On Sat, 27 Feb 2016, Tejun Heo wrote:

On Sat, Feb 27, 2016 at 12:37:23PM +0100, Thomas Gleixner wrote:

quoted

On Sat, 27 Feb 2016, Tejun Heo wrote:

quoted

On Fri, Feb 26, 2016 at 01:47:39PM -0800, Yang Shi wrote:

quoted

Call synchronize_sched() when kernfs_node is updated since tracepoints are
protected by rcu_read_lock_sched.

Adding synchronize_sched() to operations which can be triggered from
userland usually turns out to be problematic.  If this can't be solved
any other way, I think the right thing to do is replacing the path
with cgroup id as Christoph suggested.

Agreed. The path is not that important, right?

It can be, but we can print out the ino and userland can match that up
with path if necessary.

Wouldn't be cgroup id the better choice?

Thanks,

	tglx