Re: [v5,4/4] stmmac: s32: enable support for Multi-IRQ mode

[Simon Horman]

This is an AI-generated review of your patch. The human sending this
email has considered the AI review valid, or at least plausible.

For local reproduction steps see https://netdev-ai.bots.linux.dev/ai-local.html
---
stmmac: s32: enable support for Multi-IRQ mode

This commit enables Multi-IRQ mode for the S32 DWMAC driver, allowing
per-queue IRQs for improved performance. The driver checks for multiple
queues and per-queue IRQs, enabling multi-IRQ mode when both are present.

> diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac-s32.c b/drivers/net/ethernet/stmicro/stmmac/dwmac-s32.c
> index af594a096676..4aad7077c9d6 100644
> --- a/drivers/net/ethernet/stmicro/stmmac/dwmac-s32.c
> +++ b/drivers/net/ethernet/stmicro/stmmac/dwmac-s32.c

[ ... ]

> @@ -165,6 +165,16 @@ static int s32_dwmac_probe(struct platform_device *pdev)
>  	plat->core_type = DWMAC_CORE_GMAC4;
>  	plat->pmt = 1;
>  	plat->flags |= STMMAC_FLAG_SPH_DISABLE;
> +
> +	/* Check for multi-IRQ config. Assumption: symmetrical rx/tx queues */
> +	if (plat->rx_queues_to_use > 1 &&
> +	    (res.rx_irq[0] > 0 && res.tx_irq[0] > 0)) {
> +		plat->flags |= STMMAC_FLAG_MULTI_MSI_EN;
> +		dev_info(dev, "Multi-IRQ mode (per queue IRQs) selected\n");
> +	} else {
> +		dev_info(dev, "MAC IRQ mode selected\n");
> +	}

Does this validation check all queues have IRQs? The condition verifies
only res.rx_irq[0] and res.tx_irq[0] but doesn't confirm that higher
queues have IRQs when rx_queues_to_use > 1.

If a device tree specifies multiple queues but only provides IRQs for
queue 0:

    snps,rx-queues-to-use = <2>;
    interrupt-names = "macirq", "rx-queue-0", "tx-queue-0";
                      /* queue 1 IRQs missing */

The driver would enable STMMAC_FLAG_MULTI_MSI_EN, but later in
stmmac_request_irq_multi_msi() (drivers/net/ethernet/stmicro/stmmac/
stmmac_main.c), queues with rx_irq[i] == 0 are silently skipped:

    for (i = 0; i < priv->plat->rx_queues_to_use; i++) {
        if (priv->rx_irq[i] <= 0)
            continue;  /* queue 1+ skipped, no handler registered */
        ...
    }

This would leave higher-numbered queues with no interrupt handler,
causing packet loss when traffic arrives on those queues.

Would it be safer to validate all queue IRQs before enabling multi-IRQ
mode, or fall back to MAC IRQ mode if any queue IRQ is missing?

> +
>  	plat->rx_fifo_size = 20480;
>  	plat->tx_fifo_size = 20480;