Re: [PATCH v2] Exynos : Add support for Exynos random number generator

[jonghwa3.lee@samsung.com]

On 2012년 06월 28일 02:52, Stephen Boyd wrote:

> Some minor comments, otherwise this looks much better than the previous
> patch.
> 
> On 06/27/12 03:31, Jonghwa Lee wrote:
>> This patch supports Exynos SOC's PRNG driver. Exynos's PRNG has 5 seeds and
>> 5 random number outputs. Module is excuted under runtime power management control,
>> so it activates only while it's in use. Otherwise it will be suspended generally.
>> It was tested on PQ board by rngtest program.
>>
>> Signed-off-by: Jonghwa Lee <jonghwa3.lee@samsung.com>
>> Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
> 
> This is an incorrect signoff chain. Kyungmin is not sending this so why
> are you not the last one to sign off? Who is the author, Kyungmin or
> yourself?
> 


I'm author. Okay, I'll leave my name only.

>> +
>> +config HW_RANDOM_EXYNOS
>> +	tristate "EXYNOS HW random number generator support"
>> +	depends on HW_RANDOM && ARCH_EXYNOS && HAS_IOMEM && PM_RUNTIME
> 
> There is no need to depend on PM_RUNTIME or ARCH_EXYNOS.
> 
>> +
>> +static int exynos_read(struct hwrng *rng, void *buf,
>> +					size_t max, bool wait)
>> +{
>> +	struct exynos_rng *exynos_rng = container_of(rng,
>> +						struct exynos_rng, rng);
>> +	u32 *data = buf;
>> +
>> +	pm_runtime_get_sync(exynos_rng->dev);
>> +
>> +	exynos_rng_writel(exynos_rng, PRNG_START, 0);
>> +
>> +	do {
>> +		cpu_relax();
>> +	} while (!(exynos_rng_readl(exynos_rng,
>> +			EXYNOS_PRNG_STATUS_OFFSET) & PRNG_DONE));
>> +
>> +	exynos_rng_writel(exynos_rng, PRNG_DONE, EXYNOS_PRNG_STATUS_OFFSET);
> 
> Curious, is this actually required? You poll for the status to say done
> and the hardware requires you to write back the done bit after it
> signals done?
> 


Yes, It's hardware's own characteristic. It needs to be written 1 on
status register to clear it.

>> +
>> +	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
>> +	if (!res) {
>> +		clk_put(exynos_rng->clk);
>> +		return -ENODEV;
>> +	}
> 
> Pass this through directly to devm_request_and_ioremap() without
> checking the return value to save some lines.
> 

>> +

>> +	exynos_rng->mem = devm_request_and_ioremap(&pdev->dev, res);
>> +	if (!exynos_rng->mem) {
>> +		dev_err(&pdev->dev, "Ioremap failed.\n");
> 
> devm_request_and_ioremap() already prints a message on failure to remap
> so this is unnecessary printk.
> 

>> +		return -EBUSY;

>> +	}
>> +
>> +	platform_set_drvdata(pdev, exynos_rng);
>> +
>> +	pm_runtime_enable(&pdev->dev);
>> +
>> +	ret = hwrng_register(&exynos_rng->rng);
>> +	if (ret)
>> +		return ret;
>> +
>> +	return 0;
> 
> Why not just 'return hwrng_register()'?
> 

>> +

>> +static int exynos_rng_runtime_resume(struct device *dev)
>> +{
>> +	struct platform_device *pdev = to_platform_device(dev);
>> +	struct exynos_rng *exynos_rng = platform_get_drvdata(pdev);
>> +
>> +	clk_prepare_enable(exynos_rng->clk);
>> +	return 0;
> 
> Perhaps return the value of clk_prepare_enable() in case it fails for
> some reason?
> 


Okay, I agree with your opinion all.

Thanks.