sched/topology, cpuset: Avoid spurious/wrong domain rebuilds
authorPeter Zijlstra <peterz@infradead.org>
Tue, 8 Aug 2017 10:16:24 +0000 (12:16 +0200)
committerIngo Molnar <mingo@kernel.org>
Fri, 25 Aug 2017 09:12:20 +0000 (11:12 +0200)
commit77d1dfda0e79b41894880418f04794e92e4350e2
treeb5623653d670a4933ceb5e8eab97560205a4aa0a
parenta090c4f2cd2cfeb54eaf8ad1e726a6c485f3abc3
sched/topology, cpuset: Avoid spurious/wrong domain rebuilds

When disabling cpuset.sched_load_balance we expect to be able to online
CPUs without generating sched_domains. However this is currently
completely broken.

What happens is that we generate the sched_domains and then destroy
them. This is because of the spurious 'default' domain build in
cpuset_update_active_cpus(). That builds a single machine wide domain
and then schedules a work to build the 'real' domains. The work then
finds there are _no_ domains and destroys the lot again.

Furthermore, if there actually were cpusets, building the machine wide
domain is actively wrong, because it would allow tasks to 'escape' their
cpuset. Also I don't think its needed, the scheduler really should
respect the active mask.

Reported-by: Ofer Levi(SW) <oferle@mellanox.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vineet.Gupta1@synopsys.com <Vineet.Gupta1@synopsys.com>
Cc: rusty@rustcorp.com.au <rusty@rustcorp.com.au>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
kernel/cgroup/cpuset.c