On Wednesday 10 February 2010 11:34:34 am Larry Finger wrote: > On 02/10/2010 11:51 AM, Bjorn Helgaas wrote: > > Thanks a lot! Your box has 17 windows, which would overflow the > > 16-entry table, EXCEPT that we failed to insert this resource: > > [mem 0x000d0000-0x000d3fff]. > > > > Can you also send the /proc/iomem contents? That should show why this > > is allocated. > > > > Maybe there's a recent framebuffer change or something that accounts > > for this change. > > finger@larrylap:~/wireless-testing> cat /proc/iomem > > 00000000-0009dfff : System RAM > 0009e000-0009ffff : reserved > 000d2000-000fffff : reserved This 0xd2000-0xfffff entry is what conflicted with the host bridge window, and I think it came from the e820 memory map: BIOS-e820: 00000000000d2000 - 0000000000100000 (reserved) ... pci_root PNP0A08:00: can't allocate host bridge window [mem 0x000d0000-0x000d3fff] There might have been an e820-related change that triggered this. In any case, I think that if you were to change PCI_BUS_NUM_RESOURCES from 16 to 12 (or even to 15), your system would stop working with "pci=use_crs". If we then added my patches to replace the table with a list, I think it would work again. You don't need to actually try that; I think we understand the problem well enough already. I really appreciate all your help! Bjorn > 00100000-bbf4ffff : System RAM > 01000000-013169fa : Kernel code > 013169fb-01835f3f : Kernel data > 018cf000-02404667 : Kernel bss > bbf50000-bbf64fff : ACPI Tables > bbf65000-bbf65fff : ACPI Non-volatile Storage > bbf66000-bfffffff : reserved > c0000000-c001ffff : 0000:00:12.0 > d0000000-dfffffff : 0000:00:12.0 > d0000000-d3ffffff : vesafb > e0000000-efffffff : reserved > e0000000-efffffff : pnp 00:03 > e0000000-e09fffff : PCI MMCONFIG 0000 [bus 00-09] > f0000000-f0ffffff : 0000:00:12.0 > f4000000-f4ffffff : 0000:00:12.0 > f8000000-fbffffff : PCI Bus 0000:06 > fc000000-fc0fffff : PCI Bus 0000:04 > fc000000-fc003fff : 0000:04:00.0 > fc000000-fc003fff : 0000:04:00.0 > fc100000-fc1fffff : PCI Bus 0000:01 > fc100000-fc1007ff : 0000:01:09.0 > fc100800-fc1008ff : 0000:01:09.1 > fc100c00-fc100cff : 0000:01:09.2 > fc101000-fc1010ff : 0000:01:09.3 > fc101400-fc1014ff : 0000:01:09.4 > fc200000-fc27ffff : 0000:00:01.3 > fc480000-fc483fff : 0000:00:07.0 > fc480000-fc483fff : ICH HD audio > fc484000-fc485fff : 0000:00:09.0 > fc484000-fc485fff : ahci > fc486000-fc486fff : 0000:00:02.0 > fc486000-fc486fff : ohci_hcd > fc487000-fc487fff : 0000:00:04.0 > fc487000-fc487fff : ohci_hcd > fc488000-fc488fff : 0000:00:0a.0 > fc488000-fc488fff : forcedeth > fc489000-fc4890ff : 0000:00:02.1 > fc489000-fc4890ff : ehci_hcd > fc489400-fc4894ff : 0000:00:04.1 > fc489400-fc4894ff : ehci_hcd > fc489800-fc48980f : 0000:00:0a.0 > fc489800-fc48980f : forcedeth > fc489c00-fc489cff : 0000:00:0a.0 > fc489c00-fc489cff : forcedeth > fec00000-fec0ffff : reserved > fec00000-fec003ff : IOAPIC 0 > fec80000-fec80fff : pnp 00:00 > fed00000-fed003ff : HPET 0 > fee00000-fee00fff : Local APIC > fee00000-fee00fff : reserved > fef00000-fef00fff : pnp 00:00 > fff80000-ffffffff : reserved > > This version is from 2.6.33-rc7, but from the wireless-testing tree as I had > rebooted. I think it should be the same, but if you need the version from the > other kernel, let me know. > > Larry > -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
