Bug 44510

I have the same issue with a Lenovo T410s (updated to latest BIOS); I see I have the 8086:0044 device that triggers quirk_calpella_no_shadow_gtt() in intel-iommu.c, but the PCI header is such that I hit the second clause in:

static void __devinit quirk_calpella_no_shadow_gtt(struct pci_dev *dev)
{
        unsigned short ggc;

        if (pci_read_config_word(dev, GGC, &ggc))
                return;

        if (!(ggc & GGC_MEMORY_VT_ENABLED)) {
                printk(KERN_INFO "DMAR: BIOS has allocated no shadow GTT; disabling IOMMU for graphics\n");
                dmar_map_gfx = 0;
        } else if (dmar_map_gfx) {
                /* we have to ensure the gfx device is idle before we flush */
                printk(KERN_INFO "DMAR: Disabling batched IOTLB flush on Ironlake\n");
                intel_iommu_strict = 1;
       }
}

so I see:

[    0.802361] DMAR: Disabling batched IOTLB flush on Ironlake

in my kernel log, but then I get a stream of

[   98.467011] DRHD: handling fault status reg 2
[   98.467014] DMAR:[DMA Write] Request device [00:02.0] fault addr 400020000 
[   98.467038] DMAR:[fault reason 05] PTE Write access is not set

"intel_iommu=igfx_off" resolves this for me as well.

Full lspci:

00:00.0 Host bridge [0600]: Intel Corporation Core Processor DRAM Controller [8086:0044] (rev 02)
00:02.0 VGA compatible controller [0300]: Intel Corporation Core Processor Integrated Graphics Controller [8086:0046] (rev 02)
00:16.0 Communication controller [0780]: Intel Corporation 5 Series/3400 Series Chipset HECI Controller [8086:3b64] (rev 06)
00:16.3 Serial controller [0700]: Intel Corporation 5 Series/3400 Series Chipset KT Controller [8086:3b67] (rev 06)
00:19.0 Ethernet controller [0200]: Intel Corporation 82577LM Gigabit Network Connection [8086:10ea] (rev 06)
00:1a.0 USB controller [0c03]: Intel Corporation 5 Series/3400 Series Chipset USB2 Enhanced Host Controller [8086:3b3c] (rev 06)
00:1b.0 Audio device [0403]: Intel Corporation 5 Series/3400 Series Chipset High Definition Audio [8086:3b57] (rev 06)
00:1c.0 PCI bridge [0604]: Intel Corporation 5 Series/3400 Series Chipset PCI Express Root Port 1 [8086:3b42] (rev 06)
00:1c.1 PCI bridge [0604]: Intel Corporation 5 Series/3400 Series Chipset PCI Express Root Port 2 [8086:3b44] (rev 06)
00:1c.3 PCI bridge [0604]: Intel Corporation 5 Series/3400 Series Chipset PCI Express Root Port 4 [8086:3b48] (rev 06)
00:1d.0 USB controller [0c03]: Intel Corporation 5 Series/3400 Series Chipset USB2 Enhanced Host Controller [8086:3b34] (rev 06)
00:1e.0 PCI bridge [0604]: Intel Corporation 82801 Mobile PCI Bridge [8086:2448] (rev a6)
00:1f.0 ISA bridge [0601]: Intel Corporation 5 Series/3400 Series Chipset LPC Interface Controller [8086:3b0f] (rev 06)
00:1f.2 SATA controller [0106]: Intel Corporation 5 Series/3400 Series Chipset 6 port SATA AHCI Controller [8086:3b2f] (rev 06)
00:1f.3 SMBus [0c05]: Intel Corporation 5 Series/3400 Series Chipset SMBus Controller [8086:3b30] (rev 06)
00:1f.6 Signal processing controller [1180]: Intel Corporation 5 Series/3400 Series Chipset Thermal Subsystem [8086:3b32] (rev 06)
03:00.0 Network controller [0280]: Intel Corporation Centrino Wireless-N 1000 [8086:0084]
ff:00.0 Host bridge [0600]: Intel Corporation Core Processor QuickPath Architecture Generic Non-core Registers [8086:2c62] (rev 02)
ff:00.1 Host bridge [0600]: Intel Corporation Core Processor QuickPath Architecture System Address Decoder [8086:2d01] (rev 02)
ff:02.0 Host bridge [0600]: Intel Corporation Core Processor QPI Link 0 [8086:2d10] (rev 02)
ff:02.1 Host bridge [0600]: Intel Corporation Core Processor QPI Physical 0 [8086:2d11] (rev 02)
ff:02.2 Host bridge [0600]: Intel Corporation Core Processor Reserved [8086:2d12] (rev 02)
ff:02.3 Host bridge [0600]: Intel Corporation Core Processor Reserved [8086:2d13] (rev 02)

In linus/master:

commit 5c0422878fcdc279ae9a8e8b66972a15b5efb67f
Author: Ben Widawsky <ben@bwidawsk.net>
Date:   Mon Oct 17 15:51:55 2011 -0700

    drm/i915: ILK + VT-d workaround
    
    Idle the GPU before doing any unmaps. We know if VT-d is in use through
    an exported variable from iommu code.
    
    This should avoid a known HW issue.
    
    Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
    Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
    Signed-off-by: Keith Packard <keithp@keithp.com>

This does introduce a synchronous flush of the iommu tlb everytime a buffer is released which can cause stalls of ~1s every few seconds... But it does prevent the eventual machine hang. However, I would advise you to continue to use intel_iommu=igfx_off unless you truly do require virtualised graphics.