Messages in this thread | | | Subject | Re: [PATCH] sched/cputime: Ensure correct utime and stime proportion | From | Xunlei Pang <> | Date | Wed, 4 Jul 2018 14:56:42 +0800 |
| |
On 7/2/18 11:21 PM, Tejun Heo wrote: > Hello, Peter. > > On Tue, Jun 26, 2018 at 05:49:08PM +0200, Peter Zijlstra wrote: >> Well, no, because the Changelog is incomprehensible and the patch >> doesn't really have useful comments, so I'll have to reverse engineer >> the entire thing, and I've just not had time for that. > > Just as an additional data point, we also sometimes see artifacts from > cpu_adjust_time() in the form of per-task user or sys time getting > stuck for some period (in extreme cases for over a minute) while the > application isn't doing anything differently. We're telling the users > that it's an inherent sampling artifact but it'd be nice to improve it > if possible without adding noticeable overhead. No idea whether this > patch's approach is a good one tho.
The patch has no noticeable overhead except the extra cputime fileds added into task_struct. We've been running this patch on our servers for months, looks good till now.
Thanks!
| |