Bug 90014 - i915 triggering ACPI GPE storm
Summary: i915 triggering ACPI GPE storm
Status: CLOSED WORKSFORME
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/Intel (show other bugs)
Version: unspecified
Hardware: x86-64 (AMD64) Linux (All)
: medium major
Assignee: Intel GFX Bugs mailing list
QA Contact: Intel GFX Bugs mailing list
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2015-04-13 11:15 UTC by Gianguido Sora`
Modified: 2017-07-24 22:47 UTC (History)
2 users (show)

See Also:
i915 platform: HSW
i915 features: power/Other


Attachments
dmesg of running system (110.81 KB, text/plain)
2015-04-13 11:15 UTC, Gianguido Sora`
no flags Details
perf output (30.94 KB, text/plain)
2015-04-14 10:57 UTC, Gianguido Sora`
no flags Details

Description Gianguido Sora` 2015-04-13 11:15:38 UTC
Created attachment 115054 [details]
dmesg of running system

Hardware: Retina MacBook Pro 13", Intel Core i5 4258U with Intel Iris 5100 integrated graphics.

Distribution: Ubuntu 15.04 daily release
Kernel version: 3.19.0-13-generic

Output from glxgears -info:
GL_RENDERER   = Mesa DRI Intel(R) Haswell 
GL_VERSION    = 3.0 Mesa 10.5.2
GL_VENDOR     = Intel Open Source Technology Center

This bug affects all distributions using kernels after 3.14.

On my hardware, 1 core of 4 is stuck at about 70% load, even without any CPU-intensive operation running.

If the system is booted without loading i915 module the CPU load stays at normal levels.
Comment 1 Chris Wilson 2015-04-13 11:26:02 UTC
Use "sudo perf top" to see what is hogging the CPU.
Comment 2 Gianguido Sora` 2015-04-13 14:54:29 UTC
perf top reveals that:

acpi_ns_search_one_scope                                                         acpi_os_read_port                                                                acpi_os_write_port

are eating about 15% of CPU each.

If the system is loaded with "i915.modeset=0" CPU levels are normal and perf does not reveal any strange activity.
Comment 3 Chris Wilson 2015-04-13 17:51:34 UTC
Intriguing, let's see the callchain to see what it triggering the ACPI activity. So can you please 'perf record -g -a sleep 60; perf report | head -500' and attach the output.
Comment 4 Gianguido Sora` 2015-04-14 10:57:49 UTC
Created attachment 115069 [details]
perf output
Comment 5 Gianguido Sora` 2015-04-14 10:58:35 UTC
After some heavy googling, this[1] workaround seem to work.

I've attached the perf output anyway, this situation needs to be addressed.

Cheers.

[1]: http://askubuntu.com/questions/176565/why-does-kworker-cpu-usage-get-so-high
Comment 6 Chris Wilson 2015-04-14 15:17:38 UTC
So which GPE was it?
Comment 7 Gianguido Sora` 2015-04-14 15:25:47 UTC
GPE 66.
Comment 8 Jani Nikula 2016-01-18 13:14:05 UTC
Please try kernel v4.4.
Comment 9 Jani Nikula 2016-12-28 12:46:51 UTC
No response in months, closing. Please reopen if the problem persists with latest kernels.


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.