Re: [2.6.32-rc5-git5] synchronize_sched() inside spin_lock()?

["Paul E. McKenney" <paulmck@linux.vnet.ibm.com>]

On Mon, Nov 02, 2009 at 03:33:46PM -0800, Dmitry Torokhov wrote:
> On Monday 02 November 2009 02:30:28 pm Paul E. McKenney wrote:
> > On Mon, Nov 02, 2009 at 09:00:06PM +0900, Tetsuo Handa wrote:
> > > Commit: 4ea7e38696c7e798c47ebbecadfd392f23f814f9
> > >
> > > tracepoint_synchronize_unregister() calls synchronize_sched(), but it is
> > > between spin_lock() and spin_unlock(). Is it OK?
> > 
> > Calling synchronize_sched() while holding a spinlock would indeed be
> > very bad, but the code below seems to instead be invoking call_rcu(),
> > which is no problem.
> > 
> > Or am I missing something here?
> > 
> > 							Thanx, Paul
> > 
> > > static int set_all_monitor_traces(int state)
> > > {
> > > 	int rc = 0;
> > > 	struct dm_hw_stat_delta *new_stat = NULL;
> > > 	struct dm_hw_stat_delta *temp;
> > >
> > > 	spin_lock(&trace_state_lock);
> > >
> > > 	switch (state) {
> > > 	case TRACE_ON:
> > > 		rc |= register_trace_kfree_skb(trace_kfree_skb_hit);
> > > 		rc |= register_trace_napi_poll(trace_napi_poll_hit);
> > > 		break;
> > > 	case TRACE_OFF:
> > > 		rc |= unregister_trace_kfree_skb(trace_kfree_skb_hit);
> > > 		rc |= unregister_trace_napi_poll(trace_napi_poll_hit);
> > >
> > > 		tracepoint_synchronize_unregister();
> 
> This has synchronize_sched() inside.

Ah!  That would indeed be bad.

							Thanx, Paul