On 15.03.2023 22:10:40, Dario Binacchi wrote:

Add support for the basic extended CAN controller (bxCAN) found in many
low- to middle-end STM32 SoCs. It supports the Basic Extended CAN
protocol versions 2.0A and B with a maximum bit rate of 1 Mbit/s.

The controller supports two channels (CAN1 as master and CAN2 as slave)
and the driver can enable either or both of the channels. They share
some of the required logic (e. g. clocks and filters), and that means
you cannot use the slave CAN without enabling some hardware resources
managed by the master CAN.

Each channel has 3 transmit mailboxes, 2 receive FIFOs with 3 stages and
28 scalable filter banks.
It also manages 4 dedicated interrupt vectors:
- transmit interrupt
- FIFO 0 receive interrupt
- FIFO 1 receive interrupt
- status change error interrupt

Driver uses all 3 available mailboxes for transmission and FIFO 0 for
reception. Rx filter rules are configured to the minimum. They accept
all messages and assign filter 0 to CAN1 and filter 14 to CAN2 in
identifier mask mode with 32 bits width. It enables and uses transmit,
receive buffers for FIFO 0 and error and status change interrupts.

Signed-off-by: Dario Binacchi <dario.binacchi@amarulasolutions.com>
Reviewed-by: Vincent Mailhol <redacted>

[...]

[...]

+
+static inline void bxcan_rmw(struct bxcan_priv *priv, void __iomem *addr,
+			     u32 clear, u32 set)
+{
+	unsigned long flags;
+	u32 old, val;
+
+	spin_lock_irqsave(&priv->rmw_lock, flags);
+	old = readl(addr);
+	val = (old & ~clear) | set;
+	if (val != old)
+		writel(val, addr);
+
+	spin_unlock_irqrestore(&priv->rmw_lock, flags);
+}

I think you don't need the spin_lock anymore, but it's not in the hot
path, so leave it this way.

+

[...]

+static irqreturn_t bxcan_rx_isr(int irq, void *dev_id)
+{
+	struct net_device *ndev = dev_id;
+	struct bxcan_priv *priv = netdev_priv(ndev);
+
+	can_rx_offload_irq_offload_fifo(&priv->offload);
+	can_rx_offload_irq_finish(&priv->offload);
+
+	return IRQ_HANDLED;

This handler is not 100% shared IRQ safe, you have to return IRQ_NONE if
no IRQ is active.

+}
+
+static irqreturn_t bxcan_tx_isr(int irq, void *dev_id)
+{
+	struct net_device *ndev = dev_id;
+	struct bxcan_priv *priv = netdev_priv(ndev);
+	struct bxcan_regs __iomem *regs = priv->regs;
+	struct net_device_stats *stats = &ndev->stats;
+	u32 tsr, rqcp_bit;
+	int idx;
+
+	tsr = readl(&regs->tsr);
+	if (!(tsr & (BXCAN_TSR_RQCP0 | BXCAN_TSR_RQCP1 | BXCAN_TSR_RQCP2)))
+		return IRQ_HANDLED;

Is this a check for no IRQ? Then return IRQ_NONE.

+
+	while (priv->tx_head - priv->tx_tail > 0) {
+		idx = bxcan_get_tx_tail(priv);
+		rqcp_bit = BXCAN_TSR_RQCP0 << (idx << 3);
+		if (!(tsr & rqcp_bit))
+			break;
+
+		stats->tx_packets++;
+		stats->tx_bytes += can_get_echo_skb(ndev, idx, NULL);
+		priv->tx_tail++;
+	}
+
+	writel(tsr, &regs->tsr);
+
+	if (bxcan_get_tx_free(priv)) {
+		/* Make sure that anybody stopping the queue after
+		 * this sees the new tx_ring->tail.
+		 */
+		smp_mb();
+		netif_wake_queue(ndev);
+	}
+
+	return IRQ_HANDLED;
+}
+
+static void bxcan_handle_state_change(struct net_device *ndev, u32 esr)
+{
+	struct bxcan_priv *priv = netdev_priv(ndev);
+	enum can_state new_state = priv->can.state;
+	struct can_berr_counter bec;
+	enum can_state rx_state, tx_state;
+	struct sk_buff *skb;
+	struct can_frame *cf;
+
+	/* Early exit if no error flag is set */
+	if (!(esr & (BXCAN_ESR_EWGF | BXCAN_ESR_EPVF | BXCAN_ESR_BOFF)))
+		return;
+
+	bec.txerr = FIELD_GET(BXCAN_ESR_TEC_MASK, esr);
+	bec.rxerr = FIELD_GET(BXCAN_ESR_REC_MASK, esr);
+
+	if (esr & BXCAN_ESR_BOFF)
+		new_state = CAN_STATE_BUS_OFF;
+	else if (esr & BXCAN_ESR_EPVF)
+		new_state = CAN_STATE_ERROR_PASSIVE;
+	else if (esr & BXCAN_ESR_EWGF)
+		new_state = CAN_STATE_ERROR_WARNING;
+
+	/* state hasn't changed */
+	if (unlikely(new_state == priv->can.state))
+		return;
+
+	skb = alloc_can_err_skb(ndev, &cf);
+
+	tx_state = bec.txerr >= bec.rxerr ? new_state : 0;
+	rx_state = bec.txerr <= bec.rxerr ? new_state : 0;
+	can_change_state(ndev, cf, tx_state, rx_state);
+
+	if (new_state == CAN_STATE_BUS_OFF) {
+		can_bus_off(ndev);
+	} else if (skb) {
+		cf->can_id |= CAN_ERR_CNT;
+		cf->data[6] = bec.txerr;
+		cf->data[7] = bec.rxerr;
+	}
+
+	if (skb) {
+		int err;
+
+		err = can_rx_offload_queue_timestamp(&priv->offload, skb,
+						     priv->timestamp);
+		if (err)
+			ndev->stats.rx_fifo_errors++;
+	}
+}
+
+static void bxcan_handle_bus_err(struct net_device *ndev, u32 esr)
+{
+	struct bxcan_priv *priv = netdev_priv(ndev);
+	enum bxcan_lec_code lec_code;
+	struct can_frame *cf;
+	struct sk_buff *skb;
+
+	lec_code = FIELD_GET(BXCAN_ESR_LEC_MASK, esr);
+
+	/* Early exit if no lec update or no error.
+	 * No lec update means that no CAN bus event has been detected
+	 * since CPU wrote BXCAN_LEC_UNUSED value to status reg.
+	 */
+	if (lec_code == BXCAN_LEC_UNUSED || lec_code == BXCAN_LEC_NO_ERROR)
+		return;
+
+	/* Common for all type of bus errors */
+	priv->can.can_stats.bus_error++;
+
+	/* Propagate the error condition to the CAN stack */
+	skb = alloc_can_err_skb(ndev, &cf);
+	if (skb)
+		cf->can_id |= CAN_ERR_PROT | CAN_ERR_BUSERROR;
+
+	switch (lec_code) {
+	case BXCAN_LEC_STUFF_ERROR:
+		netdev_dbg(ndev, "Stuff error\n");
+		ndev->stats.rx_errors++;
+		if (skb)
+			cf->data[2] |= CAN_ERR_PROT_STUFF;
+		break;
+
+	case BXCAN_LEC_FORM_ERROR:
+		netdev_dbg(ndev, "Form error\n");
+		ndev->stats.rx_errors++;
+		if (skb)
+			cf->data[2] |= CAN_ERR_PROT_FORM;
+		break;
+
+	case BXCAN_LEC_ACK_ERROR:
+		netdev_dbg(ndev, "Ack error\n");
+		ndev->stats.tx_errors++;
+		if (skb) {
+			cf->can_id |= CAN_ERR_ACK;
+			cf->data[3] = CAN_ERR_PROT_LOC_ACK;
+		}
+		break;
+
+	case BXCAN_LEC_BIT1_ERROR:
+		netdev_dbg(ndev, "Bit error (recessive)\n");
+		ndev->stats.tx_errors++;
+		if (skb)
+			cf->data[2] |= CAN_ERR_PROT_BIT1;
+		break;
+
+	case BXCAN_LEC_BIT0_ERROR:
+		netdev_dbg(ndev, "Bit error (dominant)\n");
+		ndev->stats.tx_errors++;
+		if (skb)
+			cf->data[2] |= CAN_ERR_PROT_BIT0;
+		break;
+
+	case BXCAN_LEC_CRC_ERROR:
+		netdev_dbg(ndev, "CRC error\n");
+		ndev->stats.rx_errors++;
+		if (skb) {
+			cf->data[2] |= CAN_ERR_PROT_BIT;
+			cf->data[3] = CAN_ERR_PROT_LOC_CRC_SEQ;
+		}
+		break;
+
+	default:
+		break;
+	}
+
+	if (skb) {
+		int err;
+
+		err = can_rx_offload_queue_timestamp(&priv->offload, skb,
+						     priv->timestamp);
+		if (err)
+			ndev->stats.rx_fifo_errors++;
+	}
+}
+
+static irqreturn_t bxcan_state_change_isr(int irq, void *dev_id)
+{
+	struct net_device *ndev = dev_id;
+	struct bxcan_priv *priv = netdev_priv(ndev);
+	struct bxcan_regs __iomem *regs = priv->regs;
+	u32 msr, esr;
+
+	msr = readl(&regs->msr);
+	if (!(msr & BXCAN_MSR_ERRI))
+		return IRQ_NONE;

No IRQ, the driver returns IRQ_NONE here? Looks good!

+
+	esr = readl(&regs->esr);
+	bxcan_handle_state_change(ndev, esr);
+
+	if (priv->can.ctrlmode & CAN_CTRLMODE_BERR_REPORTING)
+		bxcan_handle_bus_err(ndev, esr);
+
+	msr |= BXCAN_MSR_ERRI;
+	writel(msr, &regs->msr);
+	can_rx_offload_irq_finish(&priv->offload);
+
+	return IRQ_HANDLED;
+}

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:   +49-5121-206917-5555 |