On T500 with 2.6.32.1 agpgart sometimes fails to initialize which results in displaying errors: [ 4.276354] Linux agpgart interface v0.103 [ 4.738113] [drm] Initialized drm 1.1.0 20060810 [ 5.075234] i915 0000:00:02.0: power state changed by ACPI to D0 [ 5.075241] i915 0000:00:02.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16 [ 5.075244] i915 0000:00:02.0: setting latency timer to 64 [ 5.075260] [drm:drm_fill_in_dev] *ERROR* Cannot initialize the agpgart module. [ 5.075265] i915 0000:00:02.0: PCI INT A disabled [ 5.075269] i915: probe of 0000:00:02.0 failed with error -22 [ 5.075320] agpgart-intel 0000:00:00.0: Intel Mobile Intel® GM45 Express Chipset [ 5.076833] agpgart-intel 0000:00:00.0: detected 32764K stolen memory [ 5.080520] agpgart-intel 0000:00:00.0: AGP aperture is 256M @ 0xd0000000 As oppose to: [ 3.053571] Linux agpgart interface v0.103 [ 3.189171] agpgart-intel 0000:00:00.0: Intel Mobile Intel® GM45 Express Chipset [ 3.190737] agpgart-intel 0000:00:00.0: detected 32764K stolen memory [ 3.194308] agpgart-intel 0000:00:00.0: AGP aperture is 256M @ 0xd0000000 [ 4.281443] [drm] Initialized drm 1.1.0 20060810 [ 4.343881] i915 0000:00:02.0: power state changed by ACPI to D0 [ 4.343893] i915 0000:00:02.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16 [ 4.343896] i915 0000:00:02.0: setting latency timer to 64 [ 4.357989] i915 0000:00:02.0: irq 31 for MSI/MSI-X [ 4.357995] [drm] set up 31M of stolen space [ 4.635800] fb0: inteldrmfb frame buffer device [ 4.638952] [drm] Initialized i915 1.6.0 20080730 for 0000:00:02.0 on minor 0 [ 5.918313] [drm] LVDS-8: set mode 1680x1050 c (Found by grep -P '(i9|agp|drm)')
I suppose it didn't happen with 2.6.31.x?
I don't remember any problems with 2.6.31 or 2.6.32. However it happend only twice so it's rather rare (however it does happen). X11 starts in such case but have artifacts on gdm and gnome+gnome-shell does not want to start.
So from the kernel timing, the failure case is that drm/i915 loads before agp/intel? which will be sure to be broken...
Still valid in 2.6.32.8 (as patched in Debian source package linux-2.6/2.6.32-8); different but similar trace (drm/i915 loads before agp/intel, although init messages are interleaved somehow): $ dmesg|grep -P '(i9|agp|drm)' [ 0.000000] Command line: BOOT_IMAGE=/vmlinuz-2.6.32-2-amd64 root=/dev/mapper/talisker-root ro cryptopts=source=/dev/hda2,lvm=talisker-root i915.powersave=0 splash [ 0.000000] Kernel command line: BOOT_IMAGE=/vmlinuz-2.6.32-2-amd64 root=/dev/mapper/talisker-root ro cryptopts=source=/dev/hda2,lvm=talisker-root i915.powersave=0 splash [ 0.843168] Linux agpgart interface v0.103 [ 1.396964] [drm] Initialized drm 1.1.0 20060810 [ 2.589844] i915 0000:00:02.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16 [ 2.600506] i915 0000:00:02.0: setting latency timer to 64 [ 2.600537] [drm:drm_fill_in_dev] *ERROR* Cannot initialize the agpgart module. [ 2.631231] i915 0000:00:02.0: PCI INT A disabled [ 2.641185] i915: probe of 0000:00:02.0 failed with error -22 [ 2.651092] agpgart-intel 0000:00:00.0: Intel 945GM Chipset [ 2.661591] agpgart-intel 0000:00:00.0: detected 7932K stolen memory [ 2.674992] agpgart-intel 0000:00:00.0: AGP aperture is 256M @ 0xd0000000
Created attachment 25179 [details] force intel agp dependence for i915 Could you test if this patch works for you?
Ok. I didn't reproduce problem on 2.6.32 (although others did) but I just installed the pached kernel.
Said commit is upstream as: commit 1f7a6e372e9cb4d749f34c0738d832e6cadb4071 Author: Zhenyu Wang <zhenyuw@linux.intel.com> Date: Tue Feb 23 14:05:24 2010 +0800 drm/i915: Add dependency on the intel agp module Please reopen if this did not fix your issue. Is this patch needed for 2.6.32.y? I don't see it there.