dptf sends thermal data to kernel in endless loop, causing EC stability problems |
|||
Issue descriptionWhen running chromeos-4.4 on chell, the following log messages are seen continuously on the EC. [11830.392457 DPTF sensor 1, threshold 5 C, index 0, enabled] [11830.402453 DPTF sensor 1, threshold 55 C, index 1, enabled] [11830.412349 DPTF sensor 1, threshold 5 C, index 0, enabled] [11830.422389 DPTF sensor 1, threshold 55 C, index 1, enabled] [11830.432986 DPTF sensor 2, threshold 5 C, index 0, enabled] [11830.443107 DPTF sensor 2, threshold 55 C, index 1, enabled] [11830.453100 DPTF sensor 2, threshold 5 C, index 0, enabled] [11830.463067 DPTF sensor 2, threshold 55 C, index 1, enabled] [11830.473775 DPTF sensor 3, threshold 5 C, index 0, enabled] [11830.483764 DPTF sensor 3, threshold 51 C, index 1, enabled] [11830.493826 DPTF sensor 3, threshold 5 C, index 0, enabled] [11830.503855 DPTF sensor 3, threshold 51 C, index 1, enabled] [11830.514855 DPTF sensor 4, threshold 5 C, index 0, enabled] [11830.524858 DPTF sensor 4, threshold 55 C, index 1, enabled] [11830.535101 DPTF sensor 4, threshold 5 C, index 0, enabled] [11830.545410 DPTF sensor 4, threshold 55 C, index 1, enabled] Random EC and thus system reboots are seen as well. Debugging on Linux reveals that the thresholds are constantly updated from user space. The problem disappears after executing "stop dptf".
,
Aug 12 2016
Hi Duncan, sorry I didn't see this bug until today. I've been monitoring partner site and not this site. This issue should have been fixed by patch https://chromium-review.googlesource.com/#/c/361787/ which was merged a few weeks ago. The root cause of the issue was that DPTF was trying to set temperature thresholds for each thermal zone at /sys/class/thermal/thermal_zoneX/trip_point_[0|1]_temp. Prior to kernel 4.4 this field was read-only so the write would fail, and DPTF would then fall back to polling. Starting with kernel 4.4 these trip points are finally writable, but we saw immediate uevent generated by writing to these nodes, which causes DPTF to think that these threshold points have been tripped, and then it will re-set these thresholds, which then causes an infinite loop. The new DPTF 8.2.10602 version disables such uevent handling and uses polling as the default just as on 3.18 kernel, so we shouldn't see such issues on kernel 4.4 anymore.
,
Aug 12 2016
Excellent. I'll update my image and close the bug after confirming that it now works (and sorry for filing the bug in the wrong system).
,
Apr 7 2017
Can we close this?
,
Apr 8 2017
Long since fixed |
|||
►
Sign in to add a comment |
|||
Comment 1 by groeck@chromium.org
, Jun 16 2016Summary: dptf sends thermal data to kernel in endless loop, causing EC stability problems (was: dptf sends thermal data to kernel in endless loop)