On 24 November 2014 at 20:39, Rafael J. Wysocki rjw@rjwysocki.net wrote:
I hope that doesn't happen under rcu_read_lock()?
No. Should it?
The modification needs to be done under dev_opp_list_lock in the first place
Yeah, its there to protect against other updaters.
in which case you don't need the _rcu version of list walking, so you simply can use list_for_each_entry_safe() here. The mutex is sufficient for the synchronization with other writers (if any). The freeing, though, has to be
Correct.
deferred until all readers drop their references to the old entry. You can use kfree_rcu() for that.
Cool. I understood the concept atleast. And yes I followed the srcu mail from paul as well.. Will reply there.