Bug 150311
| Summary: | Unnaccounted for high cpu usage while idle | ||
|---|---|---|---|
| Product: | Process Management | Reporter: | Michael Dale Long (link) |
| Component: | Other | Assignee: | process_other |
| Status: | NEW --- | ||
| Severity: | normal | CC: | blshadow |
| Priority: | P1 | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Kernel Version: | 4.6.0-rc1 and later | Subsystem: | |
| Regression: | No | Bisected commit-id: | |
| Attachments: | Kernel config used to build 4.6.0-rc1 | ||
I spent some time bisecting from v4.5.7 to v4.6-rc1, and it appears that the problem is introduced in this commit: f872f5400cc01373d8e29d9c7a5296ccfaf4ccf3 At first glance I have no idea why this commit would be the cause of this behavior, but it is clearly observable in a kernel built from that version and not in one built prior to it. I have the same problem In my case, the cause of the problems is enabled by mistake CONFIG_CONTEXT_TRACKING_FORCE option. Apparently I had the same problem. Disabling that in 4.7.1 fixed the issue. Thanks for the information! |
Created attachment 226441 [details] Kernel config used to build 4.6.0-rc1 Starting with 4.6.0-rc1, my system began having unexplained CPU usage. Even when the system is completely idle, the first core would show around 50% CPU. htop did not show any process associated with the activity, kernel or user. I run a full tickless kernel, which I thought might have been related, but even going back to a normal kernel tick just spreads the mystery usage across different cores, it doesn't go away. I ran atop to see if short lived processes were at fault, but it didn't detect any. I confirmed that this does not occur with 4.5.7, so it was introduced somewhere between 4.5.x and 4.6.0-rc1. It does still occur on 4.7.0. All this was tested on an untainted kernel built from linux-stable.