Missing VT output with all current drivers on QHD+/K2100M

Just had to register to add my feedback.

My company just purchased a few M4800 for a rather large POC using the M4800. Everything mentioned above is exactly what I have been thinking lately. I too was relying on the possibility to fall back to Intel HD if necessary and I checked out the CPU specs at Intel to really make sure it was having Intel integrated GPU. We were expecting things go rather strait forward and then make a rather big purchase for our engineers but this is from now on not a viable way and we are really disappointed. The combo with no Optimus and borked NVidia drivers and a company obviously not interested in fixing those mistakes is priceless.

/Anders

One thing is for sure : the way this is handled - or rather not handled - both by Dell (hidden Optimus design flaw) and NVidia (buggy drivers with no communication, let alone resolution, after six months) has not gone unnoticed with us.

It seems we are forced to conclude that the least troublesome path forward is not to rely on hypothetical improvements/bugfixes in a near or distant future, but to simply rule out the troublesome (mainly non open-source) brands/components from our current/future BoM requirements.

(although this is not a Dell but an NVidia forum, I’d like to note that Dell is currently explicitly advertising the M4800/QHD+/Core i7-4910MQ/K2100M WITH HD4600 support, go figure)

Not sure if this helps, but my problems with black vts are usually because buggy framebuffer drivers(but then a again I have only one graphics card and it’s desktop). What modes does hwinfo --framebuffer gives you(must be run as root).

Though it might be useless for you because of optimus; but I have always needed to follow these instructions to get somewhat working ttys:

Thanks for commenting.

Framebuffer support seems to be very limited :

# hwinfo --framebuffer

02: None 00.0: 11001 VESA Framebuffer                           
  [Created at bios.459]
  Unique ID: rdCR.XjymFJTY1+9
  Hardware Class: framebuffer
  Model: "NVIDIA GK106 Board - 20390503"
  Vendor: "NVIDIA Corporation"
  Device: "GK106 Board - 20390503"
  SubVendor: "NVIDIA"
  SubDevice: 
  Revision: "Chip Rev"
  Memory Size: 14 MB
  Memory Range: 0xf1000000-0xf1dfffff (rw)
  Mode 0x0300: 640x400 (+640), 8 bits
  Mode 0x0301: 640x480 (+640), 8 bits
  Mode 0x0303: 800x600 (+800), 8 bits
  Mode 0x030e: 320x200 (+640), 16 bits
  Mode 0x030f: 320x200 (+1280), 24 bits
  Mode 0x0311: 640x480 (+1280), 16 bits
  Mode 0x0312: 640x480 (+2560), 24 bits
  Mode 0x0314: 800x600 (+1600), 16 bits
  Mode 0x0315: 800x600 (+3200), 24 bits
  Mode 0x0330: 320x200 (+320), 8 bits
  Mode 0x0331: 320x400 (+320), 8 bits
  Mode 0x0332: 320x400 (+640), 16 bits
  Mode 0x0333: 320x400 (+1280), 24 bits
  Mode 0x0334: 320x240 (+320), 8 bits
  Mode 0x0335: 320x240 (+640), 16 bits
  Mode 0x0336: 320x240 (+1280), 24 bits
  Mode 0x033d: 640x400 (+1280), 16 bits
  Mode 0x033e: 640x400 (+2560), 24 bits
  Config Status: cfg=new, avail=yes, need=no, active=unknown

However, as noted, (high-res) VT works very well :

  • either with nouveau ;
  • with proprietary blob, during boot initram & systemd startup, up until nvidia.ko is loaded.

Hmh, quite different from mine:

$ sudo hwinfo --framebuffer
[sudo] password for tuke: 
02: None 00.0: 11001 VESA Framebuffer                           
  [Created at bios.464]
  Unique ID: rdCR.U+J2Kd4CBPF
  Hardware Class: framebuffer
  Model: "NVIDIA GM107 Board - 20100050"
  Vendor: "NVIDIA Corporation"
  Device: "GM107 Board - 20100050"
  SubVendor: "NVIDIA"
  SubDevice: 
  Revision: "Chip Rev"
  Memory Size: 14 MB
  Memory Range: 0xd9000000-0xd9dfffff (rw)
  Mode 0x0300: 640x400 (+640), 8 bits
  Mode 0x0301: 640x480 (+640), 8 bits
  Mode 0x0303: 800x600 (+800), 8 bits
  Mode 0x0305: 1024x768 (+1024), 8 bits
  Mode 0x0307: 1280x1024 (+1280), 8 bits
  Mode 0x030e: 320x200 (+640), 16 bits
  Mode 0x030f: 320x200 (+1280), 24 bits
  Mode 0x0311: 640x480 (+1280), 16 bits
  Mode 0x0312: 640x480 (+2560), 24 bits
  Mode 0x0314: 800x600 (+1600), 16 bits
  Mode 0x0315: 800x600 (+3200), 24 bits
  Mode 0x0317: 1024x768 (+2048), 16 bits
  Mode 0x0318: 1024x768 (+4096), 24 bits
  Mode 0x031a: 1280x1024 (+2560), 16 bits
  Mode 0x031b: 1280x1024 (+5120), 24 bits
  Mode 0x0330: 320x200 (+320), 8 bits
  Mode 0x0331: 320x400 (+320), 8 bits
  Mode 0x0332: 320x400 (+640), 16 bits
  Mode 0x0333: 320x400 (+1280), 24 bits
  Mode 0x0334: 320x240 (+320), 8 bits
  Mode 0x0335: 320x240 (+640), 16 bits
  Mode 0x0336: 320x240 (+1280), 24 bits
  Mode 0x033d: 640x400 (+1280), 16 bits
  Mode 0x033e: 640x400 (+2560), 24 bits
  Mode 0x0345: 1600x1200 (+1600), 8 bits
  Mode 0x0346: 1600x1200 (+3200), 16 bits
  Mode 0x034a: 1600x1200 (+6400), 24 bits                                                                                           
  Mode 0x034b: 1680x1050 (+1680), 8 bits                                                                                            
  Mode 0x034c: 1680x1050 (+3360), 16 bits                                                                                           
  Mode 0x034d: 1680x1050 (+6720), 24 bits
  Mode 0x0360: 1280x800 (+1280), 8 bits
  Mode 0x0361: 1280x800 (+5120), 24 bits
  Config Status: cfg=new, avail=yes, need=no, active=unknown

What framebuffer driver do you use(dmesg|grep framebuffer)? I’m using uvesafb, which needs userspace virtualizing daemon, called v86d. See arch wiki for information of it:
https://wiki.archlinux.org/index.php/uvesafb

I think it has something to do with framebuffer driver you are using, which is not quite compatible with binary blops. If you are using grub you can check what the resolutions are supported before binary driver is loaded, with grub console command videoinfo(or vbeinfo).

As the relevant messages have been flushed out of the dmesg buffer, I’ll need to wait until my next reboot to have a look at them.

My current grub2 loads vbe & gfxterm, and sets the gfxpayload to 3200x1800x16.

However, I remember (not so fondly) trying every grub framebuffer incarnation months ago - when I was setting up the system - to no avail (for nvidia that is, nouveau worked fine back then).

Well do you have some other log i.e. kern.log: cat /var/log/kern.log|grep framebuffer

Yeah I remember it was horrible trial and error fiddling to get framebuffer drivers to work with nvidia binary(well not that bad still X worked every time). Hopefully upcoming drivers has better kms support as talked this years xdc2014 and thus framebuffer drivers to be better supported:
http://www.x.org/wiki/Events/XDC2014/XDC2014RitgerEGLNonMesa/

# journalctl -k | egrep -i "frame|buffer|fb|vesa|nvidia"

kernel: vesafb: mode is 3200x1800x16, linelength=6400, pages=0
kernel: vesafb: scrolling: redraw
kernel: vesafb: Truecolor: size=0:5:6:5, shift=0:11:5:0
kernel: vesafb: framebuffer at 0xf1000000, mapped to 0xffffc90007680000, using 11264k, total 11264k
kernel: Console: switching to colour frame buffer device 400x112
kernel: fb0: VESA VGA frame buffer device
...
kernel: nvidia: module license 'NVIDIA' taints kernel.
kernel: nvidia: module verification failed: signature and/or  required key missing - tainting kernel
kernel: [drm] Initialized nvidia-drm 0.0.0 20130102 for 0000:01:00.0 on minor 0
kernel: NVRM: loading NVIDIA UNIX x86_64 Kernel Module  343.22  Thu Sep 11 16:27:43 PDT 2014
...
kernel: NVRM: Your system is not currently configured to drive a VGA console
kernel: NVRM: on the primary VGA device. The NVIDIA Linux graphics driver
kernel: NVRM: requires the use of a text-mode VGA console. Use of other console
kernel: NVRM: drivers including, but not limited to, vesafb, may result in
kernel: NVRM: corruption and stability problems, and is not supported.

Note that (to the best of my knowledge, if memory serves well) I did try with stock “text-mode VGA console”, with the same results : black VT’s after nvidia.ko is loaded.

Hmm you are using vesafb not uvesafb. Well there are obvious problem right there

kernel: vesafb: mode is 3200x1800x16, linelength=6400, pages=0

Which aren’t listed in your hwinfo --framebuffer while you are running X → black screen.

You could try uvesafb and v86d, but I haven’t found any how-tos to redhat/fedora distributions. Arch linux wiki has some tutorial to use them with systemd, but again there might be distribution specific anomalies that does not work between arch and fedora.

Chiming in to add one more voice: same problem. Dell M4800, nvidia-drivers 343.22-r2 on Gentoo. I’ve tried uvesafb, X86_SYSFB=y/n.

Disable all framebuffer drivers and use a plain VGA console. Also don’t boot via UEFI. Then it should work.

Still no joy. I’ve disabled all FB drivers. Played with all manner of grub options (vga=normal, nofb, nomodeset). I wasn’t expecting those to work anyway, because the symptom is that the backlight turns off in F1-F6. The consoles may be there, they just can’t be seen.

Well,

  • This blatant and crippling bug never receiving any love from the NVidia team for the past two and a half years, and
  • The closed-source nvidia.ko module, as always, tainting the kernel (so - quite understandably - no support from the kernel developers either),

I finally bit the bullet and next to a Dell Precision 7510, purchased a Dell Precision 7710, both with a sufficiently performant Intel Iris Pro 580 iGP (i915.ko, open source) and the 7710 with a nice AMD R9 M295X dGP (amdgpu.ko, GCN1.2, open source).

Notwithstanding some ACPI issues (which, all components being open source, have at least a chance of being resolved), I couldn’t be happier.

No matter how you look at it, the open source efforts of AMD cannot be denied.
Evaluating the two support models (tainted/untainted), it’s a simple but straightforward business decision : thanks for nothing, and no more NVIDIA for our team.

that is true guys…
the powermanagement is definitly very bad with nvidia driver…i think the the code musst be implemented in the kernel …like the nouveau devs did …
Just do it .
please
regards

Mounir