Re: BUG in drivers/dma/ioat/dma_v2.c:314

[David Woodhouse <dwmw2@infradead.org>]

On Tue, 2010-07-06 at 17:51 -0700, Dan Williams wrote:
> It should always trigger, and I have verified as much with the attached
> replacement patch (by forcing the error on a working system), but we run
> into a new problem.  
> 
> IOMMU: can't find DMAR for device 0000:00:0f.0
> Allocating domain for 0000:00:0f.0 failed

Yeah, we're actually doing the check in the wrong place. This should
work better, I think.

(I'm still only checking for include_all rather than checking for
!cap_write_drain() because that _really_ ought to work; looking at
Chris' latest dmesg it _is_ being assigned to the catch-all unit.)

diff --git a/drivers/pci/intel-iommu.c b/drivers/pci/intel-iommu.c
index c9171be..7df8102 100644
--- a/drivers/pci/intel-iommu.c
+++ b/drivers/pci/intel-iommu.c
@@ -3030,6 +3030,32 @@ static void __init iommu_exit_mempool(void)
 
 }
 
+static void quirk_ioat_snb_no_catchall(struct pci_dev *pdev)
+{
+	struct dmar_drhd_unit *drhd;
+	int i;
+
+	/* We know that this device on this chipset has its own
+	   IOMMU. If we find it under the catch-all IOMMU, then
+	   the BIOS is lying to us. Hope that the IOMMU for
+	   this device is actually disabled, and it needs no
+	   translation... */
+
+	for_each_drhd_unit(drhd) {
+		if (!drhd->include_all)
+			continue;
+		for (i = 0; i < drhd->devices_cnt; i++) {
+			if (drhd->devices[i] == pdev) {
+				WARN_TAINT_ONCE(1, TAINT_FIRMWARE_WORKAROUND,
+						"BIOS wrongly included I/OAT device under catch-all VT-d unit\n");
+				pdev->dev.archdata.iommu = DUMMY_DEVICE_DOMAIN_INFO;
+				return;
+			}
+		}
+	}
+}
+DECLARE_PCI_FIXUP_ENABLE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_IOAT_SNB, quirk_ioat_snb_no_catchall);
+
 static void __init init_no_remapping_devices(void)
 {
 	struct dmar_drhd_unit *drhd;


-- 
David Woodhouse                            Open Source Technology Centre
David.Woodhouse@intel.com                              Intel Corporation