-
Notifications
You must be signed in to change notification settings - Fork 142
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Ubuntu 15.04 works with upstart, doesn't with systemd #644
Comments
Looks like a compatibility issue with the modprobe configuration file on Ubuntu, please take this downstream (on Launchpad). |
Did you mean write to systemd developers on launchpad or what? |
Surely you could at least copy some details to the Launchpad tracker instead of pasting a link :-P When I said compatibility, I think that systemd(/kmod?) does not like the |
@Lekensteyn, done, I just hate launchpad because it still doesn't support any formatting and long text like this is barely readable. |
This is unrelated to systemd or any other init, but how kmod behaves. "alias nvidia off" isn't a "magic" value in any way (it's undocumented in man modprobe.d either), if you try to load that module you will get that This should be reopened. |
@martinpitt IIRC it was done to prevent Isn't the nvidia packaging on Ubuntu responsible for this module configuration snippet? |
I don't know who/what creates that "alias nividia off" thing, to be honest. It's not the bumblebee package as far as I can see, but it could certainly be the nvidia driver, or one of its configuration helpers. I also don't know what bumblebee does, so please forgive my ignorance :-) If Bumblebee makes sense to run without the "nvidia" module, then it should be robust against modprobe failing. If it doesn't make sense, then failing its .service seems exactly right? |
@martinpitt Bumblebee basically works like this:
When configured to use the nvidia driver, then the daemon fails if it cannot find the nvidia module as shown in the first post. |
Just got systemd-related update of kmod, now everything works fine) |
Thanks for reporting back! |
@nazar-pc what package did you got updated and which version? i just upgraded to 15.04 and nvidia-346, installed all updates, and having same problem. |
It should work out of the box since we should have same versions of packages (systemd and related at least). |
bumblebee-bugreport: https://www.dropbox.com/s/gun4k1xvbl86b7k/bumblebee-bugreport-20150425_131051.tar.gz?dl=0 |
Did you set
Also if you have |
installed nvidia-349 from edgers
still the same:
bumblebee-bugreport: https://www.dropbox.com/s/dmg9zj05jayzv9v/bumblebee-bugreport-20150425_133237.tar.gz?dl=0 |
I've posted my configuration for you to try it. |
WIth your configuration i'm getting black screen. |
Our configurations in fact differ only in that single line, there is no way for black screen with my config and working with changed line, because contents is identical. |
I'm having the same issue with systemd and bumblebee not starting properly. Tinkering with it, I got to the same problems than some of you had (not being able to load the gpu driver). Tried some of the solutions here with no luck. So before breaking more stuff, I wanted to ask, having these packages installed from the official ubuntu repos:
Was there a special order in which I should have installed them? The defaults in bumblebee.conf were "Driver=" and "KernelDriver=nvidia-current", but they fail to load the gpu. I tried with "Driver=nvidia" and "KernelDriver=nvidia", but they gave me a black screen on the first boot, then a normal second boot but still not able to load the gpu drivers. I also tried to manually starting bumblebee via systemctl and manually calling modprobe on nvidia_346, with the same results. |
After last bumblebee update in Ubuntu 15.10 (yes, unstable) I do not need to call |
As for the nvidia issue, that happens because nvidia is not properly unloaded since the change from rmmod to modprobe. It is a known issue for which a fix was attempted in edd81b0, but it is possible that the fix does not work. Can you report whether you can really unload a module via modprobe?
Failing this, you can run the |
Ok, I found my issue: I had to use "nvidia-346" instead of just "nvidia" in the KernelDriver setting of the [driver-nvidia] section. Having those listed packages installed (my previous comment), my working config is this:
|
@Lekensteyn, unloading is not working properly in any way you've mentioned, I've already tried all of that before, here is some details:
So it seems that driver is not loaded anymore, but discrete GPU indicator on laptop is still working. |
@nazar-pc Thanks for testing this, can you test another thing:
If this still does not work, then we must revert the change for Ubuntu users. To disable card manually, unload the driver and write OFF:
|
Unfortunately doesn't, reboot is not needed after file creation, right?
I'm on kernel 4.0-rc4 currently if it makes any difference. |
Nope, a reboot is not needed as modprobe consults the file at runtime. The rmmod->modprobe change is about to be reverted for Ubuntu, but this in turn makes it not possible to unload nvidia-uvm before unloading nvidia... Can you try another modprobe configuration in
This should work, just tested something similar. |
Sorry, probably there is some type there, since I don't have |
That was a line for a file in |
Yes, |
Hi, I'll mention this here as well (for traceability more than anything else), here's a proposed solution until the driver packages change: Bumblebee-Project/bumblebee-ppa#34 |
Currently Ubuntu switched to systemd by default, unfortunately bumblebee is not working in this case.
I have proprietary nvidia driver and bumblebee from standard repository, kernel 3.19.0.
with
Driver=
with
Driver=nvidia
When booted with upstart - works perfectly
The text was updated successfully, but these errors were encountered: