Re: [PATCH 3/5] perf tools: Allocate thread map_groups dynamicaly

[Jiri Olsa <jolsa@redhat.com>]

On Mon, Mar 17, 2014 at 04:13:53PM +0900, Namhyung Kim wrote:
> On Fri, 14 Mar 2014 15:00:04 +0100, Jiri Olsa wrote:
> > Moving towards sharing map groups within a process threads.
> >
> > Because of this we need the map groups to be dynamically
> > allocated. No other functional change is intended in here.
> > @@ -664,7 +664,7 @@ void thread__find_addr_map(struct thread *thread,
> >  	al->cpumode = cpumode;
> >  	al->filtered = false;
> >  
> > -	if (machine == NULL) {
> > +	if ((machine == NULL) || (mg == NULL)) {
> >  		al->map = NULL;
> >  		return;
> >  	}
> 
> What about the kernel threads?  I guess they're not using thread->mg but
> machine->kmaps instead?

The machine->kmaps is used to store kernel maps - core + modules.

All threads (including kernel ones) are using thread->mg,
kernel threads have empty /proc/x/maps file.

In case the sample address is detected within kernel space,
the machine->kmaps is used instead of thread->mg:

---
  void thread__find_addr_map(struct thread *thread, ...
  {
	struct map_groups *mg = thread__map_groups_get(thread);

  ...

        if (cpumode == PERF_RECORD_MISC_KERNEL && perf_host) {
                al->level = 'k';
                mg = &machine->kmaps;

  ...
	al->map = map_groups__find(mg, type, al->addr);
---

jirka