Opened 5 weeks ago

Last modified 8 days ago

#18864 new bug

KDL when shutting down

Reported by: johan Owned by: nobody
Priority: normal Milestone: Unscheduled
Component: System/Kernel Version: R1/Development
Keywords: Cc: korli
Blocked By: Blocking:
Platform: x86-64

Description

Shutting down Haiku always KDLs. Reboot does not cause any problems.

I'm running Haiku (currently hrev57668) natively on a ThinkPad X1 Yoga Gen 7 (i7-1260P @ 2.50GHz, 16GiB RAM). It's a dualboot setup using rEFInd to boot Haiku and Win 11.

This has been happening since I installed Haiku last year, not a new issue.

I tried the suggestions from #18356 (kill wpa_suplicant before shutdown, and co to continue), but the result is the same.

Attachments (2)

kdl-shutdown.jpg (791.3 KB ) - added by johan 5 weeks ago.
Photo of KDL
syslog.old (512.1 KB ) - added by johan 2 weeks ago.

Download all attachments as: .zip

Change History (8)

by johan, 5 weeks ago

Attachment: kdl-shutdown.jpg added

Photo of KDL

comment:1 by waddlesplash, 5 weeks ago

This doesn't appear to be related to #18356.

rax, rdx, and rdi appear to have garbage data in them: lots of 0x90s. That's probably the issue here.

Can you try booting with the "intel_cstates" module blacklisted and see if this changes anything?

comment:2 by waddlesplash, 5 weeks ago

Component: - GeneralSystem/Kernel

comment:3 by johan, 5 weeks ago

I tried to blacklist the module, but I still got the same problem. I added the following text to "/boot/system/settings/packages":

Package haiku {
    EntryBlacklist {
	add-ons/kernel/power/cpuidle/intel_cstates
    }
}

comment:4 by waddlesplash, 2 weeks ago

Please attach a full syslog.

by johan, 2 weeks ago

Attachment: syslog.old added

comment:5 by waddlesplash, 8 days ago

Cc: korli added

It seems strange we are setting MTRRs so often, something looks wrong there.

Presumably we are getting here via an "all CPUs ICI" invocation of the set-mtrrs. Maybe something got stuck in the ICI layer and we are somehow processing the ICI multiple times, and thus the data is invalid later times?

I note this processor appears to have 4 performance cores and 8 efficiency cores. Perhaps that's somehow related to this? CC korli, he may have more ideas.

comment:6 by korli, 8 days ago

Could it about the same as #18778 ? What are other CPUs doing?

Note: See TracTickets for help on using tickets.