Re: [PATCH v4 1/2] i2c: tegra: Fix PEC support for SMBUS block read

[Thierry Reding]

On Fri, Mar 24, 2023 at 05:29:23PM +0530, Akhil R wrote:
> Update the msg->len value correctly for SMBUS block read. The discrepancy
> went unnoticed as msg->len is used in SMBUS transfers only when a PEC
> byte is added.
> 
> Fixes: d7583c8a5748 ("i2c: tegra: Add SMBus block read function")
> Signed-off-by: Akhil R <akhilrajeev@xxxxxxxxxx>
> ---
>  drivers/i2c/busses/i2c-tegra.c | 38 +++++++++++++++++++++++-----------
>  1 file changed, 26 insertions(+), 12 deletions(-)
> 
> diff --git a/drivers/i2c/busses/i2c-tegra.c b/drivers/i2c/busses/i2c-tegra.c
> index 6aab84c8d22b..83e74b8baf67 100644
> --- a/drivers/i2c/busses/i2c-tegra.c
> +++ b/drivers/i2c/busses/i2c-tegra.c
> @@ -245,6 +245,7 @@ struct tegra_i2c_hw_feature {
>   * @msg_err: error code for completed message
>   * @msg_buf: pointer to current message data
>   * @msg_buf_remaining: size of unsent data in the message buffer
> + * @msg_len: length of message in current transfer
>   * @msg_read: indicates that the transfer is a read access
>   * @timings: i2c timings information like bus frequency
>   * @multimaster_mode: indicates that I2C controller is in multi-master mode
> @@ -279,6 +280,7 @@ struct tegra_i2c_dev {
>  	size_t msg_buf_remaining;
>  	int msg_err;
>  	u8 *msg_buf;
> +	__u16 msg_len;
>  
>  	struct completion dma_complete;
>  	struct dma_chan *tx_dma_chan;
> @@ -1169,7 +1171,7 @@ static void tegra_i2c_push_packet_header(struct tegra_i2c_dev *i2c_dev,
>  	else
>  		i2c_writel(i2c_dev, packet_header, I2C_TX_FIFO);
>  
> -	packet_header = msg->len - 1;
> +	packet_header = i2c_dev->msg_len - 1;
>  
>  	if (i2c_dev->dma_mode && !i2c_dev->msg_read)
>  		*dma_buf++ = packet_header;
> @@ -1242,20 +1244,32 @@ static int tegra_i2c_xfer_msg(struct tegra_i2c_dev *i2c_dev,
>  		return err;
>  
>  	i2c_dev->msg_buf = msg->buf;
> +	i2c_dev->msg_len = msg->len;
>  
> -	/* The condition true implies smbus block read and len is already read */
> -	if (msg->flags & I2C_M_RECV_LEN && end_state != MSG_END_CONTINUE)
> -		i2c_dev->msg_buf = msg->buf + 1;
> -
> -	i2c_dev->msg_buf_remaining = msg->len;
>  	i2c_dev->msg_err = I2C_ERR_NONE;
>  	i2c_dev->msg_read = !!(msg->flags & I2C_M_RD);
>  	reinit_completion(&i2c_dev->msg_complete);
>  
> +	/* *
> +	 * For SMBUS block read command, read only 1 byte in the first transfer.
> +	 * Adjust that 1 byte for the next transfer in the msg buffer and msg
> +	 * length.
> +	 */
> +	if (msg->flags & I2C_M_RECV_LEN) {
> +		if (end_state == MSG_END_CONTINUE) {
> +			i2c_dev->msg_len = 1;
> +		} else {
> +			i2c_dev->msg_buf += 1;
> +			i2c_dev->msg_len -= 1;
> +		}
> +	}
> +
> +	i2c_dev->msg_buf_remaining = i2c_dev->msg_len;
> +
>  	if (i2c_dev->msg_read)
> -		xfer_size = msg->len;
> +		xfer_size = i2c_dev->msg_len;
>  	else
> -		xfer_size = msg->len + I2C_PACKET_HEADER_SIZE;
> +		xfer_size = i2c_dev->msg_len + I2C_PACKET_HEADER_SIZE;
>  
>  	xfer_size = ALIGN(xfer_size, BYTES_PER_FIFO_WORD);
>  
> @@ -1295,7 +1309,7 @@ static int tegra_i2c_xfer_msg(struct tegra_i2c_dev *i2c_dev,
>  	if (!i2c_dev->msg_read) {
>  		if (i2c_dev->dma_mode) {
>  			memcpy(i2c_dev->dma_buf + I2C_PACKET_HEADER_SIZE,
> -			       msg->buf, msg->len);
> +			       msg->buf, i2c_dev->msg_len);
>  
>  			dma_sync_single_for_device(i2c_dev->dma_dev,
>  						   i2c_dev->dma_phys,
> @@ -1352,7 +1366,7 @@ static int tegra_i2c_xfer_msg(struct tegra_i2c_dev *i2c_dev,
>  						i2c_dev->dma_phys,
>  						xfer_size, DMA_FROM_DEVICE);
>  
> -			memcpy(i2c_dev->msg_buf, i2c_dev->dma_buf, msg->len);
> +			memcpy(i2c_dev->msg_buf, i2c_dev->dma_buf, i2c_dev->msg_len);
>  		}
>  	}
>  
> @@ -1408,8 +1422,8 @@ static int tegra_i2c_xfer(struct i2c_adapter *adap, struct i2c_msg msgs[],
>  			ret = tegra_i2c_xfer_msg(i2c_dev, &msgs[i], MSG_END_CONTINUE);
>  			if (ret)
>  				break;
> -			/* Set the read byte as msg len */
> -			msgs[i].len = msgs[i].buf[0];
> +			/* Set the msg length from first byte */
> +			msgs[i].len += msgs[i].buf[0];

I'm having trouble understanding why this change is needed. msg->len is
never changed in tegra_i2c_xfer_msg(), as far as I can tell, so it would
contain whatever was in it for the previous transfer. But since we want
to read the message length from the first byte, wouldn't the assignment
(i.e. the old code) be the right way to do that? If we add the length
from the first byte, we could potentially be reading more than whan the
first byte indicated.

What am I missing?

Thierry

Attachment: signature.asc
Description: PGP signature