>>>>>  arch/arm64/boot/dts/qcom/sm6115.dtsi | 2 ++
>>>>>>>>>>  1 file changed, 2 insertions(+)
>>>>>>>>>>
>>>>>>>>>> diff --git a/arch/arm64/boot/dts/qcom/sm6115.dtsi b/arch/arm64/boot/dts/qcom/sm6115.dtsi
>>>>>>>>>> index 29b5b388cd94..bfaaa1801a4d 100644
>>>>>>>>>> --- a/arch/arm64/boot/dts/qcom/sm6115.dtsi
>>>>>>>>>> +++ b/arch/arm64/boot/dts/qcom/sm6115.dtsi
>>>>>>>>>> @@ -1430,6 +1430,8 @@ gpucc: clock-controller@...0000 {
>>>>>>>>>>                        clocks = <&rpmcc RPM_SMD_XO_CLK_SRC>,
>>>>>>>>>>                                 <&gcc GCC_GPU_GPLL0_CLK_SRC>,
>>>>>>>>>>                                 <&gcc GCC_GPU_GPLL0_DIV_CLK_SRC>;
>>>>>>>>>> +                      power-domains = <&rpmpd SM6115_VDDCX>;
>>>>>>>>>> +                      required-opps = <&rpmpd_opp_low_svs>;
>>>>>>>>>
>>>>>>>>> Where is this required-opp coming from? The clocks in gpucc seem to have
>>>>>>>>> different voltage requirements depending on the rates, but we usually
>>>>>>>>> handle that in the OPP tables of the consumer.
>>>>>>>> The only lower levels defined for this SoC are VDD_MIN and VDD_RET,
>>>>>>>> but quite obviously the GPU won't work then
>>>>>>>>
>>>>>>>
>>>>>>> The levels needed for the GPU clocks to run should be in the GPU OPP
>>>>>>> table though, just like e.g. sdhc2_opp_table for the SDCC clocks.
>>>>>>>
>>>>>>> I still don't really understand why this is specified here. :)
>>>>>> The GPU_CC block needs this rail to be at a certain power level for
>>>>>> register access. This describes that requirement.
>>>>>>
>>>>>
>>>>> Can you show where this is defined downstream? On a quick look I didn't
>>>>> see something like that anywhere. Or is this from some secret
>>>>> documentation?
>>>> As far as downstream goes, you can notice that no branch's or RCG's
>>>> vdd tables ever define a level lower than the one I mentioned.
>>>>
>>>
>>> As far as I can tell the vdd tables are only used when the clock is
>>> actually enabled though, not for writing to registers while they are
>>> disabled.
>> Maybe, but you can also notice that even XO rates require at least
>> SVS_LOW to function.
> 
> But the vdd tables are related to clock rates, which, in the upstream
> design, should be voted by the consumers, not by the clock driver.
Not all of the clocks are associated with OPP tables upstream, and it
would be nice if the clock controller block had power flowing to it
in case one wanted to access a different clock.

Konrad