Re: [PATCH V2] crypto: qcom-rng - Add hw_random interface support

[Bjorn Andersson]

On Wed, Sep 20, 2023 at 08:34:08AM +0530, Om Prakash Singh wrote:
> Add hw_random interface support in qcom-rng driver as new IP block
> in Qualcomm SoC has inbuilt NIST SP800 90B compliant entropic source
> to generate true random number.
> 
> Keeping current rng_alg interface as well for random number generation
> using Kernel Crypto API.
> 
> Signed-off-by: Om Prakash Singh <quic_omprsing@xxxxxxxxxxx>
> ---
> 
> Changes in V2:
> - Updated patch to fix the return value from qcom_rng_generate() to be
>   consistent with current implementation

As far as I can tell you didn't change this, see below.

> - Updated patch to make it more concise
> - Removed unnecessary use local variable and it's initialization
> - Updated patch to use devm_hwrng_register() instead of hwrng_register()
> - Updated subject line of the patch
> 
> This patch is depends on [1]
> [1] https://lore.kernel.org/lkml/20230824-topic-sm8550-rng-v2-4-dfcafbb16a3e@xxxxxxxxxx/
> 
>  drivers/crypto/qcom-rng.c | 65 ++++++++++++++++++++++++++++++++++-----
>  1 file changed, 58 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/crypto/qcom-rng.c b/drivers/crypto/qcom-rng.c
> index fb54b8cfc35f..e5a574a3cc59 100644
> --- a/drivers/crypto/qcom-rng.c
> +++ b/drivers/crypto/qcom-rng.c
> @@ -7,6 +7,7 @@
>  #include <linux/acpi.h>
>  #include <linux/clk.h>
>  #include <linux/crypto.h>
> +#include <linux/hw_random.h>
>  #include <linux/io.h>
>  #include <linux/iopoll.h>
>  #include <linux/kernel.h>
> @@ -32,13 +33,18 @@ struct qcom_rng {
>  	struct mutex lock;
>  	void __iomem *base;
>  	struct clk *clk;
> -	unsigned int skip_init;
> +	struct qcom_rng_of_data *of_data;
>  };
>  
>  struct qcom_rng_ctx {
>  	struct qcom_rng *rng;
>  };
>  
> +struct qcom_rng_of_data {
> +	bool skip_init;
> +	bool hwrng_support;
> +};
> +
>  static struct qcom_rng *qcom_rng_dev;
>  
>  static int qcom_rng_read(struct qcom_rng *rng, u8 *data, unsigned int max)
> @@ -70,7 +76,7 @@ static int qcom_rng_read(struct qcom_rng *rng, u8 *data, unsigned int max)
>  		}
>  	} while (currsize < max);
>  
> -	return 0;
> +	return currsize;

As I pointed out in my previous review, if the qcom_rng_read() is
requested to read a number of bytes (max) that is not evenly divisible
with 4 (WORD_SZ) the loop will exit without accounting for the last
bytes copied...

>  }
>  
>  static int qcom_rng_generate(struct crypto_rng *tfm,
> @@ -92,6 +98,9 @@ static int qcom_rng_generate(struct crypto_rng *tfm,
>  	mutex_unlock(&rng->lock);
>  	clk_disable_unprepare(rng->clk);
>  
> +	if (ret == dlen)

...this means that if dlen % 4, you're changing the return value of this
function from 0 to dlen.

> +		ret = 0;
> +
>  	return ret;
>  }
>  
> @@ -101,6 +110,13 @@ static int qcom_rng_seed(struct crypto_rng *tfm, const u8 *seed,
>  	return 0;
>  }
>  
> +static int qcom_hwrng_read(struct hwrng *rng, void *data, size_t max, bool wait)
> +{
> +	struct qcom_rng *qrng = (struct qcom_rng *)rng->priv;

You missed Herbert's request in [1], which I presume implies that
qcom_hwrng should be moved into struct qcom_rng, which would mean that
you can get the qrng by container_of().

[1] https://lore.kernel.org/lkml/ZQQvlXvGy8p01uJS@xxxxxxxxxxxxxxxxxxx/

> +
> +	return qcom_rng_read(qrng, data, max);
> +}
> +
>  static int qcom_rng_enable(struct qcom_rng *rng)
>  {
>  	u32 val;
> @@ -136,7 +152,7 @@ static int qcom_rng_init(struct crypto_tfm *tfm)
>  
>  	ctx->rng = qcom_rng_dev;
>  
> -	if (!ctx->rng->skip_init)
> +	if (!ctx->rng->of_data->skip_init)
>  		return qcom_rng_enable(ctx->rng);
>  
>  	return 0;
> @@ -157,6 +173,12 @@ static struct rng_alg qcom_rng_alg = {
>  	}
>  };
>  
> +static struct hwrng qcom_hwrng = {
> +	.name = "qcom-hwrng",
> +	.read = qcom_hwrng_read,
> +	.quality = 1024,
> +};

Which would mean not adding this static global variable...

Regards,
Bjorn