-
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
Bumblebee not working on Lenovo Thinkpad P50 with Centos 6.8-6.9 and 7.3-7.5 #974
Comments
Are you using centos 7 on that machine? I'm using fedora but I have a spare machine I can try tomorrow with centos 6 or 7 depending on which version you are running. Its been a month and a half since I last tested centos. I made the packages hosted at NCSU. |
Dear gsgatlin, |
Hmnn. I installed bumblebee on a ideapad y470 notebook (A pretty old machine) using the instructions from: https://www.linux.ncsu.edu/bumblebee And it appears to be working ok with virtualgl and primus bridge. CentOS 7. Kernel is latest at 3.10.0-862.9.1.el7.x86_64 [gsgatlin@y470 ~]$ cat /etc/redhat-release Did the drivers build ok? What is the output from: bumblebee-nvidia --check ? I see there is a which seems to be similar problem? Also some notes about it here... https://wiki.debian.org/Bumblebee |
Here were some fedora issues which were similar. I know you are running centos 7 but its a similar distro to fedora. Besides posting output from bumblebee-nvidia --check Can you paste output from journalctl or /var/log/messages to https://pastebin.com/ or https://paste.fedoraproject.org/ or similar site right after you run optirun or primusrun and it fails? I'm pretty sure the elrepo bumblebee packages have been abandoned by their creator sadly. (Rob Mokkink) Thanks, |
Dear gsgatlin, nvidia.ko compiled into in the kernel tree ok. filename: /lib/modules/3.10.0-862.9.1.el7.x86_64/kernel/drivers/video/nvidia.ko Check bbswitch kernel module... bbswitch is loaded into the current kernel ok. All checks completed successfully! NVIDIA driver appears to have compiled ok. Documentation on bumblebee for RHEL / CentOS / fedora can be found at: I tried pastebin.com, but I could not figure out how to upload a file. The raw messages file is about 369 MB, but it compressed down to 14 MB, so I uploaded it here for your reference. Thank you again for your help. |
Hello. pastebin doesn't allow you to upload a file. You paste in text to a box and hit submit and it makes a temporary place to look at text output. But I was able to download your file for the messages syslog. So I'm seeing a couple of really weird things here... It may be that you've already fixed these and started out from elrepo but I want to make sure. The first is that the executable is actually called glxgears (not glxgears64 which is what you were trying to run) [gsgatlin@logicbomb ~]$ which glxgears The second involves the line in your output when first reporting the bug: [ 233.925766] [DEBUG] Primus LD Path: /usr/lib/primus:/usr/lib32/primus:/usr/lib64/primus I my box I get: [gsgatlin@t540p ~]$ optirun --debug glxgears [11398.036841] [INFO]Running application using virtualgl. (HERE IS WHERE I CLOSE SPINNING GEARS WINDOW) [VGL] ERROR: in readback-- Notice the line in my output: [11397.452131] [DEBUG] Primus LD Path: /usr/lib/primus:/usr/lib64/primus I wonder if the packages have gotten mixed up somehow with someone else's packages? I'm pretty sure this PATH is generated by: export CONF_PRIMUS_LD_PATH="/usr/lib/primus:/usr/lib64/primus" within my bumbleee rpm spec file at compile time. (Sources to all packages are at: https://github.com/gsgatlin/optimus-rpms ) Because on your box its trying to use debian style paths when you first reported this. Or maybe "all systems type paths" but its not what I made which is more red hat family specific. [gsgatlin@t540p bumblebee]$ rpm -q bumblebee --changelog | grep gsgatlin
I would try removing all bumblebee* bbswitch, primus, virtualgl packages, temporarily disabling elrepo yum repository, then adding back the packages from my repo and rebooting. Then try it again and see what you get. You can always add elrepo back while blocking certain packages like bumblebee, primus, bbswitch, etc if there are other packages you need from elrepo. |
Dear gsgatlin, https://forums.opensuse.org/showthread.php/505270-ultimate-tutorial-installing-Bumblebee-driver-for-SUSE bumblebee-nvidia.conf.txt Here is the /etc/modeprobe.d/bumblebee.conf: [wrthissell@LAPTOP-BKIJEPGK ~]$ nvidia-settings -c :8 ERROR: Unable to find display on any available system Thank you again for your help. [wrthissell@LAPTOP-BKIJEPGK ~]$ systemctl -l status bumblebeed Jul 28 15:44:18 LAPTOP-BKIJEPGK bumblebeed[1171]: [ 7303.023667] [ERROR][XORG] (EE) NVIDIA(GPU-0): Failed to initialize the NVIDIA graphics device! |
Yeah. my packages don't make a /etc/modeprobe.d/bumblebee.conf Also, my /etc/bumblebee/bumblebee.conf doesn't set PrimusLibraryPath So you probbaly want to remove that line. |
I think the forum post is specific to suse. With my packages hosted at ncsu the driver is contained within the bumblebee-nvidia package. |
[gsgatlin@t540p ~]$ cat /etc/modprobe.d/bumblebee.conf |
Sorry, correction, there is a /etc/modprobe.d/bumblebee.conf with the modules listed in previous comment. |
Also, the command to start nvidia settings is: optirun -b none nvidia-settings -c :8 So try that. |
Dear gsgatlin, [wrthissell@LAPTOP-BKIJEPGK ~]$ optirun --debug glxgears [ 89.977532] [ERROR]Cannot access secondary GPU - error: [XORG] (EE) /dev/dri/card0: failed to set DRM interface version 1.4: Permission denied [ 89.977537] [DEBUG]Socket closed. [wrthissell@LAPTOP-BKIJEPGK ~]$ bumblebee-nvidia --check nvidia.ko compiled into in the kernel tree ok. filename: /lib/modules/3.10.0-862.9.1.el7.x86_64/kernel/drivers/video/nvidia.ko Check bbswitch kernel module... bbswitch is loaded into the current kernel ok. All checks completed successfully! NVIDIA driver appears to have compiled ok. Documentation on bumblebee for RHEL / CentOS / fedora can be found at: I am attaching the most recent /var/log/messages as a tar.gz file: May the issue be related to my kernel load command? I have added several options in the year and half that I have been trying to resolve the issue. Here is the latest grub.cfg: Here is the dmesg command: [wrthissell@LAPTOP-BKIJEPGK ~]$ sudo dmesg | grep -C 10 bbswitch
|
Does it change anything if you remove rcutree.rcu_idle_gp_delay=1 nvidia-drm.modeset=1 pcie_port_pm=off acpi=on acpi_rev_override=5 from kernel boot? (Edit /etc/default/grub and run either grub2-mkconfig -o /boot/grub2/grub.cfg or grub2-mkconfig -o /boot/efi/EFI/centos/grub.cfg depending on which one you have or both if both files exists) Can you upload /var/log/Xorg.8.log ? if removing kernel arguments "rcutree.rcu_idle_gp_delay=1 nvidia-drm.modeset=1 pcie_port_pm=off acpi=on acpi_rev_override=5" doesn't help? Is hybrid graphics set up in the BIOS/UEFI rather then to always use discrete card? Evidently this is a setting that can be controlled with this particular notebook? Also, it turns out that "error: [XORG] (EE) /dev/dri/card0: failed to set DRM interface version 1.4: Permission denied" is bogus and we can just ignore that. This always fail due to problems in libdrm, but does not affect Bumblebee's functionality. See #652 for more information. I know its annoying to have useless error messages but there is little we can do about that... |
Dear gsgatlin, [ 76.456967] [ERROR]Cannot access secondary GPU - error: [XORG] (EE) /dev/dri/card0: failed to set DRM interface version 1.4: Permission denied [ 76.456973] [DEBUG]Socket closed.
Same results as above.
http://pastebin.com/raw/ea5jKJ15
|
Dear gsgatlin, |
The main problem seem to be "Failed to initialize the NVIDIA GPU at PCI:1:0:0. Please" and "Failed to initialize the NVIDIA graphics device" But I am unsure why this is happening. Perhaps one of the bumblebee developers will have some ideas assuming some of them still are reading these github issues. I pushed out a never version of the driver today. Does "yum update bumblebee-nvidia" followed by a reboot help? Should be bumblebee-nvidia-390.77-1 you would be updating to. I doubt it could help but its worth a try... This seems very similar to #542 Maybe you should add back rcutree.rcu_idle_gp_delay=1 but not the other arguments? Just to see if might help.... Or maybe if you add the arguments back you could try them all one at at time rather then all in one go? I have access to 4 different optimus laptops through my job plus ones I bought with my own money.. Three of them require no kernel arguments. The fourth requires only one argument: "acpi_osi=!Windows 2013" with fedora 25/26/27. A laptop a student brought to the help desk required acpi_osi=!Windows 2009 which I think was one of those dell xps machines. But I don't remember which one it was. I can look at adding that patch. I think it just hides the bogus error messages. I can look at it tomorrow. I haven't looked at doing a new release in a long time because there hasn't been a new official version since 3.2.1. If I have time to add that patch I will put the package on the web where you can download it to try it out. |
Dear gsgatlin, sudo yum update bumblebee-nvidia* It could not find an update. I then manually checked: https://linux.itecs.ncsu.edu/redhat/public/bumblebee-nonfree/rhel7/x86_64/ and the update is not there. Is it somewhere else? |
Sorry. I see what happened. Try it again. There was a bug in my staging script. yum clean all yum update bumblebee-nvidia |
Dear gsgatlin, rcutree.rcu_idle_gp_delay=1 Here are the identical results I received after each reboot: [wrthissell@LAPTOP-BKIJEPGK ~]$ optirun --debug glxgears [ 102.744965] [ERROR]Cannot access secondary GPU - error: [XORG] (EE) /dev/dri/card0: failed to set DRM interface version 1.4: Permission denied [ 102.744985] [DEBUG]Socket closed. nvidia.ko compiled into in the kernel tree ok. filename: /lib/modules/3.10.0-862.9.1.el7.x86_64/kernel/drivers/video/nvidia.ko Check bbswitch kernel module... bbswitch is loaded into the current kernel ok. All checks completed successfully! NVIDIA driver appears to have compiled ok. Documentation on bumblebee for RHEL / CentOS / fedora can be found at:
|
Dear gsgatlin, https://www.centos.org/forums/viewtopic.php?f=49&t=66178&p=284988#p284988
|
Hello. I looked at that patch you linked. The bottom part is already being done (and more) in the bumblebee-modprobefix.patch... https://paste.fedoraproject.org/paste/aGFXGhCXNoXQgbpi4HHDag I created a new package in a new branch called experimental1 https://github.com/gsgatlin/optimus-rpms/tree/experimental1 It appears to remove "failed to set DRM interface version 1.4: Permission denied" from the syslog. It does still appear in my /var/log/Xorg.8.log and I am unsure it would appear on command line for your situation. I don't think it would fix underlying problem sadly. ( I get that error message but my bumbleee works fine in centos 7 and fedora 27) Still, I compiled it and placed it on the web at this url: https://linux.itecs.ncsu.edu/redhat/public/bumblebee/experimental1/ Package name is bumblebee-3.2.1-14.el7.x86_64.rpm I can look into building packages off the develop branch of https://github.com/bumblebee-Project/Bumblebee/ and https://github.com/Bumblebee-Project/bbswitch if there is time tomorrow or the Friday. Perhaps I can make the version number be bumblebee 3.3 ? and bbswitch 0.8.5 ? based off of the develop branch of both repos. I'm just making up version numbers here. |
Dear gsgatlin, [wrthissell@LAPTOP-BKIJEPGK ~]$ bumblebee-nvidia --check nvidia.ko compiled into in the kernel tree ok. filename: /lib/modules/3.10.0-862.9.1.el7.x86_64/kernel/drivers/video/nvidia.ko Check bbswitch kernel module... bbswitch is loaded into the current kernel ok. All checks completed successfully! NVIDIA driver appears to have compiled ok. Documentation on bumblebee for RHEL / CentOS / fedora can be found at: [wrthissell@LAPTOP-BKIJEPGK ~]$ optirun --debug glxgears [ 729.676462] [ERROR]Cannot access secondary GPU - error: [XORG] (EE) NVIDIA(GPU-0): Failed to initialize the NVIDIA GPU at PCI:1:0:0. Please [ 729.676489] [DEBUG]Socket closed. I shall now try without the acpi_osi=!Windows 2009 kernel load command. |
Dear gsgatlin, [wrthissell@LAPTOP-BKIJEPGK ~]$ sudo dmesg | grep -C 10 bbswitch
|
Dear gsgatlin, [wrthissell@LAPTOP-BKIJEPGK ~]$ bumblebee-nvidia --check nvidia.ko compiled into in the kernel tree ok. filename: /lib/modules/3.10.0-862.9.1.el7.x86_64/kernel/drivers/video/nvidia.ko Check bbswitch kernel module... bbswitch is loaded into the current kernel ok. All checks completed successfully! NVIDIA driver appears to have compiled ok. Documentation on bumblebee for RHEL / CentOS / fedora can be found at: [wrthissell@LAPTOP-BKIJEPGK ~]$ optirun --debug glxgears [ 143.503669] [ERROR]Cannot access secondary GPU - error: [XORG] (EE) NVIDIA(GPU-0): Failed to initialize the NVIDIA GPU at PCI:1:0:0. Please [ 143.503676] [DEBUG]Socket closed. [ 155.028550] [ERROR]Cannot access secondary GPU - error: [XORG] (EE) NVIDIA(GPU-0): Failed to initialize the NVIDIA GPU at PCI:1:0:0. Please [ 155.028577] [DEBUG]Socket closed.
|
I will ask on a mailing list we have if anyone has a Lenovo Thinkpad P50 I could borrow for a few hours. It seems unlikely because we have a list of approved computer they usually buy but you never know until asking... Here are some other ideas we can try. Just out of curiosity, if you run: cat /proc/acpi/bbswitch do you get 0000:01:00.0 OFF ? Just wanted to see if its at least powering off the card or keeping it on all the time... Or if the virtual file is even there...We know you can't play 3d video games with optirun (optirun -b virtualgl) / primus not working but does it at least save power and improve battery life? (nouveau would likely also do that on centos 7, but not centos 6, unless your discrete graphics is too new for red hat to have backported that code to the 3.10 kernel yet) Does optirun -b primus glxgears -info Give similar errors? How about primusrun glxgears -info ? lsmod | grep nouveau should return nothing (no output) if its been successfully blacklisted in the kernel command line. I think it will be the same but just wanted to check. If nouveau shows up in the output from lsmod that would be why you see "Failed to initialize the NVIDIA GPU" message. I don't have centos 7 box in front of me at the moment. ( I left it at work today, I use fedora 27 on this box) But as root on my thinkpad t540p fedora box: [root@t540p ~]# modprobe --verbose nvidia [gsgatlin@t540p ~]$ lsmod | grep nvidia [root@t540p ~]# modprobe -r nvidia [gsgatlin@t540p ~]$ lsmod | grep nvidia I wonder how your machine will behave with similar commands? At least its giving you the correct problem now instead of a bogus error: "Response: No - error: [XORG] (EE) NVIDIA(GPU-0): Failed to initialize the NVIDIA GPU at PCI:1:0:0. Please" As for why its failing to "initialize the NVIDIA GPU" that I'm not sure about. Can't find much about that particular error message. I will probably make bumblebee-3.2.1-14 the default version in my yum repo since the error messages are clearer. Might save some people (and myself If I ever get new laptop) some trouble with troubleshooting these kinds of problems... I will see if I can build "develop" branches of bumblebee and or bbswitch tomorrow if there is time. I will make a second branch in my github repository for that. Maybe those "develop branch" rpms could help? I thing the debuginfo rpms are only useful if it segfaults or similar. Which I gather its not doing on your machine. That is because they "strip" the binarys in the rpms. It may be that my "bumblebee-nvida --check" section could do other checking operations on the nvidia module that are not clear to me at the moment. Perhaps something that could show a problem on your laptop. Like maybe the elrepo nvidia driver did not get completely removed somehow or something. I will give it further thought. |
Couple more thing I thought of, one of which that burned me in desktop setup with network home directory recently. Do you have ".nvidia-settings-rc" and or nv/ directory in your $HOME directory? If so you may want to delete those. Can you psatebin your /etc/bumblebee/xorg.conf.nvidia file? In the Xorg.8.log.txt file you uploaded i didn't see so I wonder if the config file is different then mine... 147115.312] () Option "IgnoreABI" "1" |
Dear gsgatlin, [wrthissell@LAPTOP-BKIJEPGK ~]$ cat /proc/acpi/bbswitch [wrthissell@LAPTOP-BKIJEPGK ~]$ optirun -b primus glxgears -info [40431.612953] [ERROR]Aborting because fallback start is disabled. [wrthissell@LAPTOP-BKIJEPGK ~]$ primusrun glxgears -info [wrthissell@LAPTOP-BKIJEPGK ~]$ lsmod | grep nouveau [wrthissell@LAPTOP-BKIJEPGK ~]$ modprobe --verbose nvidia I do not have ".nvidia-settings-rc" or nv/ directory in my $HOME directory. Below is my /etc/bumblebee/xorg.conf.nvidia: The date of the above file is yesterday, which is when I executed the yum install for your experimental bumblebee rpm. Below is the current boot Xorg.8.log file: 20180802 Xorg.8.log.txt [ 40555.271] (**) IgnoreABIOption "AutoAddDevices" "false" I also noticed the following line: [ 40555.270] (==) Using system config directory "/usr/share/X11/xorg.conf.d" Below is a zip file of this directory:
|
So it appears to me that your bbswitch module is not working. It seems the card always stays on. Perhaps some kernel command line would help. If so I am utterly clueless about what that might be. It seems like you've tried them all. What happens if you run: tee /proc/acpi/bbswitch <<<OFF as root? Does that change the status? You can check with cat /proc/acpi/bbswitch (Either it will show on or off) The "develop" branch of bbswitch is not very different from master branch. We are already patching the differences to that version on fedora with bbswitch-412.patch. On RHEL / CentOS kernel that patch is not needed due to those using older kernel with backported features from newer kernels. There are a couple of branches that were last worked on about 2 years ago. "pm-rework" and "acpi-pr3" So I tried to install the "pm-rework" branch of https://github.com/Bumblebee-Project/bbswitch/tree/pm-rework but I get this error while buidling on centos 7: https://paste.fedoraproject.org/paste/F1B2wZN-XvJw0dVCVeXY7w Using a version number I made up, bbswitch-dkms-0.8.5-1.el7.x86_64.rpm I also tried "acpi-pr3" branch and that actually compiled and loaded ok. It appears to work on my lenovo ideapad y470. Using a version number I made up, bbswitch-dkms-0.8.6-1.el7.x86_64.rpm I'm not sure any of the other branches would be worth building since they are 5 and 7 years old. In the interest of full disclosure, I have uploaded both sets of packages to: (not working for me, pm-rework) (appears to work for me, acpi-pr3) So I guess you could give those a try... If none of this helps you may wish to open a new issue report on https://github.com/Bumblebee-Project/bbswitch and perhaps the author, @Lekensteyn might have some ideas? If so reference this issue so they can see what has been found out so far. On the bumblebee side there is a "xpra-backend" branch that was updated 23 days ago by @Thulinma but I don't think it would be useful to you since it simply adds xpra support and xpra package is not available for centos 7. (first attempt at Vulcan API games support) If they (The bumblebee developers) get that sorted out I was going to look into how hard it might be to build xpra for centos 7 and add it to my yum repo. And patch the bumblebee rpm also. Also on the bumblebee side there is a "develop" branch which might be useful... ? I think more work would be needed to get this working on CentOS 6 / RHEL 6 so I will not be able to copy those rpms. It could probably be made to work I'm guessing.... Right now it will not even compile. It worked on CentOS 7 but I had to disable selinux. So clearly more work would need to be done in the bumblebee-nvidia package with its selinux policy module to make it work with selinux in enforcing mode on RHEL 7 and fedora. I will try to look into that when I have more time. (for now /etc/sysconfig/selinux SELINUX=permissive ) Using a version number I made up, bumblebee-3.3.0-1.el7.x86_64.rpm https://linux.itecs.ncsu.edu/redhat/public/bumblebee/experimental2/develop/ So I guess you could give that a try... Perhaps it could help? Your 20180802.xorg.conf.nvidia.txt file looks fine since I forgot I only patch that on fedora, not RHEL or CentOS. |
Dear gsgatlin, [wrthissell@LAPTOP-BKIJEPGK ~]$ cat /proc/acpi/bbswitch |
Dear gsgatlin, [wrthissell@LAPTOP-BKIJEPGK ~]$ bumblebee-nvidia --check nvidia.ko compiled into in the kernel tree ok. filename: /lib/modules/3.10.0-862.9.1.el7.x86_64/kernel/drivers/video/nvidia.ko Check bbswitch kernel module... bbswitch is loaded into the current kernel ok. All checks completed successfully! NVIDIA driver appears to have compiled ok. Documentation on bumblebee for RHEL / CentOS / fedora can be found at: [wrthissell@LAPTOP-BKIJEPGK ~]$ optirun --debug glxgears [ 712.319668] [ERROR]Cannot access secondary GPU - error: [XORG] (EE) NVIDIA(GPU-0): Failed to initialize the NVIDIA GPU at PCI:1:0:0. Please [ 712.319688] [DEBUG]Socket closed. [ 738.304649] [ERROR]Aborting because fallback start is disabled.
20180803 Xorg.8.log.txt
|
Dear gsgatlin, nvidia.ko compiled into in the kernel tree ok. filename: /lib/modules/3.10.0-862.9.1.el7.x86_64/kernel/drivers/video/nvidia.ko Check bbswitch kernel module... bbswitch is loaded into the current kernel ok. All checks completed successfully! NVIDIA driver appears to have compiled ok. Documentation on bumblebee for RHEL / CentOS / fedora can be found at: [wrthissell@LAPTOP-BKIJEPGK ~]$ optirun --debug glxgears [ 203.219397] [ERROR]Cannot access secondary GPU - error: [XORG] (EE) NVIDIA(GPU-0): Failed to initialize the NVIDIA GPU at PCI:1:0:0. Please [ 203.219408] [DEBUG]Socket closed. [ 230.774937] [ERROR]Aborting because fallback start is disabled. Below are the relevant log files: 20180803 Xorg.8.log#2.txt I shall now try to install bumblebee-3.3.0-1.el7.x86_64.rpm, reboot and try again.
|
Dear gsgatlin, |
Dear gsgatlin, nvidia.ko compiled into in the kernel tree ok. filename: /lib/modules/3.10.0-862.9.1.el7.x86_64/kernel/drivers/video/nvidia.ko Check bbswitch kernel module... bbswitch is loaded into the current kernel ok. All checks completed successfully! NVIDIA driver appears to have compiled ok. Documentation on bumblebee for RHEL / CentOS / fedora can be found at: [wrthissell@LAPTOP-BKIJEPGK ~]$ optirun --debug glxgears [ 240.703241] [ERROR]Cannot access secondary GPU - error: [XORG] (EE) NVIDIA(GPU-0): Failed to initialize the NVIDIA GPU at PCI:1:0:0. Please [ 240.703246] [DEBUG]Socket closed.
20180803 Xorg.8.log#3.txt
|
Dear gsgatlin, Bumblebee-Project/bbswitch#112 I just added the following text to the kernel load command: 'acpi_osi=!Windows\x202013' acpi_osi=Linux nogpumanager intel_iommu=on I shall reboot and see if this fixes the issue. |
Dear gsgatlin, Bumblebee-Project/bbswitch#119 https://github.com/doudou/bbswitch
|
Dear gsgatlin, 'acpi_osi=!Windows\x202013' acpi_osi=Linux nogpumanager intel_iommu=on
[wrthissell@LAPTOP-BKIJEPGK ~]$ optirun --debug glxgears [ 141.901769] [ERROR]Cannot access secondary GPU - error: [XORG] (EE) NVIDIA(GPU-0): Failed to initialize the NVIDIA GPU at PCI:1:0:0. Please [ 141.901775] [DEBUG]Socket closed. nvidia.ko compiled into in the kernel tree ok. filename: /lib/modules/3.10.0-862.9.1.el7.x86_64/kernel/drivers/video/nvidia.ko Check bbswitch kernel module... bbswitch is loaded into the current kernel ok. All checks completed successfully! NVIDIA driver appears to have compiled ok. Documentation on bumblebee for RHEL / CentOS / fedora can be found at: [wrthissell@LAPTOP-BKIJEPGK ~]$ cat /proc/acpi/bbswitch
|
Dear gsgatlin, https://www.centos.org/forums/viewtopic.php?t=66178 after removing the kernel load option and executing the dracut --regenerate-all --force: 'acpi_osi=!Windows\x202013' acpi_osi=Linux nogpumanager intel_iommu=on Here are the results (no change): [wrthissell@LAPTOP-BKIJEPGK ~]$ optirun --debug glxgears [ 3277.503238] [ERROR]Cannot access secondary GPU - error: [XORG] (EE) NVIDIA(GPU-0): Failed to initialize the NVIDIA GPU at PCI:1:0:0. Please [ 3277.503261] [DEBUG]Socket closed. nvidia.ko compiled into in the kernel tree ok. filename: /lib/modules/3.10.0-862.9.1.el7.x86_64/kernel/drivers/video/nvidia.ko Check bbswitch kernel module... bbswitch is loaded into the current kernel ok. All checks completed successfully! NVIDIA driver appears to have compiled ok. Documentation on bumblebee for RHEL / CentOS / fedora can be found at: [wrthissell@LAPTOP-BKIJEPGK ~]$ cat /proc/acpi/bbswitch
|
I don't want to speak for anyone here but... Since there is silence... I think maybe most of the developers of this project are no longer all that interested. Plus I think a lot of them don't use bumblebee or bbswitch anymore and use nouveau drivers for power saving instead. Its hard to program something in your spare time if you don't use it any more. This is just my observations after making some bumblebee rpm packages for myself. Maybe you could try irc.freenode.net channel bumblebee but... I'm not sure that could be any better. Its up to you. Your best bet at this point might be to try to figure out how to make drivers at elrepo.org or rpmfusion.org or negativo17.org work if you are interested in 3d opengl gaming on Linux. I think this is where you have to reboot or at least log out and then log back in to switch drivers between intel and nvidia. I'm not really sure how it works in red hat. Then if it doesn't work you can make an account at https://devtalk.nvidia.com/default/board/98/linux/ and post about your problem and a actual nvidia employee might be able to help. Or someone in the community. And then maybe that might help later with bumblebee but I'm not sure about that last part. Like here are Simone Caronni's docs: (A.K.A. negativo17) https://negativo17.org/nvidia-driver/ I've never actually tested this yet but a lot of people like what he is doing. Look at the section called "Optimus laptops" I'm pretty sure he supports RHEL 7 and CentOS 7 and fedora. You might also want to check and see if some other distros have the same problem. Like you could plug in a external usb drive and boot from that and do an install onto that to see if ubuntu or fedora or whatever distro you pick gives you better luck with bumblebee or nvidia. And it would not destroy any data on your centos 7 hard drive partitions if you take care not to select the main laptop hard disk in anaconda or maybe ubuntu or whatever installer. Backup all data first just to be extra safe though. Oh and also might want to upload your laptop info here? DSDT.dsl and SSDT tables... https://launchpad.net/~hybrid-graphics-linux maybe no one cares about that nowadays? But I guess it would cover all bases. Red Hat can't put any kind of nvidia drivers in any of their distros (RHEL, CentOS, fedora) because it violates their rules about no closed source software being allowed in the distro ever. So because of that, third party yum/dnf repos are your only chance to get nvidia working and its very hit or miss with that. Some other distros like Arch or Ubuntu have no such rules so the drivers are better integrated into the distro maybe. (although ubuntu has a ppa for nvidia also so go figure...) If all you care about is power saving maybe nouveau could work. https://nouveau.freedesktop.org/wiki/Optimus/ Then you don't have to install anything "out of tree" and theoretically everything should just work right out of the box. Performance is not that great with nouveau PRIME but some people don't care about that and they never even use their DGPU. Fundamentally IMHO the whole problem is caused by: https://github.com/torvalds/linux/blob/master/Documentation/process/stable-api-nonsense.rst Because bbswitch and nvidia modules are out of tree kernel modules. Like if nvidia would just follow the rules about GPL there would not be a need for bbswitch to even exist. And nvidia would just be a part of the kernel. And it would have power saving built in just like nouveau and raedon do. I'm sorry I was not able to help figure out what was wrong here. I've never been a kernel hacker at all. I tried to help around here a little bit and give back some since they helped me so much back in the day but I am a person of limited troubleshooting ability sadly. These are just my opinions and observations and are not a statement on behalf of bumblebee project... |
Dear gsgatlin, https://linux.itecs.ncsu.edu/redhat/public/bumblebee/experimental2/acpi-pr3/
[wrthissell@LAPTOP-BKIJEPGK ~]$ bumblebee-nvidia --check nvidia.ko compiled into in the kernel tree ok. filename: /lib/modules/3.10.0-862.9.1.el7.x86_64/kernel/drivers/video/nvidia.ko Check bbswitch kernel module... bbswitch is loaded into the current kernel ok. All checks completed successfully! NVIDIA driver appears to have compiled ok. Documentation on bumblebee for RHEL / CentOS / fedora can be found at: [wrthissell@LAPTOP-BKIJEPGK ~]$ optirun -b none nvidia-settings -c :8 [ 227.075836] [ERROR]Aborting because fallback start is disabled. [ 253.424066] [ERROR]Cannot access secondary GPU - error: [XORG] (EE) NVIDIA(GPU-0): Failed to initialize the NVIDIA GPU at PCI:1:0:0. Please [ 253.424071] [DEBUG]Socket closed.
20180812 bbswitch-0.8.7-1.tar.gz
https://bugs.launchpad.net/lpbugreporter/+bug/752542
|
Dear gsgatlin, The file below is before I installed xpra. Here is how to enable the xpra repository for Centos: https://www.xpra.org/trac/wiki/Download#Linux sudo yum install xpra* installs the latest version after placing the repo file where it belongs. 20180814 bbswitch-thinkpad.tar.gz I shall now execute dracut --regenerate-all --force reboot and try again with xpra installed. |
Hmnn. seems bbswitch still is unable to turn off your discrete card even with thinkpad patch. |
Dear gsgatlin, bumblebee-3.3.0-1.el7.x86_64.rpm See below for my attempt at a remedy: [wrthissell@LAPTOP-BKIJEPGK ~]$ sudo ausearch -c 'bumblebeed' --raw | audit2allow -M my-bumblebeed
Now I shall execute the dracut --regenerate-all --force reboot and try again with xpra installed. |
1 similar comment
Dear gsgatlin, bumblebee-3.3.0-1.el7.x86_64.rpm See below for my attempt at a remedy: [wrthissell@LAPTOP-BKIJEPGK ~]$ sudo ausearch -c 'bumblebeed' --raw | audit2allow -M my-bumblebeed
Now I shall execute the dracut --regenerate-all --force reboot and try again with xpra installed. |
Dear gsgatlin,
|
Dear gsgatlin, |
Glad you got it working. |
Dear Colleagues,
I previously had Bumblebee working on a Lenovo Thinkpad W541 with Centos 6.8. I no longer have access to that computer. I moved the partitions to a Thinkpad P50, yum update to Centos 6.9, and I also installed Centos 7.3-7.5. I have been unable to get Bumblebee to work on the new computer and the Centos versions.
I submitted a bug report last November at:
https://bugs.launchpad.net/lpbugreporter/+bug/752542
A search for my username will result in the files I uploaded then.
Here are my current results:
[wrthissell@new-host-4 ~]$ uname -r
3.10.0-862.9.1.el7.x86_64
[wrthissell@new-host-4 ~]$ optirun --debug glxgears64
[ 233.925594] [DEBUG]optirun version 3.2.1 starting...
[ 233.925631] [DEBUG]Active configuration:
[ 233.925645] [DEBUG] bumblebeed config file: /etc/bumblebee/bumblebee.conf
[ 233.925668] [DEBUG] X display: :8
[ 233.925689] [DEBUG] LD_LIBRARY_PATH: /usr/lib64/nvidia-bumblebee:/usr/lib/nvidia-bumblebee:/usr/lib64:/usr/lib
[ 233.925703] [DEBUG] Socket path: /var/run/bumblebee.socket
[ 233.925719] [DEBUG] Accel/display bridge: auto
[ 233.925733] [DEBUG] VGL Compression: proxy
[ 233.925752] [DEBUG] VGLrun extra options:
[ 233.925766] [DEBUG] Primus LD Path: /usr/lib/primus:/usr/lib32/primus:/usr/lib64/primus
[ 233.925892] [DEBUG]Using auto-detected bridge virtualgl
[ 238.692727] [INFO]Response: No - error: [XORG] (EE) /dev/dri/card0: failed to set DRM interface version 1.4: Permission denied
[ 238.692776] [ERROR]Cannot access secondary GPU - error: [XORG] (EE) /dev/dri/card0: failed to set DRM interface version 1.4: Permission denied
[ 238.692803] [DEBUG]Socket closed.
[ 238.692865] [ERROR]Aborting because fallback start is disabled.
[ 238.692899] [DEBUG]Killing all remaining processes.
[wrthissell@new-host-4 ~]$
http://elrepo.org/bugs/view.php?id=737
http://elrepo.org/bugs/view.php?id=742
20170725 Bumblebee Debug.tar.gz
The text was updated successfully, but these errors were encountered: