Re: Wrong Perf Backtraces

Hi,
Each line is correct. I mean addresses match symbols. But some callers 
and callees do not match. Perhaps callchain misses some callers in 
between.
Regards

On 2020-03-29 05:13, ahmadkhorrami wrote:

> Hi Mr. Olsa,
> Thanks for your time. The problem is still there. Repeated lines are 
> not limited to inline functions. If my script works on your system, it 
> will show these lines.
> Regards.
> 
> On 2020-03-29 04:04, Jiri Olsa wrote:
> 
> On Sat, Mar 28, 2020 at 03:42:53AM +0430, ahmadkhorrami wrote:
> 
> Hi,
> Thanks. If you suggest the potentially bogus locations of the source 
> code, I
> will give a try.
> Regards.
> heya,
> the change below 'fixes' it for me:
> 
> $ perf script ...
> ...
> evince 2220122 1605573.007639:      11759 cycles:u:
> ffffffffaec012f0 [unknown] ([unknown])
> 7f93f17116b6 __mmap64+0x26 mmap64.c:59 (inlined)
> 7f93f17116b6 __mmap64+0x26 mmap64.c:47 (inlined)
> 
> it wasn't really broken, the output is just missing the source line
> info in perf script callchain output, which adds the "missing part",
> because for inlined entries the entry address stays the same for all
> its inlined parts
> 
> could you try the change?
> 
> thanks,
> jirka
> 
> diff --git a/tools/perf/util/evsel_fprintf.c 
> b/tools/perf/util/evsel_fprintf.c
> index 3b4842840db0..7349dfbbef2e 100644
> --- a/tools/perf/util/evsel_fprintf.c
> +++ b/tools/perf/util/evsel_fprintf.c
> @@ -174,8 +174,11 @@ int sample__fprintf_callchain(struct perf_sample 
> *sample, int left_alignment,
> if (print_srcline)
> printed += map__fprintf_srcline(map, addr, "\n  ", fp);
> 
> -            if (sym && sym->inlined)
> +            if (sym && sym->inlined) {
> +                if (node->srcline)
> +                    printed += fprintf(fp, " %s", node->srcline);
> printed += fprintf(fp, " (inlined)");
> +            }
> 
> if (!print_oneline)
> printed += fprintf(fp, "\n");