https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c2 George Baltz changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |GeoBaltz@gmail.com --- Comment #2 from George Baltz 2014-05-24 14:59:27 UTC --- I have about the same problem with the 331.67 driver on 13.1 - no CUDA support found in BOINC. While looking around I found this thread https://devtalk.nvidia.com/default/topic/699610/linux/334-21-driver-returns-... which shows others with the same problem - I'm trying to adapt that to my openSUSE system but no luck yet. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c Stefan Dirsch changed: What |Removed |Added ---------------------------------------------------------------------------- Priority|P5 - None |P3 - Medium Status|NEW |ASSIGNED Component|Other |X11 3rd Party Driver Summary|1 click install nvidia |CUDA support broken with |driver 331.67-26-1x86_64 |331.67 |has no gpu support anymore | QAContact|qa-bugs@suse.de |sndirsch@suse.com -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c Stefan Dirsch changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |NEEDINFO InfoProvider| |ddadap@nvidia.com -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c5 Stefan Dirsch changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |ASSIGNED InfoProvider|ddadap@nvidia.com | --- Comment #5 from Stefan Dirsch 2014-05-28 10:39:03 UTC --- Thanks a lot for the input, Daniel. I agree, that it should be possible to fix the issue by an appropriate udev rule, so the required device files with permissions for group video are created. But first I need to find a NVIDIA gfx card with CUDA support. ;-) -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c7 Stefan Dirsch changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |NEEDINFO InfoProvider| |antonvdh@gmail.com --- Comment #7 from Stefan Dirsch 2014-05-28 13:47:55 UTC --- Anton/George, could you give the "fix" a try? Thanks! -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c9 --- Comment #9 from Stefan Dirsch 2014-05-28 15:51:57 UTC --- Good news is, that thanks to modaliases in nvidia kernel module, nvidia and - due to my changes - also nvidia_uvm are now loaded even without X in place during boot in initrd. /dev/nvidia_uvm is also created. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c11 --- Comment #11 from Daniel Dadap 2014-05-28 16:06:55 UTC ---

I need to find a NVIDIA gfx card with CUDA support.

FWIW, all GPUs supported by drivers newer than the 304 series support CUDA. So if you are running a 331 series driver, you definitely have a GPU with CUDA support. You can download the CUDA Toolkit, which includes all of the tools needed for building CUDA programs along with sample programs at: https://developer.nvidia.com/cuda-toolkit -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c13 George Baltz changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |ASSIGNED InfoProvider|antonvdh@gmail.com | --- Comment #13 from George Baltz 2014-05-29 01:40:19 UTC --- Tried both patches. The first one hung my 13.1 system at boot time, the second one worked - it built /dev/nvidia-uvm with the correct permissions, and BOINC seems happy with it. Does dracut have/support an /etc/group for chgrp to refer to? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c15 --- Comment #15 from Stefan Dirsch 2014-06-02 12:16:55 UTC --- /etc/group is available in initrd. Instead of chgrp Thomas wants to add chown, which is fine. This alone doesn't help much since in newer openSUSE releases and sle12 our standard users are no longer added to the video group. Instead "rw" permission need to be added to the NVIDIA devices for the user once he logs in. This is already done for /dev/nvidiactl and /dev/nvidiaX. So we just need to add /dev/nvidia-uvm here. Unfortunately for CUDA support /dev/nvidiactl and /dev/nvidiaX also need to exist and they aren't created before X has been started, although nvidia driver has been loaded (required by nvidia-uvm module). One could add /dev/nvidiactl and /dev/nvidiaX manually, but I'm not sure how many /dev/nvidiaX devices to create, since I'm not sure how to figure out how many GPUs do exist on the card. I believe there is one for each GPU. Maybe creating /dev/nvidiactl and /dev/nvidia0 will cover most cases. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c17 --- Comment #17 from Stefan Dirsch 2014-06-02 12:41:43 UTC --- Created an attachment (id=592974) --> (http://bugzilla.novell.com/attachment.cgi?id=592974) apply-ACL-for-nvidia-device-nodes.patch Updated systemd patch -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c19 --- Comment #19 from Stefan Dirsch 2014-06-03 13:11:39 UTC --- Created an attachment (id=593119) --> (http://bugzilla.novell.com/attachment.cgi?id=593119) apply-ACL-for-nvidia-uvm-device-node.patch Additional ACL patch on top of Ludwig's ACL patch -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c22 Stefan Dirsch changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |RESOLVED Resolution| |FIXED --- Comment #22 from Stefan Dirsch 2014-06-03 14:57:53 UTC --- Closing as fixed. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c25 --- Comment #25 from Bernhard Wiedemann 2014-06-03 19:00:13 CEST --- This is an autogenerated message for OBS integration: This bug (879767) was mentioned in https://build.opensuse.org/request/show/236162 Factory / systemd -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c31 --- Comment #31 from Swamp Workflow Management 2014-06-17 10:04:23 UTC --- openSUSE-RU-2014:0803-1: An update that has one recommended fix can now be installed. Category: recommended (low) Bug References: 879767 CVE References: Sources used: openSUSE 13.1 (src): systemd-208-23.3, systemd-mini-208-23.1, systemd-rpm-macros-2-23.1 -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c33 --- Comment #33 from Episteme PROMENEUR 2014-06-20 17:52:31 UTC --- also flashplayer no more uses hw acceleration for rendering -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c35 --- Comment #35 from Episteme PROMENEUR 2014-06-23 07:19:43 UTC --- (In reply to comment #34)

Still seem to be the old 331.79 packages, since you now also have the old 3.11.6 kernel packages in there. See bnc#881723.

as you can read in my comment #32 date of the report is date of the installation of 331.79 : 2014 06 20 you say i installed old buggy 331.79 package . where to find the non buggy 331.79 ? how to have a consistent installation of good kernel and nvidia packages from the current state of my installation ? thanks -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c37 --- Comment #37 from Episteme PROMENEUR 2014-06-23 09:47:11 UTC --- change log for - nvidia-uvm-gfxG03-kmp-desktop - NVIDIA Unified Memory kernel module ven. 06 juin 2014 14:00:00 CEST sndirsch@suse.com - get rid of the requirement to a specific kernel version (bnc#881723) mar. 03 juin 2014 14:00:00 CEST sndirsch@suse.com - updated to latest long lived branch version 331.79 * Fixed a bug that caused nvidia-installer to crash in environments where /proc is not mounted. * Fixed a bug that prevented module signing from working correctly for the NVIDIA Unified Memory kernel module when using module signing keys generated by nvidia-installer. * Fixed a bug that caused blank screens and flickering when rotating displays in a Base Mosaic layout. * Fixed a bug that caused BadRRCrtc or BadRROutput errors for big-endian X11 clients making certain XRandR requests. * Fixed a bug that corrupted certain software rendering, notably the stippled text used to represent disabled entries in xterm's pop-up menus. lun. 02 juin 2014 14:00:00 CEST sndirsch@suse.com - autoselect to install also nvidia-uvm KMP; load nvidia-uvm in addition when nvidia module gets loaded; create the appropriate NVIDIA devices with the required permissions (bnc#879767) but there is not any changelog for 11 juin 2014 thus no "* Mi Jun 11 2014 sndirsch@suse.com - require the *flavor* KMP " - nvidia-gfxG03-kmp-desktop - NVIDIA graphics driver kernel module for GeForce 8xxx and newer GPUs mar. 03 juin 2014 14:00:00 CEST sndirsch@suse.com - updated to latest long lived branch version 331.79 * Fixed a bug that caused nvidia-installer to crash in environments where /proc is not mounted. * Fixed a bug that prevented module signing from working correctly for the NVIDIA Unified Memory kernel module when using module signing keys generated by nvidia-installer. * Fixed a bug that caused blank screens and flickering when rotating displays in a Base Mosaic layout. * Fixed a bug that caused BadRRCrtc or BadRROutput errors for big-endian X11 clients making certain XRandR requests. * Fixed a bug that corrupted certain software rendering, notably the stippled text used to represent disabled entries in xterm's pop-up menus. lun. 02 juin 2014 14:00:00 CEST sndirsch@suse.com - autoselect to install also nvidia-uvm KMP; load nvidia-uvm in addition when nvidia module gets loaded; create the appropriate NVIDIA devices with the required permissions (bnc#879767) linux-b4lz:~ # getfacl /dev/nvidia* getfacl: Removing leading '/' from absolute path names # file: dev/nvidia-uvm # owner: root # group: root user::rw- group::rw- other::rw- # file: dev/nvidia0 # owner: root # group: video user::rw- group::rw- other::--- # file: dev/nvidiactl # owner: root # group: video user::rw- group::rw- other::--- linux-b4lz:~ #

...

Sure you've also updated systemd package?

you ask for me if i updated systemd ? don't know . i don't update that kind of thing myself -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c39 --- Comment #39 from Stefan Dirsch 2014-06-23 10:03:27 UTC --- You're right. Latest changes - require the *flavor* KMP (bnc#881723) (uvm KMP) - recommend the *flavor* UVM KMP (bnc#881723) (base KMP) have not been pushed to NVIDIA yet. Will be done, once a new NVIDIA driver version will become available. But these fixes are unrelated to your issue. You definitely need the systemd update. It should become available via the normal openSUSE update repo - in case it isn't there yet. If you're not interested in verifying, that this change is in, I cannot help you. Also you need to run the getfacl command as the logged in user. root can access the /dev/nvidia* devices anyway. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c41 --- Comment #41 from Episteme PROMENEUR 2014-06-23 10:17:38 UTC --- roubach@linux-b4lz:~> getfacl /dev/nvidia* getfacl : suppression du premier « / » des noms de chemins absolus # file: dev/nvidia0 # owner: root # group: video user::rw- group::rw- other::--- # file: dev/nvidiactl # owner: root # group: video user::rw- group::rw- other::--- # file: dev/nvidia-uvm # owner: root # group: root user::rw- group::rw- other::rw- roubach@linux-b4lz:~> -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c43 --- Comment #43 from Episteme PROMENEUR 2014-06-23 10:34:04 UTC --- suddenly i inderstand a thing : when executing sudo usermod -a -G video $USER i got this roubach@linux-b4lz:~> sudo usermod -a -G video $USER root's password: sudo: usermod: command not found roubach@linux-b4lz:~> then in root console i executed "usermod -a -G video $USER" this was wrong then i have just with yast tool put "roubach" in group "video" then i restarted pc then now in nvidia settings i have : - menu option vdpau - direct rendering yes but fahcontrol still says cuda not detected and still same roubach@linux-b4lz:~> getfacl /dev/nvidia* getfacl : suppression du premier « / » des noms de chemins absolus # file: dev/nvidia0 # owner: root # group: video user::rw- group::rw- other::--- # file: dev/nvidiactl # owner: root # group: video user::rw- group::rw- other::--- # file: dev/nvidia-uvm # owner: root # group: root user::rw- group::rw- other::rw- roubach@linux-b4lz:~> -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c45 --- Comment #45 from Episteme PROMENEUR 2014-06-23 11:37:26 UTC --- i go back to 208-23.3 for systemd and udev now we have user:roubach:rw- roubach@linux-b4lz:~> getfacl /dev/nvidia* getfacl : suppression du premier « / » des noms de chemins absolus # file: dev/nvidia0 # owner: root # group: video user::rw- user:roubach:rw- group::rw- mask::rw- other::--- # file: dev/nvidiactl # owner: root # group: video user::rw- user:roubach:rw- group::rw- mask::rw- other::--- # file: dev/nvidia-uvm # owner: root # group: root user::rw- user:roubach:rw- group::rw- mask::rw- other::rw- roubach@linux-b4lz:~> but still "cuda not detected" -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c47 --- Comment #47 from Episteme PROMENEUR 2014-06-23 11:51:13 UTC --- is it necessary to go back to 14-2 intead of 17-69 for kmod ? because no pb for kmod and libkmod2 but for kmod-compat yast software manager tool - does not fing module init tool - want to uninstall may packages see the capture of yast tool dialog -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c49 --- Comment #49 from Episteme PROMENEUR 2014-06-23 12:16:31 UTC --- what is the repo for cuda packages ? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c51 Stefan Dirsch changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |RESOLVED InfoProvider|antonvdh@gmail.com | Resolution| |FIXED --- Comment #51 from Stefan Dirsch 2014-06-23 12:28:53 UTC --- I don't know anything about these dkms-nvidia packages. Possibly these contained CUDA packages themselves. Permissions are now being set correctly with the official systemd package update. Closing as fixed again. I cannot reproduce that issue. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c53 Wolfgang Bauer changed: What |Removed |Added ---------------------------------------------------------------------------- Status|RESOLVED |REOPENED CC| |wbauer@tmo.at Component|X11 3rd Party Driver |X11 3rd Party Driver Resolution|FIXED | AssignedTo|sndirsch@suse.com |bnc-team-screening@forge.pr | |ovo.novell.com Product|openSUSE 12.3 |openSUSE 13.1 --- Comment #53 from Wolfgang Bauer 2014-08-27 14:23:07 UTC --- Sorry for joining in late, but I wasn't able to test this before. Unfortunately this still doesn't work with the latest 340.32-31.1 packages here on 13.1 with a GTX 760 card (and neither on another 13.1 system with a GTX 550 Ti). The problem is that there is no /dev/nvidia-uvm after boot: # ls -l /dev/nvi* crw-rw----+ 1 root video 195, 0 27. Aug 15:47 /dev/nvidia0 crw-rw----+ 1 root video 195, 255 27. Aug 15:47 /dev/nvidiactl # If I run the commands from /etc/modprobe.d/50-nvidia.conf manually, the device appears and CUDA works. A bit of investigation revealed that the nvidia module is loaded from the initrd already before the root filesystem is mounted. But the nvidia-uvm module is _not_ added to the initrd, so cannot be loaded. Adding "nvidia-uvm" to INITRD_MODULES in /etc/sysconfig/kernel fixes this, but has another problem: chown is not contained in the initrd either (on 13.1 that is, it seems to be available there in Factory), so you get error messages during boot (before plymouth shows up). /dev/nvidia-uvm therefore has the group root instead of video, but it gets the correct ACLs by systemd-logind so it works even as user. Setting NO_KMS_IN_INITRD="yes" fixes all problems, as the nvidia module is then loaded later when the root filesystem is available. So I suppose the easiest fix would be that the driver package sets NO_KMS_IN_INITRD="yes" again on installation (although I'm not really fond of that, see bnc#864701 ... ;) Btw, the problem some people reported (which I can reliably reproduce here), that the nvidia driver does not work on first boot after installation/update, is apparently also caused by this. Adding nvidia-uvm to the initrd or NO_KMS_IN_INITRD="yes" fixes this for me as well. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c55 --- Comment #55 from Wolfgang Bauer 2014-08-28 04:25:12 UTC --- (In reply to comment #54)

Hmm. With running mkinitrd, both modules - 'nvidia' and 'nvidia-uvm' - should have been added to initrd. This is at least happening with sle12 (simlar behaviour with factory).

Can you verify this?

On 13.1 there is no nvidia-uvm in the initrd: (unless I manually add it via /etc/sysconfig/kernel->INITRD_MODULES) # lsinitrd /boot/initrd|grep nvidia usr/lib64/libnvidia-glsi.so.340.32 usr/lib64/libnvidia-glcore.so.340.32 usr/lib64/tls/libnvidia-tls.so.340.32 lib/modules/3.11.10-21-desktop/weak-updates/updates/nvidia.ko etc/modprobe.d/50-nvidia.conf etc/modprobe.d/nvidia-desktop.conf # nvidia-uvm-kmp-gfxG03-desktop is installed of course: # rpm -qa | grep nvidia nvidia-gfxG03-kmp-desktop-340.32_k3.11.6_4-31.1.x86_64 nvidia-glG03-340.32-31.1.x86_64 x11-video-nvidiaG03-340.32-31.1.x86_64 nvidia-gfxG03-kmp-desktop-331.79_k3.11.6_4-27.1.x86_64 nvidia-gfxG03-kmp-desktop-331.89_k3.11.6_4-28.1.x86_64 nvidia-gfxG03-kmp-desktop-331.49_k3.11.6_4-29.1.x86_64 nvidia-uvm-gfxG03-kmp-desktop-340.32_k3.11.6_4-31.1.x86_64 nvidia-uvm-gfxG03-kmp-desktop-331.79_k3.11.6_4-27.1.x86_64 nvidia-uvm-gfxG03-kmp-desktop-331.89_k3.11.6_4-28.1.x86_64 nvidia-computeG03-340.32-31.1.x86_64 nvidia-gfxG03-kmp-desktop-331.67_k3.11.6_4-26.1.x86_64 # ls -l /lib/modules/3.11.10-21-desktop/weak-updates/updates/nvidia* lrwxrwxrwx 1 root root 47 14. Aug 19:08 /lib/modules/3.11.10-21-desktop/weak-updates/updates/nvidia.ko -> /lib/modules/3.11.6-4-desktop/updates/nvidia.ko lrwxrwxrwx 1 root root 51 14. Aug 19:08 /lib/modules/3.11.10-21-desktop/weak-updates/updates/nvidia-uvm.ko -> /lib/modules/3.11.6-4-desktop/updates/nvidia-uvm.ko And as mentioned chown is missing in the initrd as well, so even if nvidia-uvm is added, the device will be owned by root:root instead of root:video (but systemd-logind will give the user permissions to access it anyway via ACLs), and there are error messages ("chmod: command not found" or similar) after the boot menu, before the bootsplash appears. # lsinitrd /boot/initrd|grep chown # I suppose it's the same on 12.3. But there the nvidia driver package still sets NO_KMS_IN_INITRD="yes" which means that nvidia.ko is not part of the initrd either. Therefore both are loaded later when the root filesystem is mounted already, so it works. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c57 --- Comment #57 from Wolfgang Bauer 2014-08-29 07:27:08 UTC --- (In reply to comment #56)

Ok. Convinced. Going back to NO_KMS_IN_INITRD="yes" for openSUSE 13.1. Will be in place with the next driver update (> 340.32). I've reopened bnc#864701.

Thank you. This should prevent the problems on 13.1 (nvidia-uvm not loaded, driver doesn't work on first boot after installation). I just hope the latter problem does not re-appear when nvidia-uvm is not installed (it's only recommended after all). I think I'm going to test that in the next days, but it's not really related to this bug report anyway. Just another thing, as I re-read this whole bug report now: (In reply to comment #21)

- no fix required for openSUSE 12.3 (we still use video group there to access nvidia devices; new users are still added to this group)

That's wrong. 12.3 was the first release to _not_ add users to the video group. See also bnc#808319 (reported against 12.3, then later reopened and changed to 13.1 by me as the systemd patch was not working correctly there), bnc#809163, and the 12.3 release notes: https://www.suse.com/releasenotes/x86_64/openSUSE/12.3/#idm47462154153376 I leave it up to you if you want to release that update for 12.3 as well though... ;) -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=879767 https://bugzilla.novell.com/show_bug.cgi?id=879767#c59 --- Comment #59 from Wolfgang Bauer 2014-08-29 08:15:02 UTC --- (In reply to comment #58)

Seems the latest fixes/updates with the most up-to-date patch for systemd have only been done for openSUSE 13.1, and not for openSUSE 12.3.. -(

12.3's systemd does contain the same patch to give the logged in user access to /dev/nvidiaX and /dev/nvidiactl via ACLs. It was not part of the release, only added afterwards via an online update. And the patch is named differently there, logind-nvidia-acl.diff instead of apply-ACL-for-nvidia-device-nodes.patch: https://build.opensuse.org/package/view_file/openSUSE:12.3:Update/systemd/lo... But yes, it is documented, and I suppose most 12.3 users added themselves to the video group anyway to get the nvidia driver working before that update was released. And not everyone uses CUDA either... -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.