| Summary: | backport request of nvidia-current-367.57 for GTX1080 | ||
|---|---|---|---|
| Product: | Mageia | Reporter: | Chris Denice <eatdirt> |
| Component: | Backports | Assignee: | QA Team <qa-bugs> |
| Status: | RESOLVED FIXED | QA Contact: | |
| Severity: | enhancement | ||
| Priority: | Normal | CC: | davidwhodgins, isobuild, mageia, marja11, tarazed25, tmb |
| Version: | 5 | Keywords: | validated_backport |
| Target Milestone: | --- | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Whiteboard: | MGA5-64-OK | ||
| Source RPM: | nvidia-current-367.57-1.mga5.src.rpm | CVE: | |
| Status comment: | |||
|
Description
Chris Denice
2016-10-25 20:45:15 CEST
(In reply to Chris Denice from comment #0) > I have just bought a nvidia GTX1080, which does not work as expected with > mga5 (no accel) only the vesa driver allows the display. > > I have backported locally nvidia-current-367.57 from cauldron on mga5, and > everything works fine then. > > I guess we want it for 5.1! > > Cheers, > Chris. Assigning to the kernel and drivers maintainers and CC'ing the isobuilders CC:
(none) =>
isobuild, marja11 I'd rather not push it as an official update in this case as nVidia changed a lot of stuff between 352.79 (currently in updates) and 367.57 that can introduce regresseions for users of older hw depending on nvidia-current... So turning this into a Backport request. You need ldetect-lst from core/backports_testing for drakx tools to properly detect supported nvidia ids. SRPMS: ldetect-lst-0.1.346.4-2.mga5.src.rpm nvidia-current-367.57-1.mga5.nonfree.src.rpm i586: ldetect-lst-0.1.346.4-2.mga5.i586.rpm ldetect-lst-devel-0.1.346.4-2.mga5.i586.rpm dkms-nvidia-current-367.57-1.mga5.nonfree.i586.rpm nvidia-current-cuda-opencl-367.57-1.mga5.nonfree.i586.rpm nvidia-current-devel-367.57-1.mga5.nonfree.i586.rpm nvidia-current-doc-html-367.57-1.mga5.nonfree.i586.rpm x11-driver-video-nvidia-current-367.57-1.mga5.nonfree.i586.rpm x86_64: ldetect-lst-0.1.346.4-2.mga5.x86_64.rpm ldetect-lst-devel-0.1.346.4-2.mga5.x86_64.rpm dkms-nvidia-current-367.57-1.mga5.nonfree.x86_64.rpm nvidia-current-cuda-opencl-367.57-1.mga5.nonfree.x86_64.rpm nvidia-current-devel-367.57-1.mga5.nonfree.x86_64.rpm nvidia-current-doc-html-367.57-1.mga5.nonfree.x86_64.rpm x11-driver-video-nvidia-current-367.57-1.mga5.nonfree.x86_64.rpm CC:
(none) =>
tmb I hope to get round to this in a couple of days. There are four nvidia machines here suitable for testing cards older than the GTX1080. CC:
(none) =>
tarazed25 i586 on x86_64 : nvidia GT 8700M Can't seem to get started on this one. Have enabled backports testing, nonfree backports testing and nonfree testing. Installed ldetect-lst OK but cannot get the dkms-nvidia-current package installed. Have run MageiaUpdate. # urpmi dkms-nvidia-current A requested package cannot be installed: dkms-nvidia-current-367.57-1.mga5.nonfree.i586 (due to unsatisfied dkms[*][>= 2.0.19-37]) dkms is at 2.0.19-34.1 Which repository needs to be enabled for this to upgrade? Ah, sorry, I forgot to reset dkms deps in the backport... will fix that up... I've removed and rebuilt the nvidia-current rpms as listed in comment 2, so you can try again ... Thanks Thomas. That worked. There seems to be another problem;on i586 or x86_64 ldetect-lst, dkms-nvidia-current and nvidia-current-cuda-opencl install and kmod is built but neither nvidia-current-devel nor x11-driver-video-nvidia-current can be installed due to unsatisfied ldetect-lst-0.1.346.4-2 (which had just been successfully installed). The library is libldetect.so.0.12.8 but the datestamp is 2015-02-24. I am at a loss here. Sorry, seems I made another error with this backport, having a typo in requires on ldetect-lst.... :/ I've nuked the rpms again, and submitted a new build .... should be available in a an hour or so depending on mirror. Thanks again Thomas. Mate desktop nvidia driver 367.57 up and running on x86_64 nvidia GeForce GTX 970 with 4K DisplayPort monitor. x86_64 Mate nVidia GeForce GTX 770 driving 3K display on DVI NVIDIA 367.57 works fine. x86_64 Mate NVIDIA Corporation GM204M [GeForce GTX 965M] * 2 Panasonic 2880x1620 panel on GPU 0 Enabled Backports Update Testing and Nonfree Backports Updates Testing and installed the 7 packages listed in comment #2. NVIDIA 367.57 running fine. glmark2 score = 6775 Enabled SLI mode $ sudo nvidia-xconfig --sli=On Rebooted nvidia logo flashed up twice Desktop reappeared with the display connected to GPU 0 as before. GPU temperature hovering around 51°C. glmark2 score 6278. Probably no particular advantage in SLI for normal workstation operations. @tmb
Sorry to bother you again but I have not been able to install dkms-nvidia-current on i586 (installed on a 64bit machine) without getting an error relating to dkms.conf.
dkms.conf: Error! No 'DEST_MODULE_LOCATION' directive specified for record #2.
dkms.conf: Error! Directive 'DEST_MODULE_LOCATION' does not begin with
'/kernel', '/updates', or '/extra' in record #2.
# cat /etc/depmod.d/dkms.conf
search dkms dkms-binary built-in
# cat /usr/src/nvidia-current-367.57-1.mga5.nonfree/dkms.conf
PACKAGE_NAME="nvidia-current"
PACKAGE_VERSION="367.57-1.mga5.nonfree"
BUILT_MODULE_NAME[0]="nvidia"
DEST_MODULE_LOCATION[0]="/kernel/drivers/char/drm"
DEST_MODULE_NAME[0]="nvidia-current"
BUILT_MODULE_NAME[1]="nvidia-modeset"
DEST_MODULE_LOCATION[1]="/kernel/drivers/char/drm"
BUILT_MODULE_NAME[3]="nvidia-drm"
DEST_MODULE_LOCATION[3]="/kernel/drivers/char/drm"
MAKE[0]="'make' -j${parallel_jobs} SYSSRC=${kernel_source_dir} modules"
AUTOINSTALL="yes"
Apropos of comment #14 Or maybe this is a nogo anyway on a GeForce 8700M GT, currently running nvidia 340.96. Nope, you found a valid bug in the dkms packagng for nvidia-current... Apparently not many have tested this new driver in cauldron on 32bit. But yes, on your system using nvidia340 you can only test that nvidia-current builds ok, it wont support that hw. Fair enough. Thanks Thomas. Er, who should report this bug? It is a bit over my head. No need for more reports... I'm currently fixing it :) And fixed nvidia-current rpms: SRPMS: nvidia-current-367.57-1.1.mga5.nonfree.src.rpm i586: dkms-nvidia-current-367.57-1.1.mga5.nonfree.i586.rpm nvidia-current-cuda-opencl-367.57-1.1.mga5.nonfree.i586.rpm nvidia-current-devel-367.57-1.1.mga5.nonfree.i586.rpm nvidia-current-doc-html-367.57-1.1.mga5.nonfree.i586.rpm x11-driver-video-nvidia-current-367.57-1.1.mga5.nonfree.i586.rpm x86_64: dkms-nvidia-current-367.57-1.1.mga5.nonfree.x86_64.rpm nvidia-current-cuda-opencl-367.57-1.1.mga5.nonfree.x86_64.rpm nvidia-current-devel-367.57-1.1.mga5.nonfree.x86_64.rpm nvidia-current-doc-html-367.57-1.1.mga5.nonfree.x86_64.rpm x11-driver-video-nvidia-current-367.57-1.1.mga5.nonfree.x86_64.rpm x86_64 CSM laptop Lenovo Ideapad Y500 nvidia GeForce GT 650M The backports and updates installed cleanly, kmod rebuilt and system rebooted to the Mate desktop. It appears to be fully functional but there was an odd hiccup. On login the panel did not appear although it was running. Decided to kill the running instance hoping to relaunch it but just killing it was enough. It reinstated itself OK. @tmb re comment #19 Do I need to run all those earlier tests again? It has been a long day. (In reply to Len Lawrence from comment #21) > @tmb re comment #19 > Do I need to run all those earlier tests again? It has been a long day. Just test it builds ok on 32bit with the rpms in comment 19. That's enough for this backport OK Thomas and thanks for your prompt responses. Back on it tomorrow. i586 on x86_64 real hardware kernel 4.4.30-server-2.mga5 nvidia GeForce 8700M GT - 340.96 Installed the updated packages without any problems. No difficulty with rebooting. Not giving this the OK for i586 just yet - need to give others an opportunity to run it on 32bit hardware. Installed the latest RPMs and repeated the test for the GTX970 machine with the 4K DP monitor. Clean install and build and all is well after reboot. Giving this the OK for 64bits. Thanks for all your help Thomas.
Len Lawrence
2016-11-15 09:18:35 CET
Whiteboard:
(none) =>
MGA5-64-OK
Dave Hodgins
2016-11-17 20:28:25 CET
Keywords:
(none) =>
validated_backport moved Status:
NEW =>
RESOLVED For those that tested this one... we now push newest long lived branch to updates to fix security issues, so please give it a spin: https://bugs.mageia.org/show_bug.cgi?id=19654 |