[Qemu-devel] Possible regression for Heathrow (PPC), with workaround

[Stuart Brady]

Hi,

I've noticed a possible regression involving the Heathrow (PPC) machine,
caused by the PCI IRQ sharing patch commited on 2006-09-24.

The Debian Sarge (3.1) installer does not appear to run correctly since
that commit, due to lost interrupts.

(FWIW, I was following http://overselfresearch.com/kb/qemu.html)

Here is a workaround for the issue:

diff -u -r1.3 grackle_pci.c
--- hw/grackle_pci.c    24 Sep 2006 17:01:44 -0000      1.3
+++ hw/grackle_pci.c    15 Jan 2007 18:31:47 -0000
@@ -77,12 +77,14 @@
 /* Don't know if this matches real hardware, but it agrees with OHW.  */
 static int pci_grackle_map_irq(PCIDevice *pci_dev, int irq_num)
 {
-    return (irq_num + (pci_dev->devfn >> 3)) & 3;
+    /* XXX: Workaround */
+    return pci_dev->config[PCI_INTERRUPT_LINE];
 }
 
 static void pci_grackle_set_irq(void *pic, int irq_num, int level)
 {
-    heathrow_pic_set_irq(pic, irq_num + 8, level);
+    /* XXX: Workaround */
+    heathrow_pic_set_irq(pic, irq_num, level);
 }
 
 PCIBus *pci_grackle_init(uint32_t base, void *pic)
 
Cheers,
-- 
Stuart Brady