On Thu, 2008-04-10 at 10:58 -0600, Matthew Wilcox wrote: > On Thu, Apr 10, 2008 at 12:08:15PM +0100, TJ wrote: > > http://bugzilla.kernel.org/show_bug.cgi?id=10396 > > > > The fix-up checks the PCI config for the subsidiary buses and if it > > finds them scans them. This adds them to the root_pci_bus list. Later in > > the boot process the ACPI/PCI code reads the ACPI DSDT table, finds the > > PCI bus entries (PNP0A03) and tries to scan them. It fails when scanning > > the 2nd and 3rd buses with: > > > > [ 0.910906] ACPI: PCI Root Bridge [PX0B] (0000:02) > > [ 0.912085] ACPI: Bus 0000:02 not present in PCI namespace > > [ 0.917111] ACPI: PCI Root Bridge [PX1A] (0000:03) > > [ 0.920085] ACPI: Bus 0000:03 not present in PCI namespace > > > > Unfortunately, the report is misleading since the reason is that the bus > > is found to be already registered and therefore ignored. The situation > > can be worked around by booting with "pci=noacpi". > > I agree with your analysis of the problem. > Of course the patch can fix this problem. In fact we have another similar bug. http://bugzilla.kernel.org/show_bug.cgi?id=10124 The patch is already added to the -mm tree. The patch title is : acpi-unneccessary-to-scan-the-pci-bus-already-scanned.patch Thanks. Yakui > > The solution is to make the pci_fixup_i450nx() code selective based on > > the DMI of the system. I've introduced a patch that does this. Initially > > the only DMI it will match is Dell PowerEdge 6300 but if other systems > > are found to be affected the output of "dmidecode" should be captured > > and reported. Additional DMI_MATCH entries can then be added to the > > patch. > > I disagree with your solution -- it's a bit clumsy and only solves this > one system. How about this solution? > > ---- > > The PCI root bus that ACPI wants to scan may have already been discovered > by another means. In the case of Bugzilla 10396, it was found by > pci_fixup_i450nx(), but there are many other ways to find root busses. > > The problem is that pci_scan_bus() will return NULL both for busses > which don't exist and for busses which were already scanned. We could > decide to make pci_scan_bus() return an ERR_PTR() to allow callers to > distinguish the two situations, but it is simpler for ACPI to check > whether the bus already exists. > > Signed-off-by: Matthew Wilcox <willy@xxxxxxxxxxxxxxx> > > diff --git a/drivers/acpi/pci_root.c b/drivers/acpi/pci_root.c > index c3fed31..954e314 100644 > --- a/drivers/acpi/pci_root.c > +++ b/drivers/acpi/pci_root.c > @@ -302,6 +302,8 @@ static int __devinit acpi_pci_root_add(struct acpi_device *device) > * thus the root bridge's pci_dev does not exist). > */ > root->bus = pci_acpi_scan_root(device, root->id.segment, root->id.bus); > + if (!root->bus) > + root->bus = pci_find_bus(root->id.segment, root->id.bus); > if (!root->bus) { > printk(KERN_ERR PREFIX > "Bus %04x:%02x not present in PCI namespace\n", > -- To unsubscribe from this list: send the line "unsubscribe linux-acpi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
