Bug 207603
| Summary: | S2idle loops in s2idle_enter before actually entering s2idle | ||
|---|---|---|---|
| Product: | Power Management | Reporter: | Todd Brandt (todd.e.brandt) |
| Component: | Other | Assignee: | Rafael J. Wysocki (rjw) |
| Status: | RESOLVED PATCH_ALREADY_AVAILABLE | ||
| Severity: | normal | CC: | lenb |
| Priority: | P1 | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Kernel Version: | 5.7.0-rc2 | Subsystem: | |
| Regression: | No | Bisected commit-id: | |
| Bug Depends on: | |||
| Bug Blocks: | 178231 | ||
| Attachments: |
otcpl-dell-p3520-with-loop-stripped.html
dmesg.txt ftrace.txt ftrace_with_just_suspend_resume_trace_points.txt issue.def issue.def |
||
Created attachment 288963 [details]
dmesg.txt
Created attachment 288965 [details]
ftrace.txt
Created attachment 288967 [details]
ftrace_with_just_suspend_resume_trace_points.txt
This should be fixed by https://patchwork.kernel.org/patch/11538065/ if I'm not mistaken. Created attachment 289119 [details]
issue.def
Created attachment 289121 [details]
issue.def
|
Created attachment 288961 [details] otcpl-dell-p3520-with-loop-stripped.html When running stress testing, occasionally the s2idle_enter code will enter an infinite loop instead of waiting. The suspend_machine tracepoints are called hundreds of thousands of times during a 10 second period. It happens on many unrelated machines. It prevents timeline generation because the code html cannot render hundreds of thousands of suspend machine & resume machine phases. Dell p3520 Whiskey Lake U Dell XPS 13 9300 Amber Lake Y Attached are the logs for the Dell p3520 run and a timeline with the infinite loop trimmed out.