Re: [PATCH RESEND v4 5/8] can: flexcan: add FLEXCAN_QUIRK_IRQ_BERR quirk

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [High] Concurrent lockless modification of `skb_irq_queue` leading to linked list corruption.
--

commit 5585c21d3d6c362465f049b7368b7a16b3da034f
Author: Ciprian Marian Costea <ciprianmarian.costea@oss.nxp.com>

can: flexcan: add FLEXCAN_QUIRK_IRQ_BERR quirk

This commit introduces FLEXCAN_QUIRK_IRQ_BERR to handle hardware with a
dedicated interrupt line for bus errors and device state changes. It adds
the flexcan_irq_esr() handler and enables/disables the new IRQ line as
needed.

> diff --git a/drivers/net/can/flexcan/flexcan-core.c b/drivers/net/can/flexcan/flexcan-core.c
> index 32e4d4da00a1c..23ddf79106410 100644
> --- a/drivers/net/can/flexcan/flexcan-core.c
> +++ b/drivers/net/can/flexcan/flexcan-core.c
[ ... ]
> @@ -1912,6 +1929,13 @@ static int flexcan_open(struct net_device *dev)
>  			goto out_free_irq_boff;
>  	}
>  
> +	if (priv->devtype_data.quirks & FLEXCAN_QUIRK_IRQ_BERR) {
> +		err = request_irq(priv->irq_err,
> +				  flexcan_irq_esr, IRQF_SHARED, dev->name, dev);

[Severity: High]
This is a pre-existing issue, but this quirk continues a pattern where
separate interrupt handlers are registered on distinct physical IRQ lines.
Is there anything preventing flexcan_irq_esr() from executing concurrently
with the mailbox interrupt handler (flexcan_irq_mb) on a different CPU?

If they run concurrently, can this lead to lockless doubly-linked list
corruption?

The CAN rx_offload framework expects SKB enqueuing functions to be
serialized. However, if flexcan_irq_esr() calls flexcan_do_state() or
flexcan_do_berr() which use can_rx_offload_queue_timestamp(), it will
insert SKBs into offload->skb_irq_queue locklessly via
__skb_queue_add_sort().

At the same time, flexcan_do_mb() running from flexcan_irq_mb() can call
can_rx_offload_irq_offload_timestamp() or
can_rx_offload_get_echo_skb_queue_timestamp(), which also locklessly modify
the same skb_irq_queue.

Could this concurrent access scramble the list's next/prev pointers, leading
to kernel panics or infinite loops during NAPI processing?

> +		if (err)
> +			goto out_free_irq_boff;
> +	}
> +

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260603071342.641874-1-ciprianmarian.costea@oss.nxp.com?part=5