diff mbox

tpm_tis_spi: Use DMA-safe memory for SPI transfers

Message ID 20170704135609.5064-1-Alexander.Steffen@infineon.com (mailing list archive)
State New, archived
Headers show

Commit Message

Alexander Steffen July 4, 2017, 1:56 p.m. UTC
The buffers used as tx_buf/rx_buf in a SPI transfer need to be DMA-safe.
This cannot be guaranteed for the buffers passed to tpm_tis_spi_read_bytes
and tpm_tis_spi_write_bytes. Therefore, we need to use our own DMA-safe
buffer and copy the data to/from it.

Fixes: 0edbfea537d1 ("tpm/tpm_tis_spi: Add support for spi phy")
Cc: stable@vger.kernel.org
Signed-off-by: Alexander Steffen <Alexander.Steffen@infineon.com>
---
 drivers/char/tpm/tpm_tis_spi.c | 32 ++++++++++++++++++--------------
 1 file changed, 18 insertions(+), 14 deletions(-)

Comments

Jason Gunthorpe July 9, 2017, 9:11 p.m. UTC | #1
On Tue, Jul 04, 2017 at 03:56:09PM +0200, Alexander Steffen wrote:
>  struct tpm_tis_spi_phy {
>  	struct tpm_tis_data priv;
>  	struct spi_device *spi_device;
> -
> -	u8 tx_buf[4];
> -	u8 rx_buf[4];
> +	u8 *iobuf;

tpm_tis_spi_phy is already devm_kzalloc'd, why embed another kalloc
pointer inside it?

> +	phy->iobuf = devm_kmalloc(&dev->dev, MAX_SPI_FRAMESIZE, GFP_KERNEL);
> +	if (!phy->iobuf)
> +		return -ENOMEM;

Just do:

  struct tpm_tis_spi_phy {
  	struct tpm_tis_data priv;
  	struct spi_device *spi_device;
        u64 iobuf[MAX_SPI_FRAMESIZE/8];

Jason

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
Jarkko Sakkinen July 16, 2017, 11:41 a.m. UTC | #2
On Tue, 2017-07-04 at 15:56 +0200, Alexander Steffen wrote:
> The buffers used as tx_buf/rx_buf in a SPI transfer need to be DMA-safe.
> This cannot be guaranteed for the buffers passed to tpm_tis_spi_read_bytes
> and tpm_tis_spi_write_bytes. Therefore, we need to use our own DMA-safe
> buffer and copy the data to/from it.
> 
> Fixes: 0edbfea537d1 ("tpm/tpm_tis_spi: Add support for spi phy")
> Cc: stable@vger.kernel.org
> Signed-off-by: Alexander Steffen <Alexander.Steffen@infineon.com>

The commit message does not explain much how you are sorting it out and
how/why it works.

/Jarkko

> ---
>  drivers/char/tpm/tpm_tis_spi.c | 32 ++++++++++++++++++--------------
>  1 file changed, 18 insertions(+), 14 deletions(-)
> 
> diff --git a/drivers/char/tpm/tpm_tis_spi.c b/drivers/char/tpm/tpm_tis_spi.c
> index 88fe72a..5321245 100644
> --- a/drivers/char/tpm/tpm_tis_spi.c
> +++ b/drivers/char/tpm/tpm_tis_spi.c
> @@ -46,9 +46,7 @@
>  struct tpm_tis_spi_phy {
>  	struct tpm_tis_data priv;
>  	struct spi_device *spi_device;
> -
> -	u8 tx_buf[4];
> -	u8 rx_buf[4];
> +	u8 *iobuf;
>  };
>  
>  static inline struct tpm_tis_spi_phy *to_tpm_tis_spi_phy(struct tpm_tis_data *data)
> @@ -71,14 +69,14 @@ static int tpm_tis_spi_transfer(struct tpm_tis_data *data, u32 addr, u16 len,
>  	while (len) {
>  		transfer_len = min_t(u16, len, MAX_SPI_FRAMESIZE);
>  
> -		phy->tx_buf[0] = direction | (transfer_len - 1);
> -		phy->tx_buf[1] = 0xd4;
> -		phy->tx_buf[2] = addr >> 8;
> -		phy->tx_buf[3] = addr;
> +		phy->iobuf[0] = direction | (transfer_len - 1);
> +		phy->iobuf[1] = 0xd4;
> +		phy->iobuf[2] = addr >> 8;
> +		phy->iobuf[3] = addr;
>  
>  		memset(&spi_xfer, 0, sizeof(spi_xfer));
> -		spi_xfer.tx_buf = phy->tx_buf;
> -		spi_xfer.rx_buf = phy->rx_buf;
> +		spi_xfer.tx_buf = phy->iobuf;
> +		spi_xfer.rx_buf = phy->iobuf;
>  		spi_xfer.len = 4;
>  		spi_xfer.cs_change = 1;
>  
> @@ -88,9 +86,9 @@ static int tpm_tis_spi_transfer(struct tpm_tis_data *data, u32 addr, u16 len,
>  		if (ret < 0)
>  			goto exit;
>  
> -		if ((phy->rx_buf[3] & 0x01) == 0) {
> +		if ((phy->iobuf[3] & 0x01) == 0) {
>  			// handle SPI wait states
> -			phy->tx_buf[0] = 0;
> +			phy->iobuf[0] = 0;
>  
>  			for (i = 0; i < TPM_RETRY; i++) {
>  				spi_xfer.len = 1;
> @@ -99,7 +97,7 @@ static int tpm_tis_spi_transfer(struct tpm_tis_data *data, u32 addr, u16 len,
>  				ret = spi_sync_locked(phy->spi_device, &m);
>  				if (ret < 0)
>  					goto exit;
> -				if (phy->rx_buf[0] & 0x01)
> +				if (phy->iobuf[0] & 0x01)
>  					break;
>  			}
>  
> @@ -115,10 +113,9 @@ static int tpm_tis_spi_transfer(struct tpm_tis_data *data, u32 addr, u16 len,
>  
>  		if (direction) {
>  			spi_xfer.tx_buf = NULL;
> -			spi_xfer.rx_buf = buffer;
>  		} else {
> -			spi_xfer.tx_buf = buffer;
>  			spi_xfer.rx_buf = NULL;
> +			memcpy(phy->iobuf, buffer, transfer_len);
>  		}
>  
>  		spi_message_init(&m);
> @@ -127,6 +124,9 @@ static int tpm_tis_spi_transfer(struct tpm_tis_data *data, u32 addr, u16 len,
>  		if (ret < 0)
>  			goto exit;
>  
> +		if (direction)
> +			memcpy(buffer, phy->iobuf, transfer_len);
> +
>  		len -= transfer_len;
>  		buffer += transfer_len;
>  	}
> @@ -194,6 +194,10 @@ static int tpm_tis_spi_probe(struct spi_device *dev)
>  
>  	phy->spi_device = dev;
>  
> +	phy->iobuf = devm_kmalloc(&dev->dev, MAX_SPI_FRAMESIZE, GFP_KERNEL);
> +	if (!phy->iobuf)
> +		return -ENOMEM;
> +
>  	return tpm_tis_core_init(&dev->dev, &phy->priv, -1, &tpm_spi_phy_ops,
>  				 NULL);
>  }

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
Alexander Steffen July 27, 2017, 1:38 p.m. UTC | #3
> >  struct tpm_tis_spi_phy {
> >  	struct tpm_tis_data priv;
> >  	struct spi_device *spi_device;
> > -
> > -	u8 tx_buf[4];
> > -	u8 rx_buf[4];
> > +	u8 *iobuf;
> 
> tpm_tis_spi_phy is already devm_kzalloc'd, why embed another kalloc
> pointer inside it?

I was mostly going by the documentation in https://www.kernel.org/doc/Documentation/DMA-API-HOWTO.txt. Allocating a new buffer ensures that it is cacheline-aligned and not shared with other data, as advised:
"Even if those classes of memory could physically work with DMA, you'd need to ensure the I/O buffers were cacheline-aligned."
"Also, systems with caches that aren't DMA-coherent will work better when the underlying buffers don't share cache lines with other data."

Is that sufficient to justify this implementation? Then I will add some more explanations along those lines to the commit message and resubmit the patch.

Alexander
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
diff mbox

Patch

diff --git a/drivers/char/tpm/tpm_tis_spi.c b/drivers/char/tpm/tpm_tis_spi.c
index 88fe72a..5321245 100644
--- a/drivers/char/tpm/tpm_tis_spi.c
+++ b/drivers/char/tpm/tpm_tis_spi.c
@@ -46,9 +46,7 @@ 
 struct tpm_tis_spi_phy {
 	struct tpm_tis_data priv;
 	struct spi_device *spi_device;
-
-	u8 tx_buf[4];
-	u8 rx_buf[4];
+	u8 *iobuf;
 };
 
 static inline struct tpm_tis_spi_phy *to_tpm_tis_spi_phy(struct tpm_tis_data *data)
@@ -71,14 +69,14 @@  static int tpm_tis_spi_transfer(struct tpm_tis_data *data, u32 addr, u16 len,
 	while (len) {
 		transfer_len = min_t(u16, len, MAX_SPI_FRAMESIZE);
 
-		phy->tx_buf[0] = direction | (transfer_len - 1);
-		phy->tx_buf[1] = 0xd4;
-		phy->tx_buf[2] = addr >> 8;
-		phy->tx_buf[3] = addr;
+		phy->iobuf[0] = direction | (transfer_len - 1);
+		phy->iobuf[1] = 0xd4;
+		phy->iobuf[2] = addr >> 8;
+		phy->iobuf[3] = addr;
 
 		memset(&spi_xfer, 0, sizeof(spi_xfer));
-		spi_xfer.tx_buf = phy->tx_buf;
-		spi_xfer.rx_buf = phy->rx_buf;
+		spi_xfer.tx_buf = phy->iobuf;
+		spi_xfer.rx_buf = phy->iobuf;
 		spi_xfer.len = 4;
 		spi_xfer.cs_change = 1;
 
@@ -88,9 +86,9 @@  static int tpm_tis_spi_transfer(struct tpm_tis_data *data, u32 addr, u16 len,
 		if (ret < 0)
 			goto exit;
 
-		if ((phy->rx_buf[3] & 0x01) == 0) {
+		if ((phy->iobuf[3] & 0x01) == 0) {
 			// handle SPI wait states
-			phy->tx_buf[0] = 0;
+			phy->iobuf[0] = 0;
 
 			for (i = 0; i < TPM_RETRY; i++) {
 				spi_xfer.len = 1;
@@ -99,7 +97,7 @@  static int tpm_tis_spi_transfer(struct tpm_tis_data *data, u32 addr, u16 len,
 				ret = spi_sync_locked(phy->spi_device, &m);
 				if (ret < 0)
 					goto exit;
-				if (phy->rx_buf[0] & 0x01)
+				if (phy->iobuf[0] & 0x01)
 					break;
 			}
 
@@ -115,10 +113,9 @@  static int tpm_tis_spi_transfer(struct tpm_tis_data *data, u32 addr, u16 len,
 
 		if (direction) {
 			spi_xfer.tx_buf = NULL;
-			spi_xfer.rx_buf = buffer;
 		} else {
-			spi_xfer.tx_buf = buffer;
 			spi_xfer.rx_buf = NULL;
+			memcpy(phy->iobuf, buffer, transfer_len);
 		}
 
 		spi_message_init(&m);
@@ -127,6 +124,9 @@  static int tpm_tis_spi_transfer(struct tpm_tis_data *data, u32 addr, u16 len,
 		if (ret < 0)
 			goto exit;
 
+		if (direction)
+			memcpy(buffer, phy->iobuf, transfer_len);
+
 		len -= transfer_len;
 		buffer += transfer_len;
 	}
@@ -194,6 +194,10 @@  static int tpm_tis_spi_probe(struct spi_device *dev)
 
 	phy->spi_device = dev;
 
+	phy->iobuf = devm_kmalloc(&dev->dev, MAX_SPI_FRAMESIZE, GFP_KERNEL);
+	if (!phy->iobuf)
+		return -ENOMEM;
+
 	return tpm_tis_core_init(&dev->dev, &phy->priv, -1, &tpm_spi_phy_ops,
 				 NULL);
 }