On Wed, Jul 17, 2024 at 03:55:10PM -0500, Wei Huang wrote:
> From: Manoj Panicker <manoj.panicker2@xxxxxxx>
>
> Implement TPH support in Broadcom BNXT device driver by invoking
> pcie_tph_set_st() function when interrupt affinity is changed.
*and* invoking pcie_tph_set_st() when setting up the IRQ in the first
place, I guess?
I guess this gives a significant performance benefit? The series
includes "pci=nostmode" so the benefit can be quantified, so now I'm
curious about what you measured :)
> +static void bnxt_rtnl_lock_sp(struct bnxt *bp);
> +static void bnxt_rtnl_unlock_sp(struct bnxt *bp);
These duplicate declarations can't be right, can they? OK for
work-in-progress, but it doesn't look like the final solution.
> +static void __bnxt_irq_affinity_notify(struct irq_affinity_notify *notify,
> + const cpumask_t *mask)
> +{
> + struct bnxt_irq *irq;
> +
> + irq = container_of(notify, struct bnxt_irq, affinity_notify);
> + cpumask_copy(irq->cpu_mask, mask);
> +
> + if (!pcie_tph_set_st(irq->bp->pdev, irq->msix_nr,
> + cpumask_first(irq->cpu_mask),
> + TPH_MEM_TYPE_VM, PCI_TPH_REQ_TPH_ONLY))
> + netdev_dbg(irq->bp->dev, "error in setting steering tag\n");
> +
> + if (netif_running(irq->bp->dev)) {
> + rtnl_lock();
> + bnxt_close_nic(irq->bp, false, false);
> + bnxt_open_nic(irq->bp, false, false);
> + rtnl_unlock();
> + }
> +}
