Black screen with xen dom0 kernel-3.x

[geaaru]

Hi,

I'm trying to understand where could be problem on use nvidia driver with my NVIDIA GTS 250 inside a dom0 kernel 3.x (i test this with 3.0.18, 3.2.0, 3.2.5) with nvidia drivers 290.10 and 295.20.

On compilation of the nvidia kernel module I haven't problem, all works fine but when I start X but on startup monitor go on power off (not for invalid frequency... it seems that nvidia driver doesn't send data on vga output).

From X log I don't see error or others... I try to disable EDID and avoid check:

[ 240.888] (==) NVIDIA(0): Using gamma correction (1.0, 1.0, 1.0)
[ 240.888] (**) NVIDIA(0): Option "UseEDID" "False"
[ 240.888] (**) NVIDIA(0): Option "RenderAccel" "True"
[ 240.888] (**) NVIDIA(0): Option "UseEdidFreqs" "False"
[ 240.888] (**) NVIDIA(0): Option "MetaModes" "CRT-1: 1280x1024_60 +0+0"
[ 240.888] (**) NVIDIA(0): Option "AllowGLXWithComposite" "True"
[ 240.888] (**) NVIDIA(0): Option "AddARGBGLXVisuals" "True"
[ 240.888] (**) NVIDIA(0): Option "ModeValidation" "NoDFPNativeResolutionCheck, NoVirtualSizeCheck,NoMaxPClkCheck,NoHorizSyncCheck ,NoVertRefreshCheck,NoWidthAlignmentCheck"
[ 240.888] (**) NVIDIA(0): Enabling RENDER acceleration
[ 240.888] (**) NVIDIA(0): Enabling 2D acceleration
[ 240.888] (**) NVIDIA(0): Ignoring EDIDs
[ 250.458] (II) NVIDIA(GPU-0): Not probing EDID on CRT-1.
[ 250.462] (II) NVIDIA(0): NVIDIA GPU GeForce GTS 250 (G92) at PCI:1:0:0 (GPU-0)
[ 250.462] (--) NVIDIA(0): Memory: 1048576 kBytes
[ 250.462] (--) NVIDIA(0): VideoBIOS: 62.92.7d.00.00
[ 250.462] (II) NVIDIA(0): Detected PCI Express Link width: 16X
[ 250.462] (--) NVIDIA(0): Interlaced video modes are supported on this GPU
[ 250.462] (--) NVIDIA(0): Connected display device(s) on GeForce GTS 250 at PCI:1:0:0
[ 250.462] (--) NVIDIA(0): CRT-1
[ 250.462] (--) NVIDIA(0): CRT-1: 400.0 MHz maximum pixel clock
[ 253.464] (II) NVIDIA(0): Display Device found referenced in MetaMode: CRT-1
[ 253.464] (**) NVIDIA(0): Ignoring HorizSync/VertRefresh ranges from the EDID for
[ 253.464] (**) NVIDIA(0): display device CRT-1 (Using EDID frequencies has been
[ 253.464] (**) NVIDIA(0): disabled on all display devices.)
[ 253.464] (II) NVIDIA(0): Mode Validation Overrides for CRT-1:
[ 253.464] (II) NVIDIA(0): NoMaxPClkCheck
[ 253.464] (II) NVIDIA(0): NoHorizSyncCheck
[ 253.464] (II) NVIDIA(0): NoVertRefreshCheck
[ 253.464] (II) NVIDIA(0): NoWidthAlignmentCheck
[ 253.464] (II) NVIDIA(0): NoDFPNativeResolutionCheck
[ 253.464] (II) NVIDIA(0): NoVirtualSizeCheck
[ 253.473] (II) NVIDIA(0): Assigned Display Device: CRT-1
[ 253.473] (II) NVIDIA(0): Validated modes:
[ 253.473] (II) NVIDIA(0): "CRT-1:1280x1024_60+0+0"
[ 253.474] (II) NVIDIA(0): Virtual screen size determined to be 1280 x 1024
[ 253.496] (WW) NVIDIA(0): Unable to get display device CRT-1's EDID; cannot compute DPI
[ 253.496] (WW) NVIDIA(0): from CRT-1's EDID.
[ 253.496] (==) NVIDIA(0): DPI set to (75, 75); computed from built-in default
[ 253.496] (**) NVIDIA(0): Enabling 32-bit ARGB GLX visuals.
[ 253.496] (--) Depth 24 pixmap format is 32 bpp
[ 253.496] (II) NVIDIA: Using 768.00 MB of virtual memory for indirect memory access.
[ 256.502] (II) NVIDIA(0): Setting mode "CRT-1:1280x1024_60+0+0"

But doesn't work.

I don't understand where is problem. Same kernel image and nvidia kernel module works fine when i bootstrap kernel outside xen hypervisor... so, this what means? I think that driver and xorg server are correctly.

Could be relative to a wrong pcie pass-through from Xen hypervisor of the video cards?

From dmesg i see correctly irq and address of video card:

[ 219.415959] nvidia: module license 'NVIDIA' taints kernel.
[ 219.415965] Disabling lock debugging due to kernel taint
[ 219.636894] xen: registering gsi 16 triggering 0 polarity 1
[ 219.636900] xen_map_pirq_gsi: returning irq 16 for gsi 16
[ 219.636903] xen: --> pirq=16 -> irq=16 (gsi=16)
[ 219.636911] Already setup the GSI :16
[ 219.636916] nvidia 0000:01:00.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16
[ 219.636929] nvidia 0000:01:00.0: setting latency timer to 64
[ 219.636936] vgaarb: device changed decodes: PCI:0000:01:00.0,olddecodes=io+mem,decodes=nonew ns=io+mem
[ 219.637602] NVRM: loading NVIDIA UNIX x86_64 Kernel Module 295.20 Mon Feb 6 21:07:30 PST 2012
[ 219.637699] NVRM: PAT configuration unsupported, falling back to MTRRs.
[ 250.417269] i2c i2c-4: adapter [NVIDIA i2c adapter 0 at 1:00.0] registered
[ 250.417293] i2c i2c-5: adapter [NVIDIA i2c adapter 1 at 1:00.0] registered
[ 250.417314] i2c i2c-6: adapter [NVIDIA i2c adapter 2 at 1:00.0] registered
[ 259.507057] NVRM: Xid (0000:01:00): 6, PE0001
[ 259.514121] NVRM: Xid (0000:01:00): 6, PE0001
[ 262.456791] NVRM: Xid (0000:01:00): 6, PE007e
[ 262.459002] NVRM: Xid (0000:01:00): 6, PE007e
[ 262.461211] NVRM: Xid (0000:01:00): 6, PE007e
[ 262.463414] NVRM: Xid (0000:01:00): 6, PE007e
[ 262.465606] NVRM: Xid (0000:01:00): 6, PE007e
[ 262.467797] NVRM: Xid (0000:01:00): 6, PE007e

Any suggestions ?

Thanks

P.S. If i use instead kernel 2.6.38-xen on dom0 start correctly.

[geaaru]

Here nvidia-bug-report file. Sorry for delay.

[bguthro]

I am seeing this same issue on 3.2.x kernels - have you found a resolution yet?

[johnnyk]

I'd also like to get the binary driver running under Xen with kernel 3.x