Browse Source

intel-iommu: Force-disable IOMMU for iGFX on broken Cantiga revisions.

Certain revisions of this chipset appear to be broken. There is a shadow
GTT which mirrors the real GTT but contains pre-translated physical
addresses, for performance reasons. When a GTT update happens, the
translations are done once and the resulting physical addresses written
back to the shadow GTT.

Except sometimes, the physical address is actually written back to the
_real_ GTT, not the shadow GTT. Thus we start to see faults when that
physical address is fed through translation again.

Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
David Woodhouse 15 years ago
parent
commit
2d9e667efd
1 changed files with 7 additions and 1 deletions
  1. 7 1
      drivers/pci/intel-iommu.c

+ 7 - 1
drivers/pci/intel-iommu.c

@@ -340,7 +340,7 @@ int dmar_disabled = 0;
 int dmar_disabled = 1;
 #endif /*CONFIG_DMAR_DEFAULT_ON*/
 
-static int __initdata dmar_map_gfx = 1;
+static int dmar_map_gfx = 1;
 static int dmar_forcedac;
 static int intel_iommu_strict;
 
@@ -3721,6 +3721,12 @@ static void __devinit quirk_iommu_rwbf(struct pci_dev *dev)
 	 */
 	printk(KERN_INFO "DMAR: Forcing write-buffer flush capability\n");
 	rwbf_quirk = 1;
+
+	/* https://bugzilla.redhat.com/show_bug.cgi?id=538163 */
+	if (dev->revision == 0x07) {
+		printk(KERN_INFO "DMAR: Disabling IOMMU for graphics on this chipset\n");
+		dmar_map_gfx = 0;
+	}
 }
 
 DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x2a40, quirk_iommu_rwbf);