Missing information in libevdi docs

[Ferruck]

Missing information in libevdi docs

[Ferruck]

I investigated the problem a bit further and it seems that the number of elements to be written by fwrite are incorrect. sizeof(devices_to_add) is passed as the argument, which is 2, though only one char should be written.

I opened a pull request (#55) that fixes the problem. If it's to hacky, please let me know.

[displaylink-mlukaszek]

Thanks for pointing this out. I think it is actually a documentation being wrong there. It should say 0 when there was a problem, and a non-zero value otherwise (without worrying what the exact number is).

[Ferruck]

OK, so the pull request is obsolete. But I still wonder, do you really need to write the \0 into the file as well?

[displaylink-mlukaszek]

No, probably not. The value is interpreted by evdi_add_store in the kernel module code, and it should be fine without the terminator.
Btw, if you intend to play with evdi and the library (which I encourage you to do), you might find this useful as an example: https://github.com/mlukaszek/evdipp

[Ferruck]

Indeed I want to play with evdi in the next days so your repo is a great ressource, thanks!

[Ferruck]

I'm not opening a new issue as I'm probably just holding it wrong.

I tried out your C++ examples and wrote one myself, but neither work. Calling evdi_check_device on the different cards in /dev/dri/ lead to different errors:

• card0 returns a UNRECOGNIZED and on the terminal I get the additional error

  [libevdi] ioctl: drop_master error=-1
  

  I guess this is intended and correct, because this is the default device that is present without the evdi driver, as well.

• Every other device I tried (I added devices up to card10 with evdi_add_device which actually works) returns NOT_PRESENT. This seems strange to me.

This is what I get from journalctl when adding a new card (trying to access an already created card gives no output at all):

Nov 14 18:43:24 TP_Trommler kernel: evdi: [D] add_store:195 Increasing device count to 2
Nov 14 18:43:24 TP_Trommler kernel: evdi: [D] evdi_crtc_init:312 drm_crtc_init: 0
Nov 14 18:43:24 TP_Trommler kernel: evdi: [W] evdi_painter_crtc_state_notify:377 Painter does not exist!
Nov 14 18:43:24 TP_Trommler kernel: evdi: [D] evdi_detect:72 Painter is disconnected
Nov 14 18:43:24 TP_Trommler kernel: evdi evdi.1: No connectors reported connected with modes
Nov 14 18:43:24 TP_Trommler kernel: [drm] Cannot find any crtc or sizes - going 1024x768
Nov 14 18:43:24 TP_Trommler kernel: evdi evdi.1: fb2: evdidrmfb frame buffer device
Nov 14 18:43:24 TP_Trommler kernel: [drm] Supports vblank timestamp caching Rev 2 (21.10.2013).
Nov 14 18:43:24 TP_Trommler kernel: [drm] No driver support for vblank timestamp query.
Nov 14 18:43:24 TP_Trommler kernel: [drm] evdi: evdi_stats_init
Nov 14 18:43:24 TP_Trommler kernel: [drm] Initialized evdi 1.2.64 20161003 on minor 2
Nov 14 18:43:24 TP_Trommler gnome-shell[508]: Failed to apply DRM plane transform 0: Keine Berechtigung
Nov 14 18:43:24 TP_Trommler gnome-shell[508]: Failed to apply DRM plane transform 0: Keine Berechtigung

(Sorry for the German locale)

I'm running Arch Linux on a ThinkPad X230, uname -a gives the following output:

Linux TP_Trommler 4.8.6-1-ARCH #1 SMP PREEMPT Mon Oct 31 18:51:30 CET 2016 x86_64 GNU/Linux

Am I just dumb and missing some dependencies? Is there just a simple reason why this doesn't work (and probably won't ever)? Or is this a bug in the driver?

I'd really like to work with evdi and hope that I'm just holding it wrong.

Thanks in advance!

[mlukaszek]

Could you paste a snippet of code that you're trying?

[Ferruck]

I can give you the whole code I tried, see here.

But I also tried your C++ examples which give me the same errors.

[mlukaszek]

Just to double check - are you running as root? Note the docs says "when the device does not exist or cannot be opened to check". This could be a permission problem, you need to use sudo to be able to send IOCTL from the library to the kernel module.

[Ferruck]

Of course I tried as root with sudo. Running without superuser permissions lead to different errors.

Can you change log level to VERBOSE? The easiest way is to write 6 to /sys/devices/evdi/loglevel.
references [1] [2]

[Ferruck]

OK, the verbose debug output is as follows:

Nov 15 11:36:23 TP_Trommler kernel: evdi: [D] add_store:195 Increasing device count to 1
Nov 15 11:36:23 TP_Trommler kernel: evdi: [V] evdi_platform_probe:119 
Nov 15 11:36:23 TP_Trommler kernel: evdi: [V] evdi_driver_load:25 
Nov 15 11:36:23 TP_Trommler kernel: evdi: [V] evdi_driver_load:37 
Nov 15 11:36:23 TP_Trommler kernel: evdi: [V] evdi_modeset_init:361 
Nov 15 11:36:23 TP_Trommler kernel: evdi: [V] evdi_crtc_init:306 
Nov 15 11:36:23 TP_Trommler kernel: evdi: [D] evdi_crtc_init:312 drm_crtc_init: 0
Nov 15 11:36:23 TP_Trommler kernel: evdi: [V] evdi_flip_workqueue_init:324 
Nov 15 11:36:23 TP_Trommler kernel: evdi: [V] evdi_crtc_disable:112 
Nov 15 11:36:23 TP_Trommler kernel: evdi: [W] evdi_painter_crtc_state_notify:377 Painter does not exist!
Nov 15 11:36:23 TP_Trommler kernel: evdi: [V] evdi_detect:67 
Nov 15 11:36:23 TP_Trommler kernel: evdi: [D] evdi_detect:72 Painter is disconnected
Nov 15 11:36:23 TP_Trommler kernel: evdi evdi.0: No connectors reported connected with modes
Nov 15 11:36:23 TP_Trommler kernel: [drm] Cannot find any crtc or sizes - going 1024x768
Nov 15 11:36:23 TP_Trommler kernel: evdi evdi.0: fb1: evdidrmfb frame buffer device
Nov 15 11:36:23 TP_Trommler kernel: [drm] Supports vblank timestamp caching Rev 2 (21.10.2013).
Nov 15 11:36:23 TP_Trommler kernel: [drm] No driver support for vblank timestamp query.
Nov 15 11:36:23 TP_Trommler kernel: evdi: [V] evdi_painter_init:677 
Nov 15 11:36:23 TP_Trommler kernel: [drm] evdi: evdi_stats_init
Nov 15 11:36:23 TP_Trommler kernel: [drm] Initialized evdi 1.2.64 20161003 on minor 1
Nov 15 11:36:23 TP_Trommler gnome-shell[506]: Failed to apply DRM plane transform 0: Keine Berechtigung
Nov 15 11:36:23 TP_Trommler gnome-shell[506]: Failed to apply DRM plane transform 0: Keine Berechtigung

[mlukaszek]

Have you been modifying the evdi kernel module or library code at all?

What evdi_check_device does is quite simple:

• it checks e.g. /dev/dri/card1 exists at all,
• then tries to open it (which in your case must have succeeded as you see messages about failed drop master IOCTL),
• then checks that the device is EVDI - by using DRM_IOCTL_VERSION ioctl to get the version, and comparing the name of the device with "evdi".

I guess your next logical step is to attach a debugger and step through the calls, to figure out which of the above steps fail.

[Ferruck]

No, I have neither modified the kernel module nor the library. I'll debug the program in the next days.

[Ferruck]

Sorry for the really late response. I hadn't have access to the laptop I used while encountering this issue any more. On another machine I can't reproduce the problem so I'm closing it now. Thanks for your support.