Re: [PATCH v3] ftrace: add a tracepoint for__raise_softirq_irqoff()

[Mathieu Desnoyers]

* Xiao Guangrong (xiaoguangrong@xxxxxxxxxxxxxx) wrote:
> 
> 
> Mathieu Desnoyers wrote:
> > * Xiao Guangrong (xiaoguangrong@xxxxxxxxxxxxxx) wrote:
> >> From: Mathieu Desnoyers <mathieu.desnoyers@xxxxxxxxxx>
> >>
> 
> >> +#ifdef CONFIG_TRACEPOINTS
> >> +extern void __raise_softirq_irqoff(unsigned int nr);
> >> +#else
> >>  #define __raise_softirq_irqoff(nr) do { or_softirq_pending(1UL << (nr)); } while (0)
> > 
> > Can you put the 
> > 	trace_irq_softirq_raise(nr);
> > 
> > directly in the define rather than adding this weird CONFIG_TRACEPOINTS?
> > (and change the define for a static inline), something like :
> > 
> > static inline void __raise_softirq_irqoff(unsigned int nr)
> > {
> > 	trace_irq_softirq_raise(nr);
> > 	or_softirq_pending(1UL << (nr);
> > }
> > 
> > This would ensure we don't add a function call on the
> > __raise_softirq_irqoff() fast-path.
> > 
> 
> We did this in v2, and we think it is better for same reason.
> But ...
> 
> > Beware of circular include dependencies though. The tracepoints are
> > meant not to have this kind of problems (I try to keep the dependencies
> > very minimalistic), but I wonder if Steven's TRACE_EVENT is now ok on
> > this aspect.
> > 
> 
> We encount this type of problem in v2.
> So we move to this version(v3).
> 
> > If TRACE_EVENT happens to pose problems with circular header
> > dependencies, then try moving to the DECLARE_TRACE/DEFINE_TRACE scheme
> > which has been more thoroughly tested as a first step.
> > 
> 
> IMHO, TRACE_EVENT framework is better for its more generic as ingo said,
> and it also provide ftrace support which means user can view tracepoint
> information from /debug/tracing/events.
> 
> Although this TRACE_EVENT happens to expose problems with circular header
> dependencies, we should not refuse using TRACE_EVENT, instead we should
> try to fix it for the whole TRACE_EVENT facility later.
> 

I partially agree with you :

Yes, we should try to fix TRACE_EVENT, but we should fix it _before_ we
start using it widely. Circular header dependencies is a real problem
with TRACE_EVENT right now.

Until we fix this, I will be tempted to stay with a known-good solution,
which is DECLARE/DEFINE_TRACE.

Mathieu

> Thanks
> 
> > Mathieu
> > 

-- 
Mathieu Desnoyers
OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/