Myles Watson wrote: > Initialize the interrupts even if you don't generate the MP_TABLE. I > just copied the values from mptable.c. > Also set IRQ 9 to be edge-triggered to make Linux stop complaining. > > Signed-off-by: Myles Watson <[email protected]>
Acked-by: Peter Stuge <[email protected]> But some comments: > +++ cbv2/src/mainboard/tyan/s2891/acpi_tables.c > @@ -42,6 +42,19 @@ unsigned long acpi_fill_madt(unsigned lo > apic_addr = pci_read_config32(dev, PCI_BASE_ADDRESS_1) & ~0xf; > current += acpi_create_madt_ioapic((acpi_madt_ioapic_t *) > current, 4, > apic_addr, 0); > + #if !CONFIG_HAVE_MP_TABLE /* Initialize interrupt mapping. */ > + { > + u32 dword; > + dword = 0x0120d218; > + pci_write_config32(dev, 0x7c, dword); I don't think this is how Ron meant he likes CONFIG_ values in code. More like: if(!CONFIG_HAVE_MP_TABLE) { } Personally I favor letting the preprocessor do preprocessing, rather than pushing this type of code removal to the compiler. > @@ -61,7 +74,7 @@ unsigned long acpi_fill_madt(unsigned lo > > /* IRQ9 ACPI active low. */ > current += acpi_create_madt_irqoverride((acpi_madt_irqoverride_t *) > - current, 0, 9, 9, MP_IRQ_TRIGGER_LEVEL | MP_IRQ_POLARITY_LOW); > + current, 0, 9, 9, MP_IRQ_TRIGGER_EDGE | MP_IRQ_POLARITY_LOW); This is definately something that should not be unique per board. I expect not just Tyan boards will have this code. Could it be moved out to a common place? //Peter -- coreboot mailing list: [email protected] http://www.coreboot.org/mailman/listinfo/coreboot
