Re: RE: [PATCH 2/3] can: xilinx_can: Add ECC support

[Marc Kleine-Budde <mkl@pengutronix.de>]

[-- Attachment #1: Type: text/plain, Size: 3689 bytes --]

On 21.07.2023 05:23:02, Goud, Srinivas wrote:
> >> +	if (priv->ecc_enable) {
> >> +		u32 reg_ecc;
> >> +
> >> +		reg_ecc = priv->read_reg(priv, XCAN_RXFIFO_ECC_OFFSET);
> >> +		if (isr & XCAN_IXR_E2BERX_MASK) {
> >> +			priv->ecc_2bit_rxfifo_cnt +=
> >> +				FIELD_GET(XCAN_ECC_2BIT_CNT_MASK,
> >reg_ecc);
> >> +			netdev_dbg(ndev, "%s: RX FIFO 2bit ECC error count
> >%d\n",
> >> +				   __func__, priv->ecc_2bit_rxfifo_cnt);
> >> +		}
> >> +		if (isr & XCAN_IXR_E1BERX_MASK) {
> >> +			priv->ecc_1bit_rxfifo_cnt += reg_ecc &
> >> +				XCAN_ECC_1BIT_CNT_MASK;
> >
> >Please use FIELD_GET here, too.
> >
> >> +			netdev_dbg(ndev, "%s: RX FIFO 1bit ECC error count
> >%d\n",
> >> +				   __func__, priv->ecc_1bit_rxfifo_cnt);
> >> +		}
> >> +
> >> +		reg_ecc = priv->read_reg(priv, XCAN_TXOLFIFO_ECC_OFFSET);
> >> +		if (isr & XCAN_IXR_E2BETXOL_MASK) {
> >> +			priv->ecc_2bit_txolfifo_cnt +=
> >> +				FIELD_GET(XCAN_ECC_2BIT_CNT_MASK,
> >reg_ecc);
> >> +			netdev_dbg(ndev, "%s: TXOL FIFO 2bit ECC error count
> >%d\n",
> >> +				   __func__, priv->ecc_2bit_txolfifo_cnt);
> >> +		}
> >> +		if (isr & XCAN_IXR_E1BETXOL_MASK) {
> >> +			priv->ecc_1bit_txolfifo_cnt += reg_ecc &
> >> +				XCAN_ECC_1BIT_CNT_MASK;
> >
> >same here
> >
> >> +			netdev_dbg(ndev, "%s: TXOL FIFO 1bit ECC error count
> >%d\n",
> >> +				   __func__, priv->ecc_1bit_txolfifo_cnt);
> >> +		}
> >> +
> >> +		reg_ecc = priv->read_reg(priv, XCAN_TXTLFIFO_ECC_OFFSET);
> >> +		if (isr & XCAN_IXR_E2BETXTL_MASK) {
> >> +			priv->ecc_2bit_txtlfifo_cnt +=
> >> +				FIELD_GET(XCAN_ECC_2BIT_CNT_MASK,
> >reg_ecc);
> >> +			netdev_dbg(ndev, "%s: TXTL FIFO 2bit ECC error count
> >%d\n",
> >> +				   __func__, priv->ecc_2bit_txtlfifo_cnt);
> >> +		}
> >> +		if (isr & XCAN_IXR_E1BETXTL_MASK) {
> >> +			priv->ecc_1bit_txtlfifo_cnt += reg_ecc &
> >> +				XCAN_ECC_1BIT_CNT_MASK;
> >
> >same here
> >
> >> +			netdev_dbg(ndev, "%s: TXTL FIFO 1bit ECC error count
> >%d\n",
> >> +				   __func__, priv->ecc_1bit_txtlfifo_cnt);
> >> +		}
> >> +
> >> +		/* Reset FIFO ECC counters */
> >> +		priv->write_reg(priv, XCAN_ECC_CFG_OFFSET,
> >XCAN_ECC_CFG_REECRX_MASK |
> >> +				XCAN_ECC_CFG_REECTXOL_MASK |
> >XCAN_ECC_CFG_REECTXTL_MASK);
> >
> >This is racy - you will lose events that occur between reading the register value
> >and clearing the register. You can save the old value and add the difference
> >between the new and the old value to the total counter. What happens when
> >the counter overflows? The following pseudocode should handle the u16
> >counter rolling over to 0:
> As per IP specifications when counter reaching maximum value of 0xFFFF will 
> stays there until reset.
> 
> Not sure we can avoid this race condition completely, as we need to reset
> the counters after reaching the 0xFFFF to avoid losing the events.
> 
> To minimize the race condition, we can reset the counters after reaching 
> the events to 0xFFFF instead of resetting for every interrupt. 
> Can you please suggest if we can go this approach.

Ok, the counter doesn't overflow. To keep the logic in the driver
simple, I think it's best to read the value and reset the counter
directly in the next line. Please add a comment like: "The counter
reaches its maximum at 0xffff and does not overflow. Accept the small
race window between reading and resetting."

regards,
Marc

-- 
Pengutronix e.K.                 | Marc Kleine-Budde          |
Embedded Linux                   | https://www.pengutronix.de |
Vertretung Nürnberg              | Phone: +49-5121-206917-129 |
Amtsgericht Hildesheim, HRA 2686 | Fax:   +49-5121-206917-9   |

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]