Received-SPF: pass (google.com: domain of linux-kernel+bounces-93505-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
Message-ID: <83fd1995-a06e-b76a-d91b-de1c1a6ab0ea@quicinc.com>
Date: Wed, 6 Mar 2024 14:00:56 +0530
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:91.0) Gecko/20100101
 Thunderbird/91.9.0
Subject: Re: [PATCH 4/5] clk: qcom: Add camera clock controller driver for
 SM8150
Content-Language: en-US
To: Bryan O'Donoghue <bryan.odonoghue@linaro.org>,
        Bjorn Andersson
	<andersson@kernel.org>,
        Konrad Dybcio <konrad.dybcio@linaro.org>,
        "Michael
 Turquette" <mturquette@baylibre.com>,
        Stephen Boyd <sboyd@kernel.org>, Abhishek Sahu <absahu@codeaurora.org>,
        Rob Herring <robh@kernel.org>,
        Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>,
        Conor Dooley
	<conor+dt@kernel.org>
CC: Stephen Boyd <sboyd@codeaurora.org>, <linux-arm-msm@vger.kernel.org>,
        <linux-clk@vger.kernel.org>, <linux-kernel@vger.kernel.org>,
        <devicetree@vger.kernel.org>, Ajit Pandey <quic_ajipan@quicinc.com>,
        "Imran
 Shaik" <quic_imrashai@quicinc.com>,
        Taniya Das <quic_tdas@quicinc.com>,
        Jagadeesh Kona <quic_jkona@quicinc.com>
References: <20240229-camcc-support-sm8150-v1-0-8c28c6c87990@quicinc.com>
 <20240229-camcc-support-sm8150-v1-4-8c28c6c87990@quicinc.com>
 <18567989-fb60-49ae-92e6-94e1bc2fa1c7@linaro.org>
From: "Satya Priya Kakitapalli (Temp)" <quic_skakitap@quicinc.com>
In-Reply-To: <18567989-fb60-49ae-92e6-94e1bc2fa1c7@linaro.org>
Content-Type: text/plain; charset="UTF-8"; format=flowed
Content-Transfer-Encoding: 8bit


On 3/2/2024 9:43 PM, Bryan O'Donoghue wrote:
> On 29/02/2024 5:38 a.m., Satya Priya Kakitapalli wrote:
>> Add support for the camera clock controller for camera clients
>> to be able to request for camcc clocks on SM8150 platform.
>>
>> Signed-off-by: Satya Priya Kakitapalli <quic_skakitap@quicinc.com>
>> ---
>
>> +static int cam_cc_sm8150_probe(struct platform_device *pdev)
>> +{
>> +    struct regmap *regmap;
>> +    int ret;
>> +
>> +    ret = devm_pm_runtime_enable(&pdev->dev);
>> +    if (ret)
>> +        return ret;
>> +
>> +    ret = pm_runtime_resume_and_get(&pdev->dev);
>> +    if (ret)
>> +        return ret;
>> +
>> +    regmap = qcom_cc_map(pdev, &cam_cc_sm8150_desc);
>> +    if (IS_ERR(regmap)) {
>> +        pm_runtime_put(&pdev->dev);
>> +        return PTR_ERR(regmap);
>> +    }
>> +
>> +    clk_trion_pll_configure(&cam_cc_pll0, regmap, &cam_cc_pll0_config);
>> +    clk_trion_pll_configure(&cam_cc_pll1, regmap, &cam_cc_pll1_config);
>> +    clk_regera_pll_configure(&cam_cc_pll2, regmap, 
>> &cam_cc_pll2_config);
>> +    clk_trion_pll_configure(&cam_cc_pll3, regmap, &cam_cc_pll3_config);
>> +    clk_trion_pll_configure(&cam_cc_pll4, regmap, &cam_cc_pll4_config);
>> +
>> +    /* Keep the critical clock always-on */
>> +    qcom_branch_set_clk_en(regmap, 0xc1e4); /* cam_cc_gdsc_clk */
>
> Does this clock need to be specified this way ?
>

Yes, we need this clock to be always on.


> drivers/clk/qcom/camcc-sc8280xp.c::camcc_gdsc_clk specifies the gdsc 
> clock as a shared op clock.
>
> Actually it looks to be register compatible, please try defining 
> titan_top_gdsc as per the example in 8280xp.
>> +
>> +    ret = qcom_cc_really_probe(pdev, &cam_cc_sm8150_desc, regmap);
>> +
>> +    pm_runtime_put(&pdev->dev);
>> +
>> +    return ret;
>> +}
>
> So this is a pattern we keep repeating in the clock probe() functions 
> which I am writing a series to address. There's no need to continue to 
> replicate the bug in new code though.
>
> Only switch on always-on clocks if probe succeeds.
>
>     ret = qcom_cc_really_probe(pdev, &cam_cc_sm8150_desc, regmap);
>     if (ret)
>         goto probe_err;
>
>     qcom_branch_set_clk_en(regmap, 0xc1e4); /* cam_cc_gdsc_clk */
>
>     pm_runtime_put(&pdev->dev);
>
>     return 0;
>
> probe_err:
>     pm_runtime_put_sync(&pdev->dev);
>
> Alternatively switch on the always-on clocks before the really_probe() 
> but then roll back in a probe_err: goto
>
> probe_err:
>     remap_bits_update(regmap, 0xc1e4, BIT(0), 0);
>     pm_runtime_put_sync(&pdev->dev);
>
> There may be corner cases where always-on has to happen before 
> really_probe() I suppose but as a general pattern the above should be 
> how we go.
>
> Anyway I suspect the right thing to do is to define a 
> titan_top_gdsc_clk with shared ops to "park" the GDSC clock to 19.2 
> MHz instead of turning it off.
>
> You can get rid of the hard-coded always-on and indeed represent the 
> clock in /sysfs - which is preferable IMO to just whacking registers 
> to keep clocks always-on in probe anyway.
>
> Please try to define the titan_top_gdsc_clk as a shared_ops clock 
> instead of hard coding to always on.
>

Defining the gdsc clk allows consumers to control it, we do not want 
this clock to be disabled/controlled from consumers. Hence it is better 
to not model this clock and just keep it always on from probe.


> If that doesn't work for some reason, then please fix your always-on 
> logic in probe() to only make the clock fixed on, if really_probe() 
> succeeds.
>

Sure I'll do this.


> ---
> bod