Re: [PATCH v6 0/2] sched/deadline: Fix and optimize sched_dl_global_validate()

[Juri Lelli]

Hi,

On 08/10/20 23:47, Peng Liu wrote:
> When change global rt bandwidth, we check to make sure that new
> settings could accommodate the allocated dl bandwidth.
> 
> Under SMP, the dl_bw is on a per root domain basis, currently we check
> and update the new settings one cpu by one cpu, but not in the unit of
> root domain, which is either overdoing or error.
> 
> patch 1 removed the superfluous checking and updating
> patch 2 fixed the error validation
> 
> For details, please see the corresponding patch.
> 
> ----------------
> v6 <-- v5:
>  - no functional changes, just revert visit_gen back to u64;
> 
> v5 <-- v4:
>  - no functional changes, just split the v4 single patch to two to
>    obey the "one patch do only one thing" rule;
>  - turn root_domain::visit_gen from u64 to u32;
>    both suggested by Juri.
>  - refine changelog;
> 
> v4 <-- v3:
>  - refine changelog;
>  - eliminate the ugly #ifdef guys with Peter's method;
> 
> v3 <-- v2:
>  - fix build error for !CONFIG_SMP, reported by kernel test robot;
> 
> v2 <-- v1:
>  - replace cpumask_weight(cpu_rq(cpu)->rd->span) with dl_bw_cpus(cpu),
>    suggested by Juri;
> 
> Peng Liu (2):
>   sched/deadline: Optimize sched_dl_global_validate()
>   sched/deadline: Fix sched_dl_global_validate()
> 
>  kernel/sched/deadline.c | 44 +++++++++++++++++++++++++++--------
>  kernel/sched/sched.h    | 51 ++++++++++++++++++++++-------------------
>  kernel/sched/topology.c |  1 +
>  3 files changed, 63 insertions(+), 33 deletions(-)

These look now good to me. Thanks a lot!

Acked-by: Juri Lelli <juri.lelli@xxxxxxxxxx>

Best,
Juri