Re: [PATCH 2/2] perf lock: Fix output of tracing lock events

[Ingo Molnar <mingo@elte.hu>]

* Peter Zijlstra <peterz@infradead.org> wrote:

> On Thu, 2009-12-17 at 11:09 +0100, Ingo Molnar wrote:
> > * Peter Zijlstra <peterz@infradead.org> wrote:
> > 
> > > >  	TP_STRUCT__entry(
> > > > +		__field(struct lockdep_map *, lockdep_addr)
> > > >  		__field(unsigned int, flags)
> > > >  		__string(name, lock->name)
> > > >  	),
> > > 
> > > I feel a bit awkward explicitly leaking kernel pointers like that. All this 
> > > is accessible by root only (for now) so its not too harmfull, but sitll.
> > 
> > What would you suggest as a 'natural lock class key'? The name? It might not 
> > be unique enough.
> > 
> > Other kernel objects like tasks, cpus, inodes, pages all have some natural key 
> > that isnt a kernel pointer - but locks are a bit special.
> 
> Well, yeah, that's the problem, and we use the pointer for exactly this 
> purpose inside the kernel too, its just that its a blatant data leak when we 
> expose it to userspace like that.
> 
> On the other hand, adding some ID generation just so we can expose it seems 
> silly too.
> 
> Why do we need to have instance resolution?

Does the tool make use of it to classify stats?

	Ingo