On Wed, Jun 15, 2016 at 01:32:49PM +0200, Peter Zijlstra wrote: > On Wed, Jun 15, 2016 at 03:49:36PM +0530, Gautham R Shenoy wrote: > > > Also, with the first patch in the series (which ensures that > > restore_unbound_workers are called *after* the new workers for the > > newly onlined CPUs are created) and without this one, you can > > reproduce this WARN_ON on both x86 and PPC by offlining all the CPUs > > of a node and bringing just one of them online. > > Ah good. > > > I am not sure about that. The workqueue creates unbound workers for a > > node via wq_update_unbound_numa() whenever the first CPU of every node > > comes online. So that seems legitimate. It then tries to affine these > > workers to the cpumask of that node. Again this seems right. As an > > optimization, it does this only when the first CPU of the node comes > > online. Since this online CPU is not yet active, and since > > nr_cpus_allowed > 1, we will hit the WARN_ON(). > > So I had another look and isn't the below a much simpler solution? > > It seems to work on my x86 with: > > for i in /sys/devices/system/cpu/cpu*/online ; do echo 0 > $i ; done > for i in /sys/devices/system/cpu/cpu*/online ; do echo 1 > $i ; done > > without complaint.
Yup. This will work on PPC as well. We will no longer have the optimization in restore_unbound_workers_cpumask() but I suppose we don't lose much by resetting the affinity every time a CPU in the pool->attr->cpumask comes online. -- Thanks and Regards gautham.
