x86/PCI: Add "pci=big_root_window" option for AMD 64-bit windows

Only try to enable a 64-bit window on AMD CPUs when "pci=big_root_window"
is specified.

This taints the kernel because the new 64-bit window uses address space we
don't know anything about, and it may contain unreported devices or memory
that would conflict with the window.

The pci_amd_enable_64bit_bar() quirk that enables the window is specific to
AMD CPUs.  The generic solution would be to have the firmware enable the
window and describe it in the host bridge's _CRS method, or at least
describe it in the _PRS method so the OS would have the option of enabling
it.

Signed-off-by: Christian König <christian.koenig@amd.com>
[bhelgaas: changelog, extend doc, mention taint in dmesg]
Signed-off-by: Bjorn Helgaas <helgaas@kernel.org>
diff --git a/arch/x86/pci/fixup.c b/arch/x86/pci/fixup.c
index e663d6b..8bad19c 100644
--- a/arch/x86/pci/fixup.c
+++ b/arch/x86/pci/fixup.c
@@ -667,6 +667,9 @@ static void pci_amd_enable_64bit_bar(struct pci_dev *dev)
 	struct resource *res, *conflict;
 	struct pci_dev *other;
 
+	if (!(pci_probe & PCI_BIG_ROOT_WINDOW))
+		return;
+
 	/* Check that we are the only device of that type */
 	other = pci_get_device(dev->vendor, dev->device, NULL);
 	if (other != dev ||
@@ -714,7 +717,9 @@ static void pci_amd_enable_64bit_bar(struct pci_dev *dev)
 		res->start = conflict->end + 1;
 	}
 
-	dev_info(&dev->dev, "adding root bus resource %pR\n", res);
+	dev_info(&dev->dev, "adding root bus resource %pR (tainting kernel)\n",
+		 res);
+	add_taint(TAINT_FIRMWARE_WORKAROUND, LOCKDEP_STILL_OK);
 
 	base = ((res->start >> 8) & AMD_141b_MMIO_BASE_MMIOBASE_MASK) |
 		AMD_141b_MMIO_BASE_RE_MASK | AMD_141b_MMIO_BASE_WE_MASK;