Re: [PATCH v1] i2c: qcom-geni: Skip extra TX DMA TRE for single read message in GPI mode

[Mukesh Kumar Savaliya]

On 3/26/2026 10:01 AM, Aniket Randive wrote:
> In GPI mode, the I2C GENI driver incorrectly generates an extra TX DMA
> TRE on the TX channel during single read message transfer. This results
What's the impact of this extra DMA TRE ? do you see failure/timeout, 
anything ?
> in an unnecessary write operation on the I2C bus, which is not required.
>
> Update the logic to avoid generating the extra TX DMA TRE for single
> read message, ensuring correct behavior and preventing redundant
> transfers.
So for read, we do unwanted write too ? if so, please write it 
accordingly.  Correct behavior needs to be justified against wrong.
> Co-developed-by: Maramaina Naresh <naresh.maramaina@xxxxxxxxxxxxxxxx>
> Signed-off-by: Maramaina Naresh <naresh.maramaina@xxxxxxxxxxxxxxxx>
> Signed-off-by: Aniket Randive <aniket.randive@xxxxxxxxxxxxxxxx>
> ---
>   drivers/i2c/busses/i2c-qcom-geni.c | 18 +++++++++++++-----
>   1 file changed, 13 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/i2c/busses/i2c-qcom-geni.c b/drivers/i2c/busses/i2c-qcom-geni.c
> index a4acb78fafb6..2706309bbebb 100644
> --- a/drivers/i2c/busses/i2c-qcom-geni.c
> +++ b/drivers/i2c/busses/i2c-qcom-geni.c
> @@ -625,8 +625,8 @@ static int geni_i2c_gpi(struct geni_i2c_dev *gi2c, struct i2c_msg msgs[],
>   {
>   	struct gpi_i2c_config *peripheral;
>   	unsigned int flags;
> -	void *dma_buf;
> -	dma_addr_t addr;
> +	void *dma_buf = NULL;
> +	dma_addr_t addr = 0;
>   	enum dma_data_direction map_dirn;
>   	enum dma_transfer_direction dma_dirn;
>   	struct dma_async_tx_descriptor *desc;
> @@ -639,6 +639,11 @@ static int geni_i2c_gpi(struct geni_i2c_dev *gi2c, struct i2c_msg msgs[],
>   	gi2c_gpi_xfer = &gi2c->i2c_multi_desc_config;
>   	msg_idx = gi2c_gpi_xfer->msg_idx_cnt;
>   
> +	if (op == I2C_WRITE && msgs[msg_idx].flags & I2C_M_RD) {
> +		peripheral->multi_msg = true;
what's the actual meaning of multi_msg here ? IIUC, this multi_msg is 
set to true for single transfer  ? any better name if so ? Yes, need to 
change it out of this patch.
> +		goto skip_dma;
> +	}
> +
>   	dma_buf = i2c_get_dma_safe_msg_buf(&msgs[msg_idx], 1);
>   	if (!dma_buf) {
>   		ret = -ENOMEM;
> @@ -668,6 +673,7 @@ static int geni_i2c_gpi(struct geni_i2c_dev *gi2c, struct i2c_msg msgs[],
>   		flags = DMA_PREP_INTERRUPT | DMA_CTRL_ACK;
>   	}
>   
> +skip_dma:
>   	/* set the length as message for rx txn */
>   	peripheral->rx_len = msgs[msg_idx].len;
>   	peripheral->op = op;
> @@ -740,9 +746,11 @@ static int geni_i2c_gpi(struct geni_i2c_dev *gi2c, struct i2c_msg msgs[],
>   	return 0;
>   
>   err_config:
> -	dma_unmap_single(gi2c->se.dev->parent, addr,
> -			 msgs[msg_idx].len, map_dirn);
> -	i2c_put_dma_safe_msg_buf(dma_buf, &msgs[msg_idx], false);
> +	if (op == I2C_WRITE && (msgs[msg_idx].flags & I2C_M_RD)) {
> +		dma_unmap_single(gi2c->se.dev->parent, addr,
> +				 msgs[msg_idx].len, map_dirn);
> +		i2c_put_dma_safe_msg_buf(dma_buf, &msgs[msg_idx], false);
> +	}
>   
>   out:
>   	gi2c->err = ret;
>
> ---
> base-commit: 785f0eb2f85decbe7c1ef9ae922931f0194ffc2e
> change-id: 20260325-skip_extra_dma_tre-a3cf22f81d9b
>
> Best regards,
> --
> Aniket Randive <aniket.randive@xxxxxxxxxxxxxxxx>