Re: [PATCH RFC 00/11] lock monitor: Separate features related to lock

[Jason Baron <jbaron@redhat.com>]

On Wed, Mar 17, 2010 at 10:52:30AM +0100, Ingo Molnar wrote:
> * Frederic Weisbecker <fweisbec@gmail.com> wrote:
> 
> > > You add chained indirect calls into all lock ops, that's got to hurt.
> > 
> > Well, the idea was not bad at the first glance. It was separating lockdep 
> > and lock events codes.
> > 
> > But indeed, the indirect calls plus the locking are not good for such a fast 
> > path.
> 
> What would be nice to have is some sort of dynamic patching approach to enable 
> _both_ lockdep, lockstat and perf lock.
> 

right. this would allow distros to ship lockdep, lockstat in their
default kernels as a runtime option.


> If TRACE_EVENT() tracepoints were patchable we could use them. (but they arent 
> right now)
> 

right. I'm going to re-post the jump labeling work again soon, which
implicitly makes all TRACE_EVENT() tracepoints into dynamic patch
points. The jump label approach can also be deployed independently of
the tracepoints.

Also, any hints, suggestions on where to start with this type of
project? I thought a lot of the lockdep overhead was tied up in the data
structures? If its just a matter of identifying the dynamic patch
points. I can convert them to jump label and run benchmarks, pretty
easily.

thanks,

-Jason