On Mon, Apr 08, 2019 at 11:39:36AM -0400 Waiman Long wrote: > On 04/08/2019 11:14 AM, Tejun Heo wrote: > > Hello, > > > > (cc'ing Waiman and copying the whole message for him) > > > > On Fri, Apr 05, 2019 at 11:36:59AM -0400, Joel Savitz wrote: > >> If a process is limited by taskset (i.e. cpuset) to only be allowed to > >> run on cpu N, and then cpu N is offlined via hotplug, the process will > >> be assigned the current value of its cpuset cgroup's effective_cpus field > >> in a call to do_set_cpus_allowed() in cpuset_cpus_allowed_fallback(). > >> This argument's value does not makes sense for this case, because > >> task_cs(tsk)->effective_cpus is modified by cpuset_hotplug_workfn() > >> to reflect the new value of cpu_active_mask after cpu N is removed from > >> the mask. While this may make sense for the cgroup affinity mask, it > >> does not make sense on a per-task basis, as a task that was previously > >> limited to only be run on cpu N will be limited to every cpu _except_ for > >> cpu N after it is offlined/onlined via hotplug. > >> > >> Pre-patch behavior: > >> > >> $ grep Cpus /proc/$$/status > >> Cpus_allowed: ff > >> Cpus_allowed_list: 0-7 > >> > >> $ taskset -p 4 $$ > >> pid 19202's current affinity mask: f > >> pid 19202's new affinity mask: 4 > >> > >> $ grep Cpus /proc/self/status > >> Cpus_allowed: 04 > >> Cpus_allowed_list: 2 > >> > >> # echo off > /sys/devices/system/cpu/cpu2/online > >> $ grep Cpus /proc/$$/status > >> Cpus_allowed: 0b > >> Cpus_allowed_list: 0-1,3 > >> > >> # echo on > /sys/devices/system/cpu/cpu2/online > >> $ grep Cpus /proc/$$/status > >> Cpus_allowed: 0b > >> Cpus_allowed_list: 0-1,3 > >> > >> On a patched system, the final grep produces the following > >> output instead: > >> > >> $ grep Cpus /proc/$$/status > >> Cpus_allowed: ff > >> Cpus_allowed_list: 0-7 > >> > >> This patch changes the above behavior by instead simply resetting the mask > >> to cpu_possible_mask. > >> > >> Signed-off-by: Joel Savitz <jsavitz@xxxxxxxxxx> > >> --- > >> kernel/cgroup/cpuset.c | 2 +- > >> 1 file changed, 1 insertion(+), 1 deletion(-) > >> > >> diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c > >> index 479743db6c37..5f65a2167bdf 100644 > >> --- a/kernel/cgroup/cpuset.c > >> +++ b/kernel/cgroup/cpuset.c > >> @@ -3243,7 +3243,7 @@ void cpuset_cpus_allowed(struct task_struct *tsk, struct cpumask *pmask) > >> void cpuset_cpus_allowed_fallback(struct task_struct *tsk) > >> { > >> rcu_read_lock(); > >> - do_set_cpus_allowed(tsk, task_cs(tsk)->effective_cpus); > >> + do_set_cpus_allowed(tsk, cpu_possible_mask); > >> rcu_read_unlock(); > > cpuset directly mangling with per-task masks has always been weird and > > somewhat broken. Given the current cpuset behavior, I suppose this is > > the better behavior. Waiman, what do you think? > > > > Thanks. > > > I think it may be better to use cpus_allowed in the case of fallback to > make sure that the task isn't allowed to run on CPUs it is not supposed > to run on, e.g. in a VM or container under cpuset control. For tasks in > the top cpuset, it is the same as cpu_possible_mask. Of course, we are > assuming that cpus_allowed has some sane value. BTW, there should be > some comments about handling this case of cpu offlining. > This is setting cpus_allowed, so we can't use that here. This is the final fallback. We've already tried parent cpuset bits at this point and found nothing. If the parent had a mask that included a CPU that was still present we would have already used that. I believe Joel's testing included using a cpuset hierarchy and it did the right thing. I don't know if he has those notes still or not. Cheers, Phil > -Longman > > > --