Bug 198033

Summary:	REGRESSION: "do_IRQ: 4.162 No irq handler for vector" during hibernate
Product:	Platform Specific/Hardware	Reporter:	Maarten Lankhorst (bugs)
Component:	x86-64	Assignee:	platform_x86_64 (platform_x86_64)
Status:	NEW ---
Severity:	normal	CC:	marta.lofstedt, mybigspam
Priority:	P1
Hardware:	All
OS:	Linux
Kernel Version:	v4.15-rc1	Subsystem:
Regression:	No	Bisected commit-id:
Attachments:	dmesg from hibernating

Description Maarten Lankhorst 2017-11-29 13:24:50 UTC

Created attachment 260923 [details]
dmesg from hibernating

Hey,

When testing v4.15-rc1 on intel's continuous integration systems, we noticed a regression where after hibernate no interrupts are delivered to the i915 driver any more.

After some false leads, I've bisected it down to

commit fdba46ffb4c203b6e6794163493fd310f98bb4be (HEAD, refs/bisect/bad)
Author: Thomas Gleixner <tglx@linutronix.de>
Date:   Wed Sep 13 23:29:27 2017 +0200

    x86/apic: Get rid of multi CPU affinity

Reverting the commit isn't easily done, because other commits in that tree appear to depend on it.

See also https://bugs.freedesktop.org/show_bug.cgi?id=103712

A revert of the entire pull request appears to make the problem go away, see results at:

https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_7333/shards.html

As of this writing, only sandy bridge and haswell results are there, but other results will be posted when they become available.

Comment 1 Marta Löfstedt 2017-11-29 13:28:35 UTC

fyi: 
https://bugs.freedesktop.org/show_bug.cgi?id=103712
only handle the issues for SNB.
There is more information from other machines is this bug:
https://bugs.freedesktop.org/show_bug.cgi?id=103951