On 2014/11/6 18:01, Thomas Gleixner wrote:

On Tue, 4 Nov 2014, Jiang Liu wrote:

quoted

+#ifdef	CONFIG_PCI_MSI_IRQ_DOMAIN
+static inline irq_hw_number_t
+msi_get_hwirq(struct pci_dev *pdev, struct msi_desc *msidesc)
+{
+	return (irq_hw_number_t)msidesc->msi_attrib.entry_nr |
+		PCI_DEVID(pdev->bus->number, pdev->devfn) << 11 |
+		(pci_domain_nr(pdev->bus) & 0xFFFFFFFF) << 27;
+}
+
+static int msi_domain_alloc(struct irq_domain *domain, unsigned int virq,
+			    unsigned int nr_irqs, void *arg)
+{
+	int i, ret;
+	irq_hw_number_t hwirq = arch_msi_irq_domain_get_hwirq(arg);
+
+	if (irq_find_mapping(domain, hwirq) > 0)
+		return -EEXIST;
+
+	ret = irq_domain_alloc_irqs_parent(domain, virq, nr_irqs, arg);
+	if (ret >= 0)
+		for (i = 0; i < nr_irqs; i++) {
+			irq_domain_set_hwirq_and_chip(domain, virq + i,
+					hwirq + i, &msi_chip, (void *)(long)i);

I think msi_chip being a global unique thing is problematic. It does
not allow multi platform kernels to select a chip at boot time and it
does not allow per domain chip implementations when you have multiple
msi domains. Aside of that msi_chip is a pretty bad name for a global.

The solution is rather simple and msi is wide spread enough to justify
that.

struct irqdomain_msi_data {
       struct irq_chip       *irq_chip;
};

We make that a struct so we can accomodate for other special things
which might be domain rather than architecture specific. One
obvious use case would be to hold the arch_msi_irq_domain_get/set_hwirq
callbacks.

struct irq_domain *msi_create_irq_domain(struct irq_domain *parent,
       		  			 struct irqdomain_msi_data *data)
{
        struct irq_domain *domain;

        domain = irq_domain_add_tree(NULL, &msi_domain_ops, NULL);
        if (domain) {
                domain->parent = parent;
		domain->msi_data = data;
	}
        return domain;
}

Now the above becomes:

    	struct irq_chip *msi_chip = domain->msi_data->irq_chip;

	irq_domain_set_hwirq_and_chip(domain, virq + i,
				      hwirq + i, msi_chip, (void *)(long)i);

Hi Thomas,
	Actually I'm working on a patch set to improve MSI support in
the way you described above this afternoon. And I'm also trying to
split MSI code into PCI dependent part and PCI independent part.
I plan to add a file kernel/irq/msi.c to host PCI independent part,
is that OK? Or should I put it under something like drivers/msi/?
The PCI indepenent part will be used to support DMAR/HPET/HTIRQ and
some ARM/ARM64 interrupts.
Regards!
Gerry

quoted

+int msi_irq_domain_alloc_irqs(struct irq_domain *domain, int type,
+			      struct pci_dev *dev, void *arg)
+{
+	int i, virq;
+	struct msi_desc *msidesc;
+	int node = dev_to_node(&dev->dev);
+
+	list_for_each_entry(msidesc, &dev->msi_list, list) {
+		arch_msi_irq_domain_set_hwirq(arg, msi_get_hwirq(dev, msidesc));

The arch_xxx callbacks want to be documented. It's not obvious what
they are supposed to do.

Thanks,

	tglx
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo at vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/