Re: [PATCH v3 1/6] cpufreq: sun50i: add efuse_xlate to get efuse version.

[Samuel Holland]

Hello,

On 12/7/20 8:37 PM, Shuosheng Huang wrote:
> It's better to use efuse_xlate to extract the differentiated part
> regarding different SoC.
> 
> Signed-off-by: Shuosheng Huang <huangshuosheng@xxxxxxxxxxxxxxxxx>
> ---
>  drivers/cpufreq/sun50i-cpufreq-nvmem.c | 72 +++++++++++++++++---------
>  1 file changed, 48 insertions(+), 24 deletions(-)
> 
> diff --git a/drivers/cpufreq/sun50i-cpufreq-nvmem.c b/drivers/cpufreq/sun50i-cpufreq-nvmem.c
> index 9907a165135b..da23d581a4b4 100644
> --- a/drivers/cpufreq/sun50i-cpufreq-nvmem.c
> +++ b/drivers/cpufreq/sun50i-cpufreq-nvmem.c
> @@ -22,21 +22,52 @@
>  #define NVMEM_MASK	0x7
>  #define NVMEM_SHIFT	5

You should remove these definitions now that you added the ones below.

>
> +#define SUN50I_H6_NVMEM_MASK		0x7
> +#define SUN50I_H6_NVMEM_SHIFT		5
> +
> +struct sunxi_cpufreq_soc_data {
> +	u32 (*efuse_xlate)(void *efuse);
> +};
> +
>  static struct platform_device *cpufreq_dt_pdev, *sun50i_cpufreq_pdev;
>  
> +static u32 sun50i_h6_efuse_xlate(struct nvmem_cell *speedbin_nvmem)
> +{
> +	size_t len;
> +	u32 *speedbin;
> +	u32 efuse_value;
> +
> +	speedbin = nvmem_cell_read(speedbin_nvmem, &len);
> +	if (IS_ERR(speedbin))
> +		return PTR_ERR(speedbin);

You're trying to return a negative value for an error here, but the
return type is unsigned.

> +
> +	efuse_value = (*(u32 *)speedbin >> SUN50I_H6_NVMEM_SHIFT) &
> +			  SUN50I_H6_NVMEM_MASK;
> +	kfree(speedbin);
> +	/*
> +	 * We treat unexpected efuse values as if the SoC was from
> +	 * the slowest bin. Expected efuse values are 1-3, slowest
> +	 * to fastest.
> +	 */
> +	if (efuse_value >= 1 && efuse_value <= 3)
> +		return efuse_value - 1;
> +	else
> +		return 0;
> +}
> +
>  /**
>   * sun50i_cpufreq_get_efuse() - Determine speed grade from efuse value
> + * @soc_data: pointer to sunxi_cpufreq_soc_data context
>   * @versions: Set to the value parsed from efuse
>   *
>   * Returns 0 if success.
>   */
> -static int sun50i_cpufreq_get_efuse(u32 *versions)
> +static int sun50i_cpufreq_get_efuse(const struct sunxi_cpufreq_soc_data *soc_data,
> +				    u32 *versions)
>  {
>  	struct nvmem_cell *speedbin_nvmem;
>  	struct device_node *np;
>  	struct device *cpu_dev;
> -	u32 *speedbin, efuse_value;
> -	size_t len;
>  	int ret;
>  
>  	cpu_dev = get_cpu_device(0);
> @@ -63,41 +94,31 @@ static int sun50i_cpufreq_get_efuse(u32 *versions)
>  		return PTR_ERR(speedbin_nvmem);
>  	}
>  
> -	speedbin = nvmem_cell_read(speedbin_nvmem, &len);
> +	*versions = soc_data->efuse_xlate(speedbin_nvmem);
>  	nvmem_cell_put(speedbin_nvmem);
> -	if (IS_ERR(speedbin))
> -		return PTR_ERR(speedbin);
> -
> -	efuse_value = (*speedbin >> NVMEM_SHIFT) & NVMEM_MASK;
> -
> -	/*
> -	 * We treat unexpected efuse values as if the SoC was from
> -	 * the slowest bin. Expected efuse values are 1-3, slowest
> -	 * to fastest.
> -	 */
> -	if (efuse_value >= 1 && efuse_value <= 3)
> -		*versions = efuse_value - 1;
> -	else
> -		*versions = 0;
>  
> -	kfree(speedbin);
>  	return 0;

If *versions is negative, you need to pass that error back to the caller.

>  };
>  
>  static int sun50i_cpufreq_nvmem_probe(struct platform_device *pdev)
>  {
> +	const struct of_device_id *match;
>  	struct opp_table **opp_tables;
>  	char name[MAX_NAME_LEN];
>  	unsigned int cpu;
>  	u32 speed = 0;
>  	int ret;
>  
> +	match = dev_get_platdata(&pdev->dev);
> +	if (!match)
> +		return -EINVAL;
> +
>  	opp_tables = kcalloc(num_possible_cpus(), sizeof(*opp_tables),
>  			     GFP_KERNEL);
>  	if (!opp_tables)
>  		return -ENOMEM;
>  
> -	ret = sun50i_cpufreq_get_efuse(&speed);
> +	ret = sun50i_cpufreq_get_efuse(match->data, &speed);
>  	if (ret)
>  		return ret;

Since we've already decoupled the speed grade from the nvmem value, why
not make this:

	int speed;
	...
	speed = sun50i_cpufreq_get_efuse(match->data);
	if (speed < 0)
		return speed;

And have sun50i_cpufreq_get_efuse return the value from efuse_xlate
without all the pointer indirection.

And this is a separate bug in the existing code, but it leaks opp_tables
if sun50i_cpufreq_get_efuse fails. That could be fixed by moving the
call to sun50i_cpufreq_get_efuse up.

Cheers,
Samuel

>  
> @@ -163,8 +184,12 @@ static struct platform_driver sun50i_cpufreq_driver = {
>  	},
>  };
>  
> +static const struct sunxi_cpufreq_soc_data sun50i_h6_data = {
> +	.efuse_xlate = sun50i_h6_efuse_xlate,
> +};
> +
>  static const struct of_device_id sun50i_cpufreq_match_list[] = {
> -	{ .compatible = "allwinner,sun50i-h6" },
> +	{ .compatible = "allwinner,sun50i-h6", .data = &sun50i_h6_data },
>  	{}
>  };
>  
> @@ -198,9 +223,8 @@ static int __init sun50i_cpufreq_init(void)
>  	if (unlikely(ret < 0))
>  		return ret;
>  
> -	sun50i_cpufreq_pdev =
> -		platform_device_register_simple("sun50i-cpufreq-nvmem",
> -						-1, NULL, 0);
> +	sun50i_cpufreq_pdev = platform_device_register_data(NULL,
> +		"sun50i-cpufreq-nvmem", -1, match, sizeof(*match));
>  	ret = PTR_ERR_OR_ZERO(sun50i_cpufreq_pdev);
>  	if (ret == 0)
>  		return 0;
>