Re: [PATCH v1] i2c: designware: Disable TX_EMPTY irq while waiting for block length byte

[Yann Sionneau <ysionneau@kalrayinc.com>]

Hi Tam,

On 9/29/23 05:53, Tam Nguyen wrote:
> During SMBus block data read process, we have seen high interrupt rate
> because of TX_EMPTY irq status while waiting for block length byte (the
> first data byte after the address phase). The interrupt handler does not
> do anything because the internal state is kept as STATUS_WRITE_IN_PROGRESS.
> Hence, we should disable TX_EMPTY irq until I2C DW receives first data
> byte from I2C device, then re-enable it.
>
> It takes 0.789 ms for host to receive data length from slave.
> Without the patch, i2c_dw_isr is called 99 times by TX_EMPTY interrupt.
> And it is none after applying the patch.
>
> Cc: stable@vger.kernel.org
> Signed-off-by: Chuong Tran <chuong@os.amperecomputing.com>
> Signed-off-by: Tam Nguyen <tamnguyenchi@os.amperecomputing.com>
> ---
>   drivers/i2c/busses/i2c-designware-master.c | 13 +++++++++++++
>   1 file changed, 13 insertions(+)
>
> diff --git a/drivers/i2c/busses/i2c-designware-master.c b/drivers/i2c/busses/i2c-designware-master.c
> index 55ea91a63382..2152b1f9b27c 100644
> --- a/drivers/i2c/busses/i2c-designware-master.c
> +++ b/drivers/i2c/busses/i2c-designware-master.c
> @@ -462,6 +462,13 @@ i2c_dw_xfer_msg(struct dw_i2c_dev *dev)
>   		if (buf_len > 0 || flags & I2C_M_RECV_LEN) {
>   			/* more bytes to be written */
>   			dev->status |= STATUS_WRITE_IN_PROGRESS;
> +			/*
> +			 * In I2C_FUNC_SMBUS_BLOCK_DATA case, there is no data
> +			 * to send before receiving data length from slave.
> +			 * Disable TX_EMPTY while waiting for data length byte
> +			 */
> +			if (flags & I2C_M_RECV_LEN)
> +				intr_mask &= ~DW_IC_INTR_TX_EMPTY;
>   			break;
>   		} else
>   			dev->status &= ~STATUS_WRITE_IN_PROGRESS;
> @@ -485,6 +492,7 @@ i2c_dw_recv_len(struct dw_i2c_dev *dev, u8 len)
>   {
>   	struct i2c_msg *msgs = dev->msgs;
>   	u32 flags = msgs[dev->msg_read_idx].flags;
> +	u32 intr_mask;
>   
>   	/*
>   	 * Adjust the buffer length and mask the flag
> @@ -495,6 +503,11 @@ i2c_dw_recv_len(struct dw_i2c_dev *dev, u8 len)
>   	msgs[dev->msg_read_idx].len = len;
>   	msgs[dev->msg_read_idx].flags &= ~I2C_M_RECV_LEN;
>   
> +	/* Re-enable TX_EMPTY interrupt. */
> +	regmap_read(dev->map, DW_IC_INTR_MASK, &intr_mask);
> +	intr_mask |= DW_IC_INTR_TX_EMPTY;
> +	regmap_write(dev->map, DW_IC_INTR_MASK, intr_mask);
> +
>   	return len;
>   }

I tested this patch on Kalray k200 and k200lp boards (Coolidge SoC, kvx 
arch). With this patch all our CI pipelines are green.

Tested-by: Yann Sionneau <ysionneau@kalrayinc.com>

-- 

Yann