Fedora Account System
Red Hat Associate
Red Hat Customer
The akmods command (akmods-0.6.2-9) doesn't log properly enough: - if a build or install of a kmod fails, akmods output to stdout a [ OK ] at the end of the line Checking kmods exist for KERNEL this is misleading - it doesn't indicate which kmod are checked - there is no indication of the context in which akmods is called that makes /var/log/akmods/akmods.log of little use for debugging I suggest thus to change the akmods command as follows (example with the kmods: v4l2loopback VirtualBox wl): from: Checking kmods exist for 6.19.7-200.fc43.x86_64 [ OK ] Checking kmods exist for 6.18.16-200.fc43.x86_64 [ OK ] to: Start checking kmods interactively: ALL for kernels: 6.19.7-200.fc43.x86_64 6.18.16-200.fc43.x86_64 with akmods-VERSION-RELEASE Checking kmods exist for 6.19.7-200.fc43.x86_64 kmod-v4l2loopback-6.19.7-200.fc43.x86_64 0.15.3-1.fc43 I [ OK ] kmod-VirtualBox-6.19.7-200.fc43.x86_64 7.2.6-2.fc43 I [ OK ] kmod-wl-6.19.7-200.fc43.x86_64 6.30.223.271-61.fc43 I [ OK ] Checking kmods exist for 6.18.16-200.fc43.x86_64 kmod-v4l2loopback-6.18.16-200.fc43.x86_64 0.15.3-1.fc43 I [ OK ] kmod-VirtualBox-6.18.16-200.fc43.x86_64 7.2.6-2.fc43 I [ OK ] kmod-wl-6.18.16-200.fc43.x86_64 6.30.223.271-61.fc43 I [ OK ] End akmods: exit code: 0, elapsed time: 1 seconds, locked for 0 seconds If called from akmods.service one would have: Start checking kmods from init ... Similar with from kernel-posttrans ... Also, add a blank line in /var/log/akmods/akmods.log after each call to akmods. This will permit to get the traces in error with a simple: awk -v RS= -v ORS='\n\n' '/exit code: [^0]/' /var/log/akmods/akmods.log The I before the [ OK ] means Installed. In case of failure we will output: missing U (for Uninstalled) In addition akmods-0.6.2-9 always exit 0. Change this to exit 1 if at least one kmod fails to build or install This will make the akmods.service fail if akmods fails, but this is in fact also better. Finally, for debugging purpose add a -n,--dry-run option like make does. Note: This has been initially done in: https://src.fedoraproject.org/rpms/akmods/pull-request/32 https://bugzilla.redhat.com/show_bug.cgi?id=2440497 Reproducible: Always
pull-request done: https://src.fedoraproject.org/rpms/akmods/pull-request/33 Successful test with 2 kmods, that gives as /var/log/akmods/akmods.log: (note the locked duration) 2026/03/18 10:58:40 akmods: Start checking kmods from akmod-posttrans: wl 2026/03/18 10:58:40 akmods: for kernels: 6.19.7-200.fc43.x86_64 6.18.16-200.fc43.x86_64 2026/03/18 10:58:40 akmods: with akmods-0.6.2-9.fc43 2026/03/18 10:58:40 akmods: Checking kmods exist for 6.19.7-200.fc43.x86_64 2026/03/18 10:58:40 akmods: Building and installing wl-kmod 2026/03/18 10:58:40 akmods: Building rpms using the command '/usr/sbin/akmodsbuild --kernels 6.19.7-200.fc43.x86_64 /usr/src/akmods/wl-kmod.latest' 2026/03/18 10:58:45 akmods: Installing newly built rpms 2026/03/18 10:58:45 akmods: DNF detected 2026/03/18 10:58:53 akmods: Successful. 2026/03/18 10:58:53 akmods: kmod-wl-6.19.7-200.fc43.x86_64 6.30.223.271-61.fc43 I 2026/03/18 10:58:53 akmods: Checking kmods exist for 6.18.16-200.fc43.x86_64 2026/03/18 10:58:53 akmods: Building and installing wl-kmod 2026/03/18 10:58:53 akmods: Building rpms using the command '/usr/sbin/akmodsbuild --kernels 6.18.16-200.fc43.x86_64 /usr/src/akmods/wl-kmod.latest' 2026/03/18 10:58:59 akmods: Installing newly built rpms 2026/03/18 10:58:59 akmods: DNF detected 2026/03/18 10:59:06 akmods: Successful. 2026/03/18 10:59:06 akmods: kmod-wl-6.18.16-200.fc43.x86_64 6.30.223.271-61.fc43 I 2026/03/18 10:59:06 akmods: Restarting systemd-modules-load.service 2026/03/18 10:59:07 akmods: End akmods: exit code: 0, elapsed time: 27 seconds, locked for 0 seconds 2026/03/18 10:59:07 akmods: Start checking kmods from akmod-posttrans: v4l2loopback 2026/03/18 10:59:07 akmods: for kernels: 6.19.7-200.fc43.x86_64 6.18.16-200.fc43.x86_64 2026/03/18 10:59:07 akmods: with akmods-0.6.2-9.fc43 2026/03/18 10:59:07 akmods: Checking kmods exist for 6.19.7-200.fc43.x86_64 2026/03/18 10:59:07 akmods: Building and installing v4l2loopback-kmod 2026/03/18 10:59:07 akmods: Building rpms using the command '/usr/sbin/akmodsbuild --kernels 6.19.7-200.fc43.x86_64 /usr/src/akmods/v4l2loopback-kmod.latest' 2026/03/18 10:59:10 akmods: Installing newly built rpms 2026/03/18 10:59:10 akmods: DNF detected 2026/03/18 10:59:17 akmods: Successful. 2026/03/18 10:59:17 akmods: kmod-v4l2loopback-6.19.7-200.fc43.x86_64 0.15.3-1.fc43 I 2026/03/18 10:59:17 akmods: Checking kmods exist for 6.18.16-200.fc43.x86_64 2026/03/18 10:59:17 akmods: Building and installing v4l2loopback-kmod 2026/03/18 10:59:17 akmods: Building rpms using the command '/usr/sbin/akmodsbuild --kernels 6.18.16-200.fc43.x86_64 /usr/src/akmods/v4l2loopback-kmod.latest' 2026/03/18 10:59:20 akmods: Installing newly built rpms 2026/03/18 10:59:20 akmods: DNF detected 2026/03/18 10:59:27 akmods: Successful. 2026/03/18 10:59:27 akmods: kmod-v4l2loopback-6.18.16-200.fc43.x86_64 0.15.3-1.fc43 I 2026/03/18 10:59:27 akmods: Restarting systemd-modules-load.service 2026/03/18 10:59:28 akmods: End akmods: exit code: 0, elapsed time: 21 seconds, locked for 27 seconds The diff (of only the akmods script) is rather big, but I don't want to split this. I forgot to mention that I also changed the saved kmodlogfile, adding the date before .log (or .failed.log): do not loose any log. Thoughts ?
The dry-run option permits to see trivially and *without* doing any action, the bug when giving more than one --kmod: OK: kmod-wl-6.18.16-200.fc43.x86_64 has been checked akmods -n --kmod wl --kernel $(uname -r) Start checking kmods interactively (dry-run): wl for kernels: 6.18.16-200.fc43.x86_64 with akmods-0.6.2-9.fc43 Checking kmods exist for 6.18.16-200.fc43.x86_64 kmod-wl-6.18.16-200.fc43.x86_64 6.30.223.271-61.fc43 I [ OK ] End akmods: exit code: 0, elapsed time: 0 seconds, locked for 0 seconds root@fmpc KO: no kmods was checked: akmods -n --kmod wl --kmod v4l2loopback --kernel $(uname -r) Start checking kmods interactively (dry-run): wl v4l2loopback for kernels: 6.18.16-200.fc43.x86_64 with akmods-0.6.2-9.fc43 Checking kmods exist for 6.18.16-200.fc43.x86_64 End akmods: exit code: 0, elapsed time: 0 seconds, locked for 0 seconds root@fmpc I did fixed that bug in the pre-install-not-as-rpm branch, but will not in this one (improve-logs).
@Nicolas Chauvet: are you okay to merge this pull-request ? https://src.fedoraproject.org/rpms/akmods/pull-request/33
I don't see any build for this pull-request. Have I missed something ? Thanks. While we are at those cosmetic changes, I would like to add, just after all the functions are defined (just before parsing the arguments): [[ $0 == *bash ]] && return This permits to debug the functions with: ( source ./akmods || exit; FUNC ARGS ) I used that for more than 20 years in many bash system scripts I wrote (when working at Inria). Do you agree ? Thanks.
(In reply to Francis Montagnac from comment #4) > I don't see any build for this pull-request. Have I missed something ? Fixed simply by doing a new commit in this branch ... weird. Changelog: * Fri Mar 20 2026 Francis Montagnac <francis.montagnac> - 0.6.2-30 - akmods: Allow debugging functions with: ( source ./akmods || exit; FUNC ARGS ) Builds: https://kojipkgs.fedoraproject.org//work/tasks/4314/143534314/akmods-0.6.2-30.fc45.noarch.rpm https://kojipkgs.fedoraproject.org//work/tasks/4314/143534314/akmods-0.6.2-30.fc45.src.rpm I installed akmods-0.6.2-30.fc45.noarch.rpm on F43: it works ... of course.
2 more small commits leading to: https://kojipkgs.fedoraproject.org//work/tasks/7192/143567192/akmods-0.6.2-32.fc45.noarch.rpm I don't see other changes to do in the scope of this pull-request. Comments ? A next pull-request. may be to wait for the end of the dnf transaction (as dnf4 did). This is already implemented in the pre-install-not-as-rpm branch.
@Nicolas Chauvet: when do you think to have the time to merge this pull-request ? https://src.fedoraproject.org/rpms/akmods/pull-request/33 (no commits since past Saturday)
Any updates for this pull-request ? https://src.fedoraproject.org/rpms/akmods/pull-request/33 Thanks.
May I have an answer about this PR please ? FWIW This PR would have helped in this rfbz: https://bugzilla.rpmfusion.org/show_bug.cgi?id=7454 The reporter had lost the initial failed.log: > cannot attach the fail log since I lost it after getting a typo bug during manual patching which completely overrode it. Ref: commit 1fa0f8602f20d279f64398a27e8a83d2ce01783a Author: Francis Montagnac <francis.montagnac> Date: Mon Mar 2 07:37:58 2026 +0100 akmods: buildinstall_kmod: keep all logs with a date suffix before [.failed].log
FWIW This PR would have helped also in https://discussion.fedoraproject.org/t/problems-with-upgrade-to-f44-and-nvidia-gtx-970m/191189/41
No comments for this RFE for 3 months. Should I assume that this is a NO-GO ?