Created attachment 77065 [details] dmesg with drm.debug=0xe Linux boots, but display 'fails' when i915 KMS kicks in. Console input is possible, and system is useable via ssh; X can be launched from console. 'Fails': on tested 2.6 and 3.2 kernels, the entire screen is taken over by a colourful snowstorm. On 3.8.4, the snowstorm flashes briefly then the screen blanks; disconnecting and reconnecting the display again yields a colourful snowstorm. Launching X will also transform the black screen into a snowstorm. I've tried with built-in i915 (fails on transition from initrd), loadable module (fails when i915 loads), and with i915 in initrd (fails immediately). The system is a Acer Revo RL80, which has only DVI-I output. The display is a projector with native resolution 1024x576. With KMS disabled in the kernel, the failure doesn't occur; but i915 cannot be used for X.
Created attachment 77066 [details] Xorg.log
# lsmod Module Size Used by fbcon 33837 71 bitblit 4278 1 fbcon softcursor 937 1 bitblit font 7016 1 fbcon xhci_hcd 77701 0 i915 456317 2 cfbfillrect 3126 1 i915 cfbimgblt 1931 1 i915 cfbcopyarea 2890 1 i915 intel_agp 8521 1 i915 intel_gtt 9951 2 i915,intel_agp drm_kms_helper 23671 1 i915 fb 43517 5 i915,fbcon,drm_kms_helper,softcursor,bitblit fbdev 498 2 fb,fbcon lspci -v (extract): 00:02.0 VGA compatible controller: Intel Corporation 2nd Generation Core Processor Family Integrated Graphics Controller (rev 09) (prog-if 00 [VGA controller]) Subsystem: Acer Incorporated [ALI] Device 074c Flags: bus master, fast devsel, latency 0, IRQ 43 Memory at f7800000 (64-bit, non-prefetchable) [size=4M] Memory at e0000000 (64-bit, prefetchable) [size=256M] I/O ports at f000 [size=64] Expansion ROM at <unassigned> [disabled] Capabilities: [90] MSI: Enable+ Count=1/1 Maskable- 64bit- Capabilities: [d0] Power Management version 2 Capabilities: [a4] PCI Advanced Features Kernel driver in use: i915 # lshw -c video *-display description: VGA compatible controller product: 2nd Generation Core Processor Family Integrated Graphics Controller vendor: Intel Corporation physical id: 2 bus info: pci@0000:00:02.0 version: 09 width: 64 bits clock: 33MHz capabilities: msi pm vga_controller bus_master cap_list rom configuration: driver=i915 latency=0 resources: irq:43 memory:f7800000-f7bfffff memory:e0000000-efffffff ioport:f000(size=64)
Created attachment 77067 [details] Picture of 'colourful snowstorm' display
Nothing really jumps out from the debug logs, seems to be bog-standard dvi. Can you please grab the latest git of intel-gpu-tools from http://cgit.freedesktop.org/xorg/app/intel-gpu-tools/ and then attach the output of intel_reg_dumper both when booting with modeset enabled and with modesetting disabled? Hopefully the register setup points at an obvious differences between i915.ko and what the bios does.
Created attachment 77101 [details] intel_reg_dump with kernel param i915.modeset=0 I did this: mount -t debugfs none /sys/kernel/debug But running the command still gave this error: Couldn't find path to dri/debugfs entry Didn't happen with the run with no kernel params. I hope this is harmless!
Created attachment 77102 [details] intel_reg_dumper with no kernel params Gave no error iff debugfs was mounted.
This is _really_ strange. Both dumps have apparently no mode configured at all. The message about the lack of debugfs is ok for the nomodeset case, we only need that to ask the kernel to wake up the gpu. If you don't use kms, that's not required. Can you try to install the vesa X driver and use that one to adjust the modes with xrandr (obviously requires that kms is disabled). Then pls grab a two reg dumps for different modes. This is just to figure out where exactly the failure is, since according to the reg dumps here neither kms nor non-kms should work at all ...
Created attachment 77113 [details] Intel reg_dump after setting 640x480 xrandr reports: "Failed to get size of gamma for output default" The default mode on launching X is 1024x768.
Created attachment 77114 [details] intel_reg_dump for 1024x768 As before, xrandr complains about not being able to get the size of the gamma. So I launched X, and got 1024x768. I used xrandr to set mode 800x600; then to set mode 640x480; then I took the first reg_dump. Then I set mode 1024x768; then I took the second reg_dump.
Hm, I've misread the dumps :( So now I don't see anything suspicious any longer ...
The dmesg does not log any attempt to actually set the fbcon mode, and the intel_reg_dump for i915.modeset=1 also seems to imply that no fb had been set. Bizarre. Can you please grab a fresh drm.debug=6 dmesg to confirm, and make sure that it is complete?
Created attachment 77125 [details] dmesg from the last vesa test This is the dmesg from the boot corresponding to the foregoing vesa regdumps.
Ah - no debug=0xe. Sorry.
Created attachment 77126 [details] A new dmesg I will redo the regdumps and xrandr exercises under this boot.
Created attachment 77128 [details] reg dump after booting, before X
Created attachment 77129 [details] regs after startx
Created attachment 77130 [details] regs after xrandr mode=800x600
Created attachment 77131 [details] Xorg.log from the same session, after the xrandr step
Sorry, I should have said I wanted the drm.debug=6 dmesg without i915.modeset=0.
Created attachment 77147 [details] dmseg with drm.debug=6, no modeset kernel param
We seem to successfully register the fbdev, but the ->set_par callback never happens, leaving us in that funny state. Can you please attach your kernel .config?
Created attachment 77149 [details] Kernel config
Created attachment 77153 [details] Yet another dmesg I realised that the last dmesg I posted was for a kernel that had KMS disabled by default. This one has KMS enabled; and the behaviour is as I would expect: coloured snowstorm when i915 gets loaded.
Created attachment 77154 [details] Kernel config updated with KMS enabled I also enabled several i2c and GPIO modules. I don't know if GPIO is used, but it seems from the dmesg that i2c *is* used. I don't know what i2c hardware is available, so I enabled (loadable) anything thatn said it was Intel.
FDI link train fail is for Imre. Attachment situation is a bit confusion, the real deal happens in the last one: https://bugs.freedesktop.org/attachment.cgi?id=77153
Wonder if this is a dupe of https://bugs.freedesktop.org/show_bug.cgi?id=60029? Jack, can you try Paulo's patch? If not, I'll cook up a SNB FDI training patch and see if that helps.
Gentlemen, that seems to have worked. I applied the patch to http://cgit.freedesktop.org/~danvet/drm-intel/snapshot/drm-intel-next-2013-03-23.tar.gz, and built and installed the kernel. The system came up with the console in 1024x768. When I launched X, it was still in 1024x768; however I was able to xrandr newmode and addmode to 1024x576 (i.e. native), and then set the native mode. So I guess I should be able to force this in xorg.conf. Thanks a lot guys, I'm very impressed by your responsiveness!
*** This bug has been marked as a duplicate of bug 60029 ***
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.