Messages in this thread | | | Date | Tue, 22 Nov 2022 13:55:53 +0800 | From | Aaron Lu <> | Subject | Re: [PATCH v3] sched: async unthrottling for cfs bandwidth |
| |
On Mon, Nov 21, 2022 at 11:31:20AM -0800, Josh Don wrote: > > Please test the final version as found here: > > > > https://git.kernel.org/pub/scm/linux/kernel/git/peterz/queue.git/commit/?h=sched/core&id=4e3c1b7b489e218dfa576cd6af0680b975b8743e > > Thanks Peter, the patch looks good to me and continues to work > correctly. Note that I needed to make two edits to get it to build > though on sched/core: > > - local_cfq_rq doesn't match variable name used in function (local_unthrottle) > - rq_lock_irqrestore -> rq_unlock_irqrestore
I also tested it on a desktop bare metal with 800 cgroups under 2 cgroups that have quota set. Before this patch, the majority of durations for sched_cfs_period_timer() are in the range of 512us-1ms, with outliers in the range of 1ms-4ms; after this patch, the majority are in the range of 128us-512us with no outliers above 512us.
| |