Re: [PATCH 2/2] scsi: ufs: add inline crypto support to UFS HCD

[Greg KH]

On Tue, Dec 11, 2018 at 09:50:27AM +0000, Parshuram Thombare wrote:
> Add real time crypto support to UFS HCD using new device
> mapper 'crypto-ufs'. dmsetup tool can be used to enable
> real time / inline crypto support using device mapper
> 'crypt-ufs'.
> 
> Signed-off-by: Parshuram Thombare <pthombar@xxxxxxxxxxx>

As you cc:ed me, I'll provide some minor review comments:

> +config BLK_DEV_HW_RT_ENCRYPTION
> +	bool
> +	depends on SCSI_UFSHCD_RT_ENCRYPTION
> +	default n

n is always the default, you never need to list that.

> +
>  source block/Kconfig.iosched
> diff --git a/drivers/scsi/ufs/Kconfig b/drivers/scsi/ufs/Kconfig
> index 2ddbb26..09a3ec0 100644
> --- a/drivers/scsi/ufs/Kconfig
> +++ b/drivers/scsi/ufs/Kconfig
> @@ -136,3 +136,15 @@ config SCSI_UFS_BSG
>  
>  	  Select this if you need a bsg device node for your UFS controller.
>  	  If unsure, say N.
> +
> +config SCSI_UFSHCD_RT_ENCRYPTION
> +	bool "Universal Flash Storage Controller RT encryption support"
> +	depends on SCSI_UFSHCD
> +	default n

Same here.

> +	select BLK_DEV_HW_RT_ENCRYPTION if SCSI_UFSHCD_RT_ENCRYPTION
> +	select BLK_DEV_DM if SCSI_UFSHCD_RT_ENCRYPTION
> +	help
> +	Universal Flash Storage Controller RT encryption support
> +
> +	Select this if you want to enable real time encryption on UFS controller
> +	If unsure, say N.

Don't you need to indent the help lines?

> +int ufshcd_crypto_init(struct ufs_hba *hba);
> +void ufshcd_crypto_remove(struct ufs_hba *hba);
> +void ufshcd_prepare_for_crypto(struct ufs_hba *hba, struct ufshcd_lrb *lrbp);
> +#endif

You need to provide inline functions for when your config option is not
enabled here.

That way you don't have to do this mess:

> diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
> index 86fe114..a96b038 100644
> --- a/drivers/scsi/ufs/ufshcd.c
> +++ b/drivers/scsi/ufs/ufshcd.c
> @@ -47,6 +47,9 @@
>  #include "unipro.h"
>  #include "ufs-sysfs.h"
>  #include "ufs_bsg.h"
> +#ifdef CONFIG_SCSI_UFSHCD_RT_ENCRYPTION
> +#include "ufshcd-crypto.h"
> +#endif

No need for #ifdefs in .c files at all.  Always include the .h file, and
then since your functions are all inline void functions, the code just
compiles away into nothing.


>  
>  #define CREATE_TRACE_POINTS
>  #include <trace/events/ufs.h>
> @@ -2198,6 +2201,14 @@ static void ufshcd_prepare_req_desc_hdr(struct ufshcd_lrb *lrbp,
>  
>  	dword_0 = data_direction | (lrbp->command_type
>  				<< UPIU_COMMAND_TYPE_OFFSET);
> +#ifdef CONFIG_SCSI_UFSHCD_RT_ENCRYPTION
> +	if (lrbp->cci >= 0) {
> +		dword_0 |= (1 << CRYPTO_UTP_REQ_DESC_DWORD0_CE_SHIFT);
> +		dword_0 |= ((lrbp->cci << CRYPTO_UTP_REQ_DESC_DWORD0_CCI_SHIFT)
> +				& CRYPTO_UTP_REQ_DESC_DWORD0_CCI_MASK);
> +	} else
> +		dword_0 &= ~CRYPTO_UTP_REQ_DESC_DWORD0_CE_MASK;
> +#endif

Some comments on what this is trying to do here?


>  	if (lrbp->intr_cmd)
>  		dword_0 |= UTP_REQ_DESC_INT_CMD;
>  
> @@ -2462,6 +2473,12 @@ static int ufshcd_queuecommand(struct Scsi_Host *host, struct scsi_cmnd *cmd)
>  	lrbp->intr_cmd = !ufshcd_is_intr_aggr_allowed(hba) ? true : false;
>  	lrbp->req_abort_skip = false;
>  
> +#ifdef CONFIG_SCSI_UFSHCD_RT_ENCRYPTION
> +	lrbp->cci = -1;

What does -1 mean?  You should have a type for it if it is something
"special".

> +	/* prepare block for crypto */
> +	if (hba->capabilities & MASK_CRYPTO_SUPPORT)
> +		ufshcd_prepare_for_crypto(hba, lrbp);
> +#endif

Again, no #ifdefs needed please.


>  	ufshcd_comp_scsi_upiu(hba, lrbp);
>  
>  	err = ufshcd_map_sg(hba, lrbp);
> @@ -8119,6 +8136,11 @@ void ufshcd_remove(struct ufs_hba *hba)
>  	ufs_bsg_remove(hba);
>  	ufs_sysfs_remove_nodes(hba->dev);
>  	scsi_remove_host(hba->host);
> +#ifdef CONFIG_SCSI_UFSHCD_RT_ENCRYPTION
> +	if (hba->capabilities & MASK_CRYPTO_SUPPORT)
> +		ufshcd_crypto_remove(hba);
> +#endif

Same ifdef here, and everywhere else in this file.

> --- a/drivers/scsi/ufs/ufshcd.h
> +++ b/drivers/scsi/ufs/ufshcd.h
> @@ -193,6 +193,9 @@ struct ufshcd_lrb {
>  	ktime_t compl_time_stamp;
>  
>  	bool req_abort_skip;
> +#ifdef CONFIG_SCSI_UFSHCD_RT_ENCRYPTION
> +	int cci;
> +#endif

No need for a #ifdef here.  But comment on exactly what "cci" means,
that seems to not make any sense to me.

>  };
>  
>  /**
> @@ -706,6 +709,9 @@ struct ufs_hba {
>  
>  	struct device		bsg_dev;
>  	struct request_queue	*bsg_queue;
> +#ifdef CONFIG_SCSI_UFSHCD_RT_ENCRYPTION
> +	struct ufshcd_crypto_ctx *cctx;
> +#endif

or here.

>  };
>  
>  /* Returns true if clocks can be gated. Otherwise false */
> diff --git a/drivers/scsi/ufs/ufshci.h b/drivers/scsi/ufs/ufshci.h
> index 6fa889d..fe5a92d 100644
> --- a/drivers/scsi/ufs/ufshci.h
> +++ b/drivers/scsi/ufs/ufshci.h
> @@ -90,6 +90,7 @@ enum {
>  	MASK_64_ADDRESSING_SUPPORT		= 0x01000000,
>  	MASK_OUT_OF_ORDER_DATA_DELIVERY_SUPPORT	= 0x02000000,
>  	MASK_UIC_DME_TEST_MODE_SUPPORT		= 0x04000000,
> +	MASK_CRYPTO_SUPPORT                     = 0x10000000,

Why are you all not using the BIT(N) macro for these?

thanks,

greg k-h