On Sun, 14 Dec 2014 10:08:54 -0800 "Paul E. McKenney" <[email protected]> wrote:
> On Sun, Dec 14, 2014 at 11:53:32AM -0500, Steven Rostedt wrote: > > On Sun, 14 Dec 2014 11:41:05 -0500 > > Steven Rostedt <[email protected]> wrote: > > > > > From: "Steven Rostedt (Red Hat)" <[email protected]> > > > > > > In order to move enabling of trace events to just after mm_init(), the > > > tracepoint enable code can not use call_rcu_sched() because rcu isn't > > > even initialized yet. Since this can only happen before SMP is set up > > > (and even before interrupts are set up), there's no reason to use > > > call_rcu_sched() at this point. > > > > > > Instead, create a variable called tracepoint_rcu_safe that gets enabled > > > via early_initcall() and if that is not set, free the code directly > > > instead of using call_rcu_sched(). > > > > > > This allows us to enable tracepoints early without issues. > > > > > > Cc: Mathieu Desnoyers <[email protected]> > > > Cc: Paul E. McKenney <[email protected]> > > > Cc: Thomas Gleixner <[email protected]> > > > Signed-off-by: Steven Rostedt <[email protected]> > > With the addition of read_mostly, and given that I am not going to mess > with call_rcu() this late in the 3.19 process without a blazingly good > reason: > > Reviewed-by: Paul E. McKenney <[email protected]> Thanks! > > Please note that you can use call_rcu() and friends as soon as rcu_init() > returns. The callbacks won't be invoked until early_initcall() time, > but they will be properly queued. > > Please note also that there are places where turning a call_rcu() into > a direct function call don't work, even at times when preemption is > disabled and there is only one CPU. One example is where single-threaded > code uses call_rcu() on a list element of a list that it is traversing > within an RCU read-side critical section. A direct call to the RCU > callback could potentially destroy the pointers that the traversal was > going to use to find the next element. This means that we cannot make > call_rcu() do direct calls to the callback, as that would break quite > a bit of existing code. > > Is there some definite point during boot before which you won't need to > invoke call_rcu_sched() for tracing? I am guessing "no", but have to ask. > I can probably make call_rcu_sched() work arbitrarily early, but it is a > bit uglier. And this assumes that irqs_disabled_flags(local_irq_save()) > returns true during early boot. I would -hope- this would be true! ;-) > With your feed back, and because I would like this to go into 3.19, I would like to keep the current patch as is (with the read_mostly update, which I'm currently testing). We can always change it later after call_rcu() has been changed. -- Steve -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
