On Tue, Jun 16, 2015 at 11:43:42AM -0400, Steven Rostedt wrote:
On Tue, 16 Jun 2015 07:16:26 -0700
"Paul E. McKenney" <paulmck@xxxxxxxxxxxxxxxxxx> wrote:
Just for the record: Using a thread for freeing the memory is curing the
problem without the need to modify rcu_is_watching.
I must confess to liking this approach better than guaranteeing full-up
reentrancy in call_rcu() and kfree_rcu(). ;-)
Then reentrancy must be really bad if you prefer a spinning thread that
polls constantly just to free an item ;-)
I was (perhaps naively) assuming that they would use a less aggressive
approach at some point. ;-)