[ofa-general] Re: mlx4 violating radix tree API locking rules?

[Roland Dreier]

 > Why not just call synchronize_rcu instead?

Not sure I understand.  Where would you put the synchronize_rcu and
what would it protect against?  RCU is being used to protect the radix
tree internals, not the mlx4 data structures.

 > > I guess CQ spinlock implies rcu_read_lock - is that right?
 > > But I do not see any synchronize_rcu calls anywhere in mlx4.
 > > Should destroy QP and friends call synchronize_rcu after
 > > removing the QP from radix tree but before freeing the QP structure?

By the way, replying to this earlier bit: I don't think the CQ
spinlock is equivalent to an rcu_read_lock().  In most configurations
it may be but I suspect the assumption would be broken by PREEMPT_RT
or the like.

 - R.