Created attachment 1514197 [details]
4.19.8 boot kernelmsgs with crash
Description of problem:
Specific kernels in the 4.19 series crash during boot in the acpi component on specific hardware (see below).
Version-Release number of selected component (if applicable):
4.19.8-300 crashes as well as 4.19.7, not 4.19.3
Steps to Reproduce:
crash during boot
boot without crash
This is is an old HP/Compag 6715b laptop. lscpu and lspci output is attached. 4.18 kernels worked without problems. Problem appeared with 4.19.1, disappeared with 4.19.3 (and 5), reappeared with 4.19.7 and 4.19.8.
Note that the koj build from https://koji.fedoraproject.org/koji/taskinfo?taskID=31411873 ("atomacpi") does work nicely!
Created attachment 1514198 [details]
4.19.8 atomacpi boot kernelmsgs without crash
Created attachment 1514199 [details]
4.19.7 boot kernelmsgs with crash
Created attachment 1514200 [details]
4.19.3 boot kernelmsgs without crash
Created attachment 1514201 [details]
Created attachment 1514202 [details]
Thanks for filing this.
What does the v4.19.1/v4.19.2 crash look like? The patch I reverted in https://koji.fedoraproject.org/koji/taskinfo?taskID=31411873 arrived in v4.19.6 so it's probably an unrelated problem that got fixed, but it'd be good to see it.
Also, are you okay with me adding you to the Cc list when I email upstream about this?
Created attachment 1514466 [details]
4.19.2-300 boot kernelmsgs with crash
Created attachment 1514467 [details]
4.19.2-301 boot kernelmsgs with crash
Created attachment 1514468 [details]
4.19.6 boot kernelmsgs with crash
Created attachment 1514469 [details]
4.19.9 boot kernelmsgs with crash
Created attachment 1514470 [details]
4.19.4-300 boot kernelmsgs without crash
Created attachment 1514472 [details]
4.19.5 boot kernelmsgs without crash
BZ needs attachment multi-upload (or I need to find out how)...
Here are a few more data points for that same machine. (cc upstream is okay as confirmed by pm already)
Can confirm I have the exact same issue with a custom blade server from Online.net, with what seems to be a completely different hardware than OP.
This server is a production one so tests are limited for now.
I only have 3 kernels available:
4.19.8-200 : Crash not long after the call trace (acpi_ns_get_normalized_pathname)
4.19.10-200 : Same thing, but I got it to boot (kind of) successfully one time only. Ethernet device was not detected and server crash during the shutdown sequence few minutes after.
4.18.18-200 : Currently running kernel. Work just fine
Created attachment 1517144 [details]
Kernel logs + hardware info
This log include the 4.19.10-200 log (the one that completed the boot sequence for some reason, without the eth device and crash at shutdown)
and the 4.19.8-200 log, which didn't succeed
Created attachment 1517367 [details]
4.20.0 boot kernelmsgs with crash
This is with 4.20.0 from knurd's kernel-vanilla-stable. Things won't get better wih 4.20, it seems.
4.19.10 does work if I revert the same commit as jcline in his atomacpi build, see https://copr.fedorainfracloud.org/coprs/mjg/kernel-hp6715b/
I guess I should not pester all current Fedora kernel updates with negative karma because of this issue, should I?
Yes, this is an upstream issue so it needs to get addressed there. The patch in question, commit 4abb951b73ff ("ACPICA: AML interpreter: add region addresses in global list during initialization") was first introduced in upstream v4.19.2 (commit 22083c028d0b), reverted in v4.19.3 (commit 8ef305fbc50d), and finally included again in v4.19.6 (commit 87403e35bc56). I sent an email in response to its inclusion in v4.19.6, but it must have slipped through the cracks for the ACPI maintainers. I think the best thing to do would be to report it upstream: https://bugzilla.kernel.org/enter_bug.cgi?product=ACPI.
*********** MASS BUG UPDATE **************
We apologize for the inconvenience. There are a large number of bugs to go through and several of them have gone stale. Due to this, we are doing a mass bug update across all of the Fedora 29 kernel bugs.
Fedora 29 has now been rebased to 4.20.5-200.fc29. Please test this kernel update (or newer) and let us know if you issue has been resolved or if it is still present with the newer kernel.
If you experience different issues, please open a new bug report for those.
Created attachment 1525141 [details]
Created attachment 1525142 [details]
4.20.4 no crash with revert patch
Created attachment 1525143 [details]
Created attachment 1525145 [details]
4.20.5 no crash with revert patch
(In reply to Justin M. Forbes from comment #19)
> *********** MASS BUG UPDATE **************
> We apologize for the inconvenience. There are a large number of bugs to go
> through and several of them have gone stale. Due to this, we are doing a
> mass bug update across all of the Fedora 29 kernel bugs.
> Fedora 29 has now been rebased to 4.20.5-200.fc29. Please test this kernel
> update (or newer) and let us know if you issue has been resolved or if it is
> still present with the newer kernel.
> If you experience different issues, please open a new bug report for those.
There is no change for the better with 4.20.4 nor 4.20.5. Same crash, and the crash goes away with the same revert patch.
ACPI people did not even bother to answer GKH's question why they went back and forth with this patch, it seems. I doubt filing a bug an the kernel's bz would help :(
There is a new patch by Eric Schmauss who has been very patient with my rebuild/reboot/debug cycles: see the kernel bz at https://bugzilla.kernel.org/attachment.cgi?id=281267
Eric would be happy to see more testing by others who have experienced the same issue. His patch avoids the revert and fixes the problem on my machine. A Fedora 29 build (Fedora 4.20.5-200 kernel + Eric's patch) can be found here:
Let me know if you want to test on other archs and need a kernel build.
Hello, still got this bug with kernel 5.0.5-200.x86_64 on f29. Is a newer build available with Eric's patch ?
The patch is not yet in Fedora's kernel. It suppose it will be in 5.1.
I regularly build Fedora kernels with Eric's patch at: https://copr.fedorainfracloud.org/coprs/mjg/kernel-hp6715b/
Note that I build them usually as soon as they hit updates-testing, so you might pick up a kernel there before the regular Fedora kernel hits the updates repo.