Re: [PATCH] net: can: flexcan: reset the error counter after leaving passive state

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

[-- Attachment #1: Type: text/plain, Size: 5257 bytes --]

On 07/25/2014 08:51 PM, Sebastian Andrzej Siewior wrote:
> The reasoning of why is done what is done is described in the longer
> comment which I do not repeat here.
> While pleaying with the HALT/Freeze mode/bit I noticed that the manual
> says that one has to wait until the core completed the request or
>  "otherwise FLEXCAN may operate in an unpredictable way"
> therefore flexcan_wait_for_frz() is added to places where the HALT bit
> is set / cleared.

Please have a look at a recent kernel, especially patch

    b1aa1c7 can: flexcan: fix transition from and to freeze mode
                          in chip_{,un}freeze

> 
> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
> ---
>  drivers/net/can/flexcan.c | 61 +++++++++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 61 insertions(+)
> 
> diff --git a/drivers/net/can/flexcan.c b/drivers/net/can/flexcan.c
> index f6f95ae..0ac99c5 100644
> --- a/drivers/net/can/flexcan.c
> +++ b/drivers/net/can/flexcan.c
> @@ -441,6 +441,61 @@ static int flexcan_poll_bus_err(struct net_device *dev, u32 reg_esr)
>  	return 1;
>  }
>  
> +static void flexcan_wait_for_frz(struct net_device *dev, int en)
> +{
> +	struct flexcan_priv *priv = netdev_priv(dev);
> +	struct flexcan_regs __iomem *regs = priv->base;
> +	u32 reg_mcr;
> +	u32 loops = 25;
> +
> +	do {
> +		reg_mcr = flexcan_read(&regs->mcr) & FLEXCAN_MCR_FRZ_ACK;
> +		if ((en && reg_mcr) || (!en && !reg_mcr))
> +			return;
> +
> +		loops--;
> +		udelay(1);
> +	} while (loops);
> +
> +	netdev_err(dev, "Timeout during for the FRZ bit\n");
> +}
> +
> +static void flexcan_reset_err_reg(struct net_device *dev)
> +{
> +	struct flexcan_priv *priv = netdev_priv(dev);
> +	struct flexcan_regs __iomem *regs = priv->base;
> +	u32 reg_mcr;
> +
> +	/*
> +	 * The reset of the error counter is ugly but I don't see any other way.
> +	 * Open the CAN bus, send packet => HW goes to ERR-pasive, TX-err
> +	 * counter is around 128 (maybe 129). Close the CAN-Bus, the TX-err
> +	 * counter drops down to somewhere between 126 … 127, HW goes to
> +	 * ERR-Warning, everything is fine.
> +	 *
> +	 * Now: Repeat the above procedure. What happens is that on send the
> +	 * TX-err increases to around 134…136 and on connect it drops to
> +	 * around 130. Occording to ESR the HW is still in passive mode _but_ it
> +	 * is possible send packets - that means can send packets but has no
> +	 * clue about it.

As far as I know, it's perfectly legal to send CAN frame while in "error
passive" mode. While in error passive mode you cannot send an active
error flag, just a passive one. I don't have a link to the standard at
hand, but have a look at the CAN wiki:

http://www.can-wiki.info/doku.php?id=can_faq_erors

> +	 * To get a consistent behavior here, the error counter are reset so we
> +	 * fall back to Err-Active mode and the second "can send on open bus"
> +	 * behaves just like the first one.
> +	 */
> +
> +	reg_mcr = flexcan_read(&regs->mcr);
> +	reg_mcr |= FLEXCAN_MCR_HALT;
> +	flexcan_write(reg_mcr, &regs->mcr);
> +
> +	flexcan_wait_for_frz(dev, 1);
> +
> +	flexcan_write(0, &regs->ecr);
> +
> +	reg_mcr &= ~FLEXCAN_MCR_HALT;
> +	flexcan_write(reg_mcr, &regs->mcr);
> +	flexcan_wait_for_frz(dev, 0);
> +}
> +
>  static void do_state(struct net_device *dev,
>  		     struct can_frame *cf, enum can_state new_state)
>  {
> @@ -499,6 +554,8 @@ static void do_state(struct net_device *dev,
>  		cf->data[1] = (bec.txerr > bec.rxerr) ?
>  			CAN_ERR_CRTL_TX_WARNING :
>  			CAN_ERR_CRTL_RX_WARNING;
> +
> +		flexcan_reset_err_reg(dev);
>  		break;
>  	case CAN_STATE_ERROR_ACTIVE:
>  		netdev_dbg(dev, "Error Active\n");
> @@ -801,6 +858,7 @@ static int flexcan_chip_start(struct net_device *dev)
>  		FLEXCAN_MCR_MAXMB(FLEXCAN_TX_BUF_ID);
>  	netdev_dbg(dev, "%s: writing mcr=0x%08x", __func__, reg_mcr);
>  	flexcan_write(reg_mcr, &regs->mcr);
> +	flexcan_wait_for_frz(dev, 1);
>  
>  	/*
>  	 * CTRL
> @@ -852,6 +910,7 @@ static int flexcan_chip_start(struct net_device *dev)
>  	reg_mcr = flexcan_read(&regs->mcr);
>  	reg_mcr &= ~FLEXCAN_MCR_HALT;
>  	flexcan_write(reg_mcr, &regs->mcr);
> +	flexcan_wait_for_frz(dev, 0);
>  
>  	priv->can.state = CAN_STATE_ERROR_ACTIVE;
>  
> @@ -885,6 +944,7 @@ static void flexcan_chip_stop(struct net_device *dev)
>  	reg = flexcan_read(&regs->mcr);
>  	reg |= FLEXCAN_MCR_MDIS | FLEXCAN_MCR_HALT;
>  	flexcan_write(reg, &regs->mcr);
> +	flexcan_wait_for_frz(dev, 1);
>  
>  	/* Disable all interrupts */
>  	flexcan_write(0, &regs->imask1);
> @@ -1018,6 +1078,7 @@ static int register_flexcandev(struct net_device *dev)
>  	reg |= FLEXCAN_MCR_FRZ | FLEXCAN_MCR_HALT |
>  		FLEXCAN_MCR_FEN | FLEXCAN_MCR_SUPV;
>  	flexcan_write(reg, &regs->mcr);
> +	flexcan_wait_for_frz(dev, 1);
>  
>  	/*
>  	 * Currently we only support newer versions of this core
> 

Marc

-- 
Pengutronix e.K.                  | Marc Kleine-Budde           |
Industrial Linux Solutions        | Phone: +49-231-2826-924     |
Vertretung West/Dortmund          | Fax:   +49-5121-206917-5555 |
Amtsgericht Hildesheim, HRA 2686  | http://www.pengutronix.de   |


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 242 bytes --]