Yinghai Lu <yinghai@xxxxxxxxxx> writes:
> Iranna D Ankad reported that IBM x3950 systems have boot problems
> after this commit:
>
> |
> | commit b9c61b70075c87a8612624736faf4a2de5b1ed30
> |
> | x86/pci: update pirq_enable_irq() to setup io apic routing
> |
>
> As explained in the previous patch ("x86: Fix out of order gsi - partial)
> try to remap those gsis
>
> This patch adds boot_ioapic_idx and gsi_to_irq/irq_to_gsi
A couple of nits, that we will eventually want to fix.
> Index: linux-2.6/arch/x86/kernel/apic/io_apic.c
> ===================================================================
> --- linux-2.6.orig/arch/x86/kernel/apic/io_apic.c
> +++ linux-2.6/arch/x86/kernel/apic/io_apic.c
> @@ -97,6 +97,41 @@ int mp_irq_entries;
> /* GSI interrupts */
> static int nr_irqs_gsi = NR_IRQS_LEGACY;
>
> +/* By default isa irqs are identity mapped to gsis */
> +unsigned int isa_irq_to_gsi[NR_IRQS_LEGACY] = {
> + 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15
> +};
> +
> +static int boot_ioapic_idx;
> +static int gsi_delta;
> +int gsi_to_irq(unsigned int gsi)
> +{
> + unsigned int irq = gsi;
> + unsigned int i;
> +
> + irq += gsi_delta;
> + for (i = 0; i < NR_IRQS_LEGACY; i++) {
> + if (isa_irq_to_gsi[i] == gsi) {
> + irq = i;
> + break;
> + }
> + }
> +
> + return irq;
> +}
> +
> +unsigned int irq_to_gsi(int irq)
> +{
> + unsigned int gsi;
> +
> + if (irq < NR_IRQS_LEGACY)
> + gsi = isa_irq_to_gsi[irq];
> + else
> + gsi = irq - gsi_delta;
> +
> + return gsi;
> +}
This should really live in arch/x86/kernel/acpi/boot.c or similar as
it has everything to do with acpi and gsis and nothing to do with the
ioapics.
Not this merge window, but ultimately we want a fixed value of 16 for
gsi_delta, and we want to always use it. This has a greater chance of
breaking things but it will ensure in the long run that we flush out
every place that actually needs translation from irqs to gsis.
> Index: linux-2.6/drivers/pnp/pnpacpi/rsparser.c
> ===================================================================
> --- linux-2.6.orig/drivers/pnp/pnpacpi/rsparser.c
> +++ linux-2.6/drivers/pnp/pnpacpi/rsparser.c
> @@ -123,6 +123,14 @@ static void pnpacpi_parse_allocated_irqr
> }
>
> flags = irq_flags(triggering, polarity, shareable);
> +#ifdef CONFIG_X86_IO_APIC
> + /*
> + * looks like IBM x3950 is using irq instead of gsi etc...
> + * convert it back at first
> + */
> + if (acpi_irq_model == ACPI_IRQ_MODEL_IOAPIC)
> + gsi = irq_to_gsi(gsi);
> +#endif
I would like to understand this better. I suspect the actual problem
is that we are getting passed bus_irq instead of gsi from acpi. If we
are getting passed the bus_irq than doing this just for x3950 is wrong.
I really dislike having special cases for a specific motherboard. Both
because they are hard to maintain (as quickly no one has that board) and
because may times they are a symptom of a bug elsewhere in the code
that we are trying to patch over instead of actually fix properly.
> irq = acpi_register_gsi(&dev->dev, gsi, triggering, polarity);
> if (irq >= 0)
> pcibios_penalize_isa_irq(irq, 1);
Eric
--
To unsubscribe from this list: send the line "unsubscribe linux-tip-commits" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
|