Re: [PATCH RFC] tracepoint: Introduce tracepoint callbacks executing with preempt on

[Mathieu Desnoyers]

----- On Apr 27, 2018, at 11:40 AM, rostedt rostedt@xxxxxxxxxxx wrote:

> On Fri, 27 Apr 2018 08:38:26 -0700
> "Paul E. McKenney" <paulmck@xxxxxxxxxxxxxxxxxx> wrote:
> 
>> On Fri, Apr 27, 2018 at 10:47:47AM -0400, Steven Rostedt wrote:
>> > On Fri, 27 Apr 2018 10:26:29 -0400 (EDT)
>> > Mathieu Desnoyers <mathieu.desnoyers@xxxxxxxxxxxx> wrote:
>> >   
>> > > The general approach and the implementation look fine, except for
>> > > one small detail: I would be tempted to explicitly disable preemption
>> > > around the call to the tracepoint callback for the rcuidle variant,
>> > > unless we plan to audit every tracer right away to remove any assumption
>> > > that preemption is disabled in the callback implementation.
>> > 
>> > I'm thinking that we do that audit. There shouldn't be many instances
>> > of it. I like the idea that a tracepoint callback gets called with
>> > preemption enabled.
>> 
>> Are you really sure you want to increase your state space that much?
> 
> Why not? The code I have in callbacks already deals with all sorts of
> context - normal, softirq, irq, NMI, preemption disabled, irq
> disabled.

It does so by disabling preemption in the callbacks, even when it's
redundant with the guarantees already provided by tracepoint-sched-rcu
and by kprobes. It's not that great for a fast-path.

Thanks,

Mathieu


-- 
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com