Re: [RESEND PATCH 02/10] crypto: sa2ul: Add crypto driver

[Eric Biggers]

On Fri, Jun 28, 2019 at 09:57:37AM +0530, Keerthy wrote:
> The Security Accelerator (SA2_UL) subsystem provides hardware
> cryptographic acceleration for the following use cases:
> â Encryption and authentication for secure boot
> â Encryption and authentication of content in applications
>   requiring DRM (digital rights management) and
>   content/asset protection
> The device includes one instantiation of SA2_UL named SA2_UL0
> 
> SA2_UL supports the following cryptographic industry standards to enable data authentication, data
> integrity and data confidentiality.
> 
> Crypto function library for software acceleration
> o AES operation
> o 3DES operation
> o SHA1 operation
> o MD5 operation
> o SHA2 â 224, 256, 384, 512 operation
> 
> Authentication supported via following hardware cores
> o SHA1
> o MD5
> o SHA2 -224
> o SHA2-256
> o SHA2-384
> o SHA2-512

What about HMAC?  

Your actual driver only exposes HMAC-SHA*, not SHA* anything.

What does the hardware actually support?

> diff --git a/drivers/crypto/Kconfig b/drivers/crypto/Kconfig
> index 603413f28fa3..b9a3fa026c74 100644
> --- a/drivers/crypto/Kconfig
> +++ b/drivers/crypto/Kconfig
> @@ -785,4 +785,21 @@ config CRYPTO_DEV_CCREE
>  
>  source "drivers/crypto/hisilicon/Kconfig"
>  
> +config CRYPTO_DEV_SA2UL
> +	tristate "Support for TI security accelerator"
> +	depends on ARCH_K3 || COMPILE_TEST
> +	select ARM64_CRYPTO
> +	select CRYPTO_AES
> +	select CRYPTO_AES_ARM64
> +	select CRYPTO_SHA1
> +	select CRYPTO_MD5
> +	select CRYPTO_ALGAPI
> +	select CRYPTO_AUTHENC
> +	select HW_RANDOM
> +	default m if ARCH_K3
> +	help
> +	  Keystone devices include a security accelerator engine that may be
> +	  used for crypto offload.  Select this if you want to use hardware
> +	  acceleration for cryptographic algorithms on these devices.

This shouldn't be enabled by default.  Note that arm64 defconfig sets ARCH_K3 as
well as lots of other ARCH_* options, so clearly just because ARCH_K3 is set
doesn't mean the kernel is being built specifically for your platform.

> +/*
> + * Mode Control Instructions for various Key lengths 128, 192, 256
> + * For CBC (Cipher Block Chaining) mode for encryption
> + */
> +static u8 mci_cbc_enc_array[3][MODE_CONTROL_BYTES] = {
> +	{	0x21, 0x00, 0x00, 0x18, 0x88, 0x0a, 0xaa, 0x4b, 0x7e, 0x00,
> +		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
> +		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00	},
> +	{	0x21, 0x00, 0x00, 0x18, 0x88, 0x4a, 0xaa, 0x4b, 0x7e, 0x00,
> +		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
> +		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00	},
> +	{	0x21, 0x00, 0x00, 0x18, 0x88, 0x8a, 0xaa, 0x4b, 0x7e, 0x00,
> +		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
> +		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00	},
> +};

Use 'const' for static constants.

> +static int sa_aes_cbc_setkey(struct crypto_ablkcipher *tfm, const u8 *key,
> +			     unsigned int keylen)
> +{
> +	struct algo_data *ad = kzalloc(sizeof(*ad), GFP_KERNEL);

Need to check from error for all memory allocations.

> +static struct sa_alg_tmpl sa_algs[] = {
> +	{.type = CRYPTO_ALG_TYPE_ABLKCIPHER,

ablkcipher API is deprecated.  Use skcipher instead.

(To be clear, these are just a few things I happened to notice from very quickly
skimming through this patch.  I don't have time to do a proper review of random
drivers.)

- Eric