On Thu, Nov 14, 2024 at 04:14:00PM +0000 Juri Lelli wrote: > Thanks Waiman and Phil for the super quick review/test of this v2! > > On 14/11/24 14:28, Juri Lelli wrote: > > ... > > > In all honesty, I still see intermittent issues that seems to however be > > related to the dance we do in sched_cpu_deactivate(), where we first > > turn everything related to a cpu/rq off and revert that if > > cpuset_cpu_inactive() reveals failing DEADLINE checks. But, since these > > seem to be orthogonal to the original discussion we started from, I > > wanted to send this out as an hopefully meaningful update/improvement > > since yesterday. Will continue looking into this. > > About this that I mentioned, it looks like the below cures it (and > hopefully doesn't regress wrt the other 2 patches). > > What do everybody think? > I think that makes sense. I think it's better not to have that deadline call buried the cpuset code as well. Reviewed-by: Phil Auld <pauld@xxxxxxxxxx> > --- > Subject: [PATCH] sched/deadline: Check bandwidth overflow earlier for hotplug > > Currently we check for bandwidth overflow potentially due to hotplug > operations at the end of sched_cpu_deactivate(), after the cpu going > offline has already been removed from scheduling, active_mask, etc. > This can create issues for DEADLINE tasks, as there is a substantial > race window between the start of sched_cpu_deactivate() and the moment > we possibly decide to roll-back the operation if dl_bw_deactivate() > returns failure in cpuset_cpu_inactive(). An example is a throttled > task that sees its replenishment timer firing while the cpu it was > previously running on is considered offline, but before > dl_bw_deactivate() had a chance to say no and roll-back happened. > > Fix this by directly calling dl_bw_deactivate() first thing in > sched_cpu_deactivate() and do the required calculation in the former > function considering the cpu passed as an argument as offline already. > > Signed-off-by: Juri Lelli <juri.lelli@xxxxxxxxxx> > --- > kernel/sched/core.c | 9 +++++---- > kernel/sched/deadline.c | 12 ++++++++++-- > 2 files changed, 15 insertions(+), 6 deletions(-) > > diff --git a/kernel/sched/core.c b/kernel/sched/core.c > index d1049e784510..43dfb3968eb8 100644 > --- a/kernel/sched/core.c > +++ b/kernel/sched/core.c > @@ -8057,10 +8057,6 @@ static void cpuset_cpu_active(void) > static int cpuset_cpu_inactive(unsigned int cpu) > { > if (!cpuhp_tasks_frozen) { > - int ret = dl_bw_deactivate(cpu); > - > - if (ret) > - return ret; > cpuset_update_active_cpus(); > } else { > num_cpus_frozen++; > @@ -8128,6 +8124,11 @@ int sched_cpu_deactivate(unsigned int cpu) > struct rq *rq = cpu_rq(cpu); > int ret; > > + ret = dl_bw_deactivate(cpu); > + > + if (ret) > + return ret; > + > /* > * Remove CPU from nohz.idle_cpus_mask to prevent participating in > * load balancing when not active > diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c > index 267ea8bacaf6..6e988d4cd787 100644 > --- a/kernel/sched/deadline.c > +++ b/kernel/sched/deadline.c > @@ -3505,6 +3505,13 @@ static int dl_bw_manage(enum dl_bw_request req, int cpu, u64 dl_bw) > } > break; > case dl_bw_req_deactivate: > + /* > + * cpu is not off yet, but we need to do the math by > + * considering it off already (i.e., what would happen if we > + * turn cpu off?). > + */ > + cap -= arch_scale_cpu_capacity(cpu); > + > /* > * cpu is going offline and NORMAL tasks will be moved away > * from it. We can thus discount dl_server bandwidth > @@ -3522,9 +3529,10 @@ static int dl_bw_manage(enum dl_bw_request req, int cpu, u64 dl_bw) > if (dl_b->total_bw - fair_server_bw > 0) { > /* > * Leaving at least one CPU for DEADLINE tasks seems a > - * wise thing to do. > + * wise thing to do. As said above, cpu is not offline > + * yet, so account for that. > */ > - if (dl_bw_cpus(cpu)) > + if (dl_bw_cpus(cpu) - 1) > overflow = __dl_overflow(dl_b, cap, fair_server_bw, 0); > else > overflow = 1; > --