Re: [tip:perf/core] perf: Add non-exec mmap() tracking

[Peter Zijlstra <peterz@infradead.org>]

On Wed, 2010-06-09 at 13:44 +0300, Pekka Enberg wrote:
> > @@ -3830,6 +3834,14 @@ static void perf_event_mmap_event(struct perf_mmap_event *mmap_event)
> >                if (!vma->vm_mm) {
> >                        name = strncpy(tmp, "[vdso]", sizeof(tmp));
> >                        goto got_name;
> > +               } else if (vma->vm_start <= vma->vm_mm->start_brk &&
> > +                               vma->vm_end >= vma->vm_mm->brk) {
> > +                       name = strncpy(tmp, "[heap]", sizeof(tmp));
> > +                       goto got_name;
> > +               } else if (vma->vm_start <= vma->vm_mm->start_stack &&
> > +                               vma->vm_end >= vma->vm_mm->start_stack) {
> > +                       name = strncpy(tmp, "[stack]", sizeof(tmp));
> > +                       goto got_name;
> >                }
> >
> >                name = strncpy(tmp, "//anon", sizeof(tmp));
> 
> Doesn't this change here break the JIT generated code region detection
> in map__new() of tools/perf/util/map.c? We generate a new
> "/tmp/perf-<pid>.map" filename for anonymous memory regions and check
> for that in dso__load() of tools/perf/util/symbol.c. 

Not unless you stick your executable code in the heap or on the stack.

If you use something like mmap(NULL, size, PROT_READ|PROT_WRITE|
PROT_EXEC, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0); to allocate your memory
everything should be fine.