Bug 58791

Summary:	X86_X2APIC + MAXSMP: Oops, sluggish, no SMP, no 8042 keyboard, CPU overheating
Product:	Platform Specific/Hardware	Reporter:	Roman Žilka (roman.zilka)
Component:	x86-64	Assignee:	platform_x86_64 (platform_x86_64)
Status:	NEW ---
Severity:	normal	CC:	kernel
Priority:	P1
Hardware:	All
OS:	Linux
Kernel Version:	3.9.3	Subsystem:
Regression:	No	Bisected commit-id:
Attachments:	.config cpuinfo dmesg dmidecode kmemleak lspci -vvv System.map vmlinuz

Description Roman Žilka 2013-05-24 17:03:23 UTC

Experimenting with 3.9.3 yielded a find: X86_X2APIC + MAXSMP = serious CPU issues. (Neither of them causes trouble when the other one is missing.)

* sluggish kernel boot
* a couple times during boot: "BUG: unable to handle kernel" / Oops / kernel panic, NULL pointer dereference, "CPU.*Stuck"
* ~twice slower HDD I/O
* very high CPU temperature despite no CPU load (reported by physical fan speed and /usr/bin/acpi)
* only 1 CPU available in the booted system (should be 2 + 2/HT)
* "ACPI Warning.*SystemIO conflicts with Region \GPIO 1"
* driver iwlwifi cannot load firmware (could be due to the prolonged boot, but it wasn't *that* long)
* "i8042: Can't write CTR while closing KBD port", "i8042: Can't reactivate KBD port"
* when running, every now and then the kernel pauses for a fraction of a second

This is the first time (vanilla 3.9.3) I tried turning on MAXSMP at all. I suppose testing the combination on an older version is in order. Please, suggest which specific version(s) are interesting in this respect.

Comment 1 Roman Žilka 2013-05-24 17:04:40 UTC

Created attachment 102461 [details]
.config

Comment 2 Roman Žilka 2013-05-24 17:05:00 UTC

Created attachment 102471 [details]
cpuinfo

Comment 3 Roman Žilka 2013-05-24 17:05:31 UTC

Created attachment 102481 [details]
dmesg

Comment 4 Roman Žilka 2013-05-24 17:05:51 UTC

Created attachment 102491 [details]
dmidecode

Comment 5 Roman Žilka 2013-05-24 17:06:14 UTC

Created attachment 102501 [details]
kmemleak

Comment 6 Roman Žilka 2013-05-24 17:06:35 UTC

Created attachment 102511 [details]
lspci -vvv

Comment 7 Roman Žilka 2013-05-24 17:07:20 UTC

Created attachment 102521 [details]
System.map

Comment 8 Roman Žilka 2013-05-24 17:13:10 UTC

Created attachment 102531 [details]
vmlinuz

SHA256: 647fb1d91aa39900e269a08e46e76f00eba3fecf9c40f5f045f6b30546b97ede

Comment 9 Roman Žilka 2013-05-24 17:33:47 UTC

Apologies, the "ACPI Warning.*SystemIO conflicts with Region \GPIO 1" problem is irrelevant to this bug's topic. It's there even without X2APIC and MAXSMP.

I suppose the kmemleak contents have nothing to do with this bug as well (just turned on kmemleak for the sake of this bug). If so, please, let me know - I have a graphics-related ugliness to pursue as well (again, not based on X2APIC/MAXSMP).

Comment 10 Tom Wijsman 2013-05-24 18:02:23 UTC

Downstream bug is at https://bugs.gentoo.org/show_bug.cgi?id=470792