insight please on why display is crashing (framebuffer/xorg?)

C

chessguy64

Hello. I've been using FreeBSD for 2 weeks. 13.0 to 13-1-RELEASE now. I've been able to minimize random crashes from pretty often to once in a while, but for some reason no matter what I do my display will hang with a black screen and grey rectangles at least some of the time, forcing me to tap the power button to restart. My nvidia-driver-390 is pkg installed and properly set up in /usr/local/etc/X11/xorg.conf.d. Lately it's only when I log out of icewm or e16 and return to text console that I get this bug. 80-90% of the time I just return to text console fine, but a few times the display is messed up and I have to use the power button to restart. Nothing in /var/crash (though I didn't wait a long time if a core dump was being written before powering off) and no evidence in Xorg.0.log indicating X did not exit cleanly. Other than this: ->
[ 537.043] (EE) event4 - IntelliMouse: client bug: event processing lagging behind by 17ms, your system is too slow
[ 545.213] (EE) event4 - IntelliMouse: client bug: event processing lagging behind by 14ms, your system is too slow
.
Someone has posted that they used the same driver with a card almost as old as mine, and it works well. And most of the time it works well for me, so I'm not 100% sure it's just an issue with the 390 nvidia driver and it's compatibility with other parts of FreeBSD. I've tried different versions of the 340/390 driver, same thing happens. (just less frequently now in 13.1). Is this related at all to framebuffer on my video card (I have a GT 430) Or how xorg shuts down, and returns to text console? If so, or if not, what's the best way to troubleshoot this? (including changing BIOS settings). Thanks
 
In short, I've not been able to solve the issue. I'm using the GeForce 1030 graphics card, but I'm not at all thinking it is the issue. I use a single HDD 1 Tb for Ubuntu and disconnect all of my other HDDs that I use FreeBsd 13.1 on. When I use Ubuntu the logging out or shutdown command works without any issues that we have encountered. Thus, It must be an issue with Xorg, but I don't have the experience or knowledge as to research or find any Xorg current bug data. Any way - Good Luck!
 
I don't know if it's strictly necessary, but for Nvidia cards I've had the following line in my /boot/loader.conf for as long as I can recall:

hw.vga.textmode="1"

Why? Because at one point in time when you switch from X to another virtual console (Ctrl-Alt-FN where FN is a "function key across the top of the keyboard") the console would not render correctly if at all. That line in the loader "fixes" it. You do lose any kind of "high resolution console" but my eyes are old enough that I don't care to look at tiny text.

I periodically get the "system too slow" because of events on the mouse on systems that use nvidia and Intel; it's obviously a "system taking too long getting around to processing an event from libevent" but I've only seen it related to an USB mouse so hard to say exactly where it is.
 
I had a similar problem with an Nvidia 1080Ti. I would get a black screen with a white square on the upper left side when I quit Openbox. Only way out of it was reboot (power switch still worked). It went away when I upgraded to 12.3.
 
Jose said:
I had a similar problem with an Nvidia 1080Ti. I would get a black screen with a white square on the upper left side when I quit Openbox. Only way out of it was reboot (power switch still worked). It went away when I upgraded to 12.3.
Click to expand...

Thanks Jose, you're correct. When I was using 12.3 - No Problem! When I upgraded to both 13.0 and now 13.1 - current issue. Then only change that has made a difference as to a black screen of death, is I change my BIOS to read the PCIe slot that my GeForece 1030 is using. But, the blinking on and off is still there. I'm simply running out of options. Mer, shared - hw.vga.textmode="1", which was nice, but that code did not change my issue. Thanks to all!! - I'll keep reading. researching, and trying and if or when I find a solution, I'll post ASAP!
 
mer said:
I don't know if it's strictly necessary, but for Nvidia cards I've had the following line in my /boot/loader.conf for as long as I can recall:

hw.vga.textmode="1"
Click to expand...

This actually might be the root cause of why the display is crashing for so many. I didn't have that in my /boot/loader.conf, and did Ctrl+Alt+F1 in icewm, and my screen went garbled with distorted colors. Then I did Ctrl+Alt+F2 and noticed the exact same black screen I was getting before with grey rectangles when it was hanging. Once I added that line to /boot/loader.conf it doesn't do that anymore and returns normally when I do Ctrl+Alt+Fn. Also, I'm getting those "your system is too slow" errors using a PS/2 mouse, not USB.
 
  • Like
Reactions: mer
grahamperrin said:
If you habitually flush swap devices before doing so, then are symptoms reproducible?
Click to expand...

I'd have to test it a lot. It's not something you can easily reproduce, because it happens so infrequently. But now I have that hw.vga setting in /boot/loader.conf, and if that is fixing things, I don't want to remove that line and test if things are working correctly now without any crashes in the future.
 
Well, I just spent two hours installing and then configuring BSD 12.3 that I had used 13 months ago on the exact same box! Guess WHAT, it did not resolve the issue of the Black Screen of Death. However, that issue was not a problem until I installed 13.0 and now it is an issue with 12.3! Thus, I'm going to shutdown and reinstall 13.1 and keep trying to resolve the issue. The only thing that I saw go by while installing the gnome and mate pkg was a message that the D-BUS was not recognized. At least the D-BUS was highlighted in yellow, or I wouldn't even notice that quick message as it was installing the pkg. So, maybe when I have 13.1 up and configured for my taste, I'll research that in issue. Thanks, I'll talk with you later.
 
rpowell47 said:
Well, I just spent two hours installing and then configuring BSD 12.3 that I had used 13 months ago on the exact same box! Guess WHAT, it did not resolve the issue of the Black Screen of Death. However, that issue was not a problem until I installed 13.0 and now it is an issue with 12.3!
Click to expand...

I don't know why if it was working without any problems in 12.3 before, then installing 13.0, using it for a while, and wiping the drive and reinstalling 12.3, why the issue would come back. I don't think it's related to your problem that you installed 13.0 on the hard drive before going back to 12.3 Also, did you add

dbus_enable="YES" to /etc/rc.conf before you installed gnome or mate? Keep in mind I don't have dbus enabled and i've had this crash happen before on logout of a wm. But I didn't have that hw.vga setting in /boot/loader.conf while the crash happened.
 
I'm sorry you went through all that trouble for naught. I have Dbus disabled on my system, FWIW. I doubt that's the problem, though.
 
Jose said:
I'm sorry you went through all that trouble for naught. I have Dbus disabled on my system, FWIW. I doubt that's the problem, though.
Click to expand...

What wm do you use? I'm thinking of migrating to 12.3. You haven't had the crash happen even a single time? Also do you have hw.vga.textmode="1" in /boot/loader.conf right now?
 
chessguy64 said:
What wm do you use?
Click to expand...
I use x11-wm/openbox.

chessguy64 said:
I'm thinking of migrating to 12.3. You haven't had the crash happen even a single time?
Click to expand...
Keep in mind my video card is much newer than yours, and is supported by the latest driver, which is what I use. It's unlikely 12.3 is the answer to your problems.

chessguy64 said:
Also do you have hw.vga.textmode="1" in /boot/loader.conf right now?
Click to expand...
Code: 
$ cat /boot/loader.conf                                                                                   
nvidia-modeset_load="YES"
hw.vga.textmode=1
fuse_load="YES"
 
Jose said:
Keep in mind my video card is much newer than yours, and is supported by the latest driver
Click to expand...

Well according to nvidia and FreeBSD, my video card is supported by nvidia-driver-390. Given that I saw the same graphic on screen when I didn't have hw.vga.textmode=1 in /boot/loader.conf, and did Ctrl+Alt+F1, Ctrl+Alt+F2 in X, (crash) matching the graphic on screen when I logged out of wm and it crashed.. I'd say there's a high probability it's just because I didn't have that line in /boot/loader.conf, and it has little to do with nvidia driver compatibility. I'd have an uptime of 13.5 hours before in a wm without a crash. It was only crashing recently on 13.1 when I logged out of a wm (sometimes) But yeah, I can probably use 13.1 successfully without reinstalling to 12.3. I have to reinstall anyway, because I didn't set a single / partition and did /var /usr /tmp / partitions. Might want to go with 12.3 because it is older and more stable.
 
chessguy64 said:
… I'd say there's a high probability it's just because I didn't have that line in /boot/loader.conf, …
Click to expand...

If attention to loader.conf(5) improves things, then (yes) you can ignore my suggestion to flush swap devices.

(The suggestion was loosely based on situations that precede very rare freezes with my AMD graphics hardware. So rare that I haven't bothered to properly diagnose things.)

Remind me, please:

chessguy64 said:
… tap the power button to restart. …
Click to expand...

Those were short taps, yes? To gracefully stop.

(Not pressing and holding to force a stop.)

chessguy64 said:
… 12.3 because it is older and more stable.
Click to expand...

Not necessarily.

Some things that are (or will be) fixed for releng/13 will never be fixed for releng/12.
 
Jose
I would remove nvidia module lines from loader.conf and add them to the kld_list line in /etc/rc.conf. I know at one point in time loader.conf was the recommended way, but I've had boot issues in the past and I think others have reported issues too when it's in loader.conf.
Just my opinions.
 
Well spotted by mer.

Also, fuse_load="YES" is (technically) useless. because fuse is not the name of the module. Maybe the result of following outdated guidance.

If FUSE is needed:

sysrc kld_list+=fusefs && kldload fusefs

Here, as a result of the rc.conf(5) entry:

Code: 
% kldstat | grep fusefs
13    1 0xffffffff84493000    12db0 fusefs.ko
%
 
You must log in or register to reply here.
Back
Top