Message-ID: <54EC9D66.10001@linaro.org>
Date: Tue, 24 Feb 2015 17:48:54 +0200
From: Georgi Djakov <georgi.djakov@linaro.org>
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:31.0) Gecko/20100101 Thunderbird/31.4.0
MIME-Version: 1.0
To: Stephen Boyd <sboyd@codeaurora.org>, mturquette@linaro.org
CC: linux-kernel@vger.kernel.org, linux-arm-msm@vger.kernel.org
Subject: Re: [PATCH v1] clk: qcom: Add MSM8916 Global Clock Controller support
References: <1423249118-22132-1-git-send-email-georgi.djakov@linaro.org> <54EBADB3.7030005@codeaurora.org>
In-Reply-To: <54EBADB3.7030005@codeaurora.org>
Content-Type: text/plain; charset=windows-1252
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 2679
Lines: 84

On 02/24/2015 12:46 AM, Stephen Boyd wrote:
> On 02/06/15 10:58, Georgi Djakov wrote:
>> [...]
[..]
>> +
>> +	/* Vote for GPLL0 to turn on */
>> +	regmap_read(regmap, 0x45000, &val);
>> +	val |= BIT(0);
>> +	regmap_write(regmap, 0x45000, val);
> 
> Hm.. I guess this is for the CPU to stay on, otherwise GPLL0 might turn
> off? But if we expose this register and the bit as gpll0_vote then we're
> going to turn it off once the last user turns off GPLL0. So I'm not sure
> how to handle this, but it certainly seems like we can just remove this
> bit of code and not be any worse off than we are right now. We need to
> figure out some way to make this work though.
> 

Ok, agree.

> Here's the problem. GPLL0 is used by the CPU running this code. It's
> also used by random other clocks in the system. If the code for the CPU
> clock is not compiled in (i.e. clock-a53.c or whatever we call it), then
> it is very possible that we'll disable the last clock that's using the
> PLL according to what software knows, that will in turn disable the PLL
> and then the CPU will die.
> 
> Of course, it's very possible that we'll never actually turn GPLL0 off
> because it's used for quite a few clocks in the system. So the chances
> of running into this problem are almost entirely theoretical.

Yes, I am investigating this scenario. Thanks for your comment.

> 
>> +
>> +	return qcom_cc_really_probe(pdev, &gcc_msm8916_desc, regmap);
>> +}
>> +
>> +static int gcc_msm8916_remove(struct platform_device *pdev)
>> +{
>> +	qcom_cc_remove(pdev);
>> +	return 0;
>> +}
>> +
>> +static struct platform_driver gcc_msm8916_driver = {
>> +	.probe		= gcc_msm8916_probe,
>> +	.remove		= gcc_msm8916_remove,
>> +	.driver		= {
>> +		.name	= "gcc-msm8916",
> 
> We need owner = THIS_MODULE here because the platform_driver_module
> macro isn't being used.
> 

But below i use the platform_driver_register macro and therefore it seems
not needed.

>> +		.of_match_table = gcc_msm8916_match_table,
>> +	},
>> +};
>> +
>> +static int __init gcc_msm8916_init(void)
>> +{
>> +	return platform_driver_register(&gcc_msm8916_driver);
>> +}
>> +core_initcall(gcc_msm8916_init);
>> +
>> +static void __exit gcc_msm8916_exit(void)
>> +{
>> +	platform_driver_unregister(&gcc_msm8916_driver);
>> +}
>> +module_exit(gcc_msm8916_exit);
>>

Thank you for taking a look at this preliminary stuff. I'll send in the
next few days updated and tested version.

BR,
Georgi
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/