Re: [Question] Hooks for scheduler tracing (CFS)

[ankita@in.ibm.com (Ankita Garg)]

On Thu, Jul 26, 2007 at 12:22:31PM -0400, Frank Ch. Eigler wrote:
> Hi -
> 
> On Thu, Jul 26, 2007 at 11:02:26AM -0400, Mathieu Desnoyers wrote:
> > [...]
> > > > The problem is also in _stp_print_flush, not *only* in relay code:
> > > > void _stp_print_flush (void)
> > > > ...
> > > >                 spin_lock(&_stp_print_lock);
> > > >                 spin_unlock(&_stp_print_lock);
> > > > 
> > > > Those will turn into mutexes with -rt.
> > > 
> > > Indeed,
> 
> (Though actually that bug was fixed some time ago.)
> 
> 
> > > plus systemtap-generated locking code uses rwlocks,
> > > local_irq_save/restore or preempt_disable, in various places.  Could
> > > someone point to a place that spells out what would be more
> > > appropriate way of ensuring atomicity while being compatible with -rt?
> > 
> > AFAIK, for your needs either:
> > [...]
> > - Use per-cpu data with preempt disabling/irq disabling
> 
> As in local_irq_save / preempt_disable?  Yes, already done.
> 
> > - Use the original "real" spin locks/rwlocks (raw_*).
> > [...]
> 
> It was unclear from the OLS paper whether the spin_lock_irq* family of
> functions also had to be moved to the raw forms.

By making the locks of the raw_* type, spin_lock_irq* functions would
then automatically disable hardware interrupts.

> 
> > You just don't want to sleep in the tracing code. [...]  Since you
> > will likely disable preemption, make sure your tracing code executes
> > in a deterministic time.
> 
> Definitely, that has always been the case.
> 
> > Make sure that the sub-buffer switch code respects that too [...]
> 
> We will review that part of the related code.
> 
> - FChE

-- 
Regards,
Ankita Garg (ankita@in.ibm.com)
Linux Technology Center
IBM India Systems & Technology Labs, 
Bangalore, India