Re: [PATCH v2 0/6] Add tracepoints to trace all system IRQs

[Frederic Weisbecker <fweisbec@gmail.com>]

On Thu, Jul 28, 2011 at 02:12:49PM +0200, Peter Zijlstra wrote:
> On Wed, 2011-07-27 at 11:57 -0700, Vaibhav Nagarnaik wrote:
> > This is a patch series aiming to trace all the system IRQs. The
> > tracepoints make it easier to understand IRQ interaction with other
> > system events and find out how much time is spent in kernel space versus
> > user space.
> 
> That doesn't cover the actual rational, you can get that information by
> using CONFIG_IRQ_TIME_ACCOUNTING.
> 
> > Patches 1-5 add tracepoints to the generic kernel IRQ handlers to
> > benefit all the architectures, as suggested by Frederic. Patch#6 adds
> > trace events to trace all platform specific interrupt vectors which do
> > not have any generic handlers. This patch traces x86 IRQ vectors but can
> > be used by other architectures to trace their vectors too.
> > 
> > Changelog v2-v1:
> > * Fix patch#3 to trace IRQ work run vector even when the irq work list
> >   is empty
> > * Fix patch#5 to use updated trace event names,
> >   call_function_ipi_(entry|exit) and
> >   call_function_single_ipi_(entry|exit)
> > * Rebase to 3.0
> > * Update changelog
> 
> I still completely hate how you make new events for every interrupt
> source. That's a massive pollution of the event space.
> 
> Also, since you do arch specific things anyway, why not from a single
> tracepoint into the arch interrupt stubs and be done with it, no need to
> puke all over the rest of the code.

Because that requires to put the tracepoints in every interrupt sources
for every archs instead of putting one in the generic space.