Created attachment 1129123 [details]
Logs in couple of instances where failure is observed. This log changes often resulting in no log at all
Description of problem:
Kexec fails on Fedora 23 ppc64le
We've reproduced a 1-1 Kexec failure on Fedora 23 ppc64le Bare Metal Install (Stock OS). We also tested it on live host OS and also observed that it fails too.
Here's all the stuff we tried:
Kexec From Kexec To Status
Fedora 23 Fedora 23 Failure
Fedora 23 Ubuntu 15.04 Failure
Fedora 23 Ubuntu 15.10 Failure
Kexec works on following ppc64le Stock Host OS without issues: Ubuntu 15.10, 15.04, CentOS 7.2
Kexec Fails on following ppc64le Stock Host OS: Fedora 23 , Debian 8.3
Logs: System hang happens with random logs (More often with no logs at all). I've attached couple of snippets of logs we were able to obtain.
Version-Release number of selected component (if applicable):
[root@ppc-builder ~]# uname -a
Linux ppc-builder 4.2.6-301.fc23.ppc64le #1 SMP Thu Nov 26 19:44:15 UTC 2015 ppc64le ppc64le ppc64le GNU/Linux
[root@ppc-builder ~]# cat /etc/os-release
VERSION="23 (Server Edition)"
PRETTY_NAME="Fedora 23 (Server Edition)"
Steps to Reproduce:
kexec -l vmlinuz —initrd=“initrd.img” —command-line=`cat /proc/cmdline`
Host Console hangs indefinitely, without providing ability to get any logs.
kexec -e results in a system hang (with different logs)
kexec -e should result in reloading of new kernel and initrd provided
I tested these two patches and it seemed to get kexec functional:
Here is a patch posted to kexec-tools upstream:
I think ubuntu works just because they are using old version of kexec-tools
Antony, can you retest with my patch in comment #2, It seems another problem in the patches in comment #1, but we may have multiple problems in the report though.
Patch in comment #2 works for me. But I will appreciate for your further testing because it is the only patch I need to make kexec work.
Antony, sorry, I find my test environment is not correct, I used a latest branch locally. In F23 patch in comment #2 is not necessary. Will see if I can test F23 on a ppc64 machine.
I reproduced the problem in F23, patches mentioned in comment #1 works, seems I can not reproduce the problem in my patch (comment #2) with Fedora kernel. But it can be easily reproduced with RHEL kernel in Fedora with latest upstream kexec-tools.
Test result like below for me:
Fedora kernel load Fedora kernel: need only the ABI v2 patches.
Fedora kernel load RHEL kernel: need only ABI v2 patches.
Fedora kernel load RHEL kernel with upstream kexec-tools: need both ABI v2 patches and the fs2dt fix in comment #2
So till now for Fedora 23 we should only need the ABI v2 patches, when we rebase kexec-tools in rawhide we may need both.
I'm still not sure why the different kernel version matters. Will discuss it in kexec list further.
Above is the upstream patches, but it need review from powerpc people.
Any of IBM people in cc list can help review them?
Cool, thanks a lot. Will post patch for Fedora 24 kexec-tools.
firstname.lastname@example.org, do you need it in Fedora 23?
Yes, if we could please. We're actively working with Fedora 23 and would like do some more testing with it. If it's too much trouble, I can always get a build in copr.
Thanks for everyone's help so far!
Ok, will fix it in f23 as well.
kexec-tools-2.0.10-10.fc23 has been submitted as an update to Fedora 23. https://bodhi.fedoraproject.org/updates/FEDORA-2016-4e3a1242aa
kexec-tools-2.0.10-10.fc23 has been pushed to the Fedora 23 testing repository. If problems still persist, please make note of it in this bug report.
See https://fedoraproject.org/wiki/QA:Updates_Testing for
instructions on how to install test updates.
You can provide feedback for this update here: https://bodhi.fedoraproject.org/updates/FEDORA-2016-4e3a1242aa
kexec-tools-2.0.10-10.fc23 has been pushed to the Fedora 23 stable repository. If problems still persist, please make note of it in this bug report.