Messages in this thread Patch in this message | | | Date | Mon, 20 Jan 2003 04:53:09 -0800 | From | William Lee Irwin III <> | Subject | Re: pci_child_fixup() |
| |
On Sun, Jan 19, 2003 at 07:52:17PM -0800, William Lee Irwin III wrote: >> child = pci_add_new_bus(bus, dev, 0); >> - child->primary = buses & 0xFF; >> - child->secondary = (buses >> 8) & 0xFF; >> - child->subordinate = (buses >> 16) & 0xFF; >> - child->number = child->secondary; >> + pci_child_fixup(bus, child, buses);
On Mon, Jan 20, 2003 at 02:57:54PM +0300, Ivan Kokshaysky wrote: > The "bus" argument seems to be redundant. Why not use "child->parent" > just filled in by pci_add_new_bus() instead? > Ivan.
Good point. Here's an adjusted patch. I added gregkh to the cc: list in case he's interested (he generally seems to be wrt. PCI).
Thanks, Bill
The NUMA-Q BIOS reports bus numbers different from the physical ones programmed into PCI-PCI bridges etc. PCI config cycles (and almost everything else) are meant to use translation tables handed to us by the BIOS. No other extant i386 port has such a beast, and there's currently no way to override bus numbers gotten from bus number registers in a meaningful way; bus number conflicts error out prior to any call into arch code. All of the IRQ routing, interrupt assignment, IO-APIC, PCI bus, etc. etc. MP table entries use these numbers, so we're stuck with them.
This provides a fixup hook so it's possible to override the standard interpretation of the bus number registers in PCI-PCI bridges.
arch/i386/pci/numa.c | 9 +++++++++ drivers/pci/probe.c | 5 +---- include/asm-i386/pci.h | 11 +++++++++++ 3 files changed, 21 insertions(+), 4 deletions(-)
diff -urpN mpc-2.5.59-1/arch/i386/pci/numa.c mpc-2.5.59-2/arch/i386/pci/numa.c --- mpc-2.5.59-1/arch/i386/pci/numa.c 2003-01-19 19:01:38.000000000 -0800 +++ mpc-2.5.59-2/arch/i386/pci/numa.c 2003-01-20 04:37:02.000000000 -0800 @@ -88,6 +88,15 @@ static struct pci_ops pci_direct_conf1_m .write = pci_conf1_mq_write }; +void pci_child_fixup(struct pci_bus *child, int buses) +{ + int quad = BUS2QUAD(child->parent->number); + child->primary = QUADLOCAL2BUS(quad, buses & 0xFF); + child->secondary = QUADLOCAL2BUS(quad, (buses >> 8) & 0xFF); + child->subordinate = QUADLOCAL2BUS(quad, (buses >> 16) & 0xFF); + child->number = child->secondary; +} + static void __devinit pci_fixup_i450nx(struct pci_dev *d) { diff -urpN mpc-2.5.59-1/drivers/pci/probe.c mpc-2.5.59-2/drivers/pci/probe.c --- mpc-2.5.59-1/drivers/pci/probe.c 2003-01-16 18:22:24.000000000 -0800 +++ mpc-2.5.59-2/drivers/pci/probe.c 2003-01-20 04:37:32.000000000 -0800 @@ -271,10 +271,7 @@ int __devinit pci_scan_bridge(struct pci if (pass) return max; child = pci_add_new_bus(bus, dev, 0); - child->primary = buses & 0xFF; - child->secondary = (buses >> 8) & 0xFF; - child->subordinate = (buses >> 16) & 0xFF; - child->number = child->secondary; + pci_child_fixup(child, buses); cmax = pci_do_scan_bus(child); if (cmax > max) max = cmax; } else { diff -urpN mpc-2.5.59-1/include/asm-i386/pci.h mpc-2.5.59-2/include/asm-i386/pci.h --- mpc-2.5.59-1/include/asm-i386/pci.h 2003-01-16 18:22:04.000000000 -0800 +++ mpc-2.5.59-2/include/asm-i386/pci.h 2003-01-20 04:39:12.000000000 -0800 @@ -100,6 +100,17 @@ static inline int pci_controller_num(str extern int pci_mmap_page_range(struct pci_dev *dev, struct vm_area_struct *vma, enum pci_mmap_state mmap_state, int write_combine); +#ifdef CONFIG_X86_NUMAQ +void pci_child_fixup(struct pci_bus *, int); +#else +static inline void pci_child_fixup(struct pci_bus *child, int buses) +{ + child->primary = buses & 0xFF; + child->secondary = (buses >> 8) & 0xFF; + child->subordinate = (buses >> 16) & 0xFF; + child->number = child->secondary; +} +#endif #endif /* __KERNEL__ */ /* implement the pci_ DMA API in terms of the generic device dma_ one */ - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |