Re: [PATCH RESEND WITH CCs v3 3/4] perf tools: enable dwarf_callchain_users on aarch64

[Leo Yan]

Hi Alexandre,

On Thu, Mar 04, 2021 at 04:32:54PM +0000, Alexandre Truong wrote:
> On arm64, enable dwarf_callchain_users which will be needed
> to do a dwarf unwind in order to get the caller of the leaf frame.
> 
> Signed-off-by: Alexandre Truong <alexandre.truong@xxxxxxx>
> Cc: John Garry <john.garry@xxxxxxxxxx>
> Cc: Will Deacon <will@xxxxxxxxxx>
> Cc: Mathieu Poirier <mathieu.poirier@xxxxxxxxxx>
> Cc: Leo Yan <leo.yan@xxxxxxxxxx>
> Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
> Cc: Ingo Molnar <mingo@xxxxxxxxxx>
> Cc: Arnaldo Carvalho de Melo <acme@xxxxxxxxxx>
> Cc: Mark Rutland <mark.rutland@xxxxxxx>
> Cc: Alexander Shishkin <alexander.shishkin@xxxxxxxxxxxxxxx>
> Cc: Jiri Olsa <jolsa@xxxxxxxxxx>
> Cc: Namhyung Kim <namhyung@xxxxxxxxxx>
> Cc: Kemeng Shi <shikemeng@xxxxxxxxxx>
> Cc: Ian Rogers <irogers@xxxxxxxxxx>
> Cc: Andi Kleen <ak@xxxxxxxxxxxxxxx>
> Cc: Kan Liang <kan.liang@xxxxxxxxxxxxxxx>
> Cc: Jin Yao <yao.jin@xxxxxxxxxxxxxxx>
> Cc: Adrian Hunter <adrian.hunter@xxxxxxxxx>
> Cc: Suzuki K Poulose <suzuki.poulose@xxxxxxx>
> Cc: Al Grant <al.grant@xxxxxxx>
> Cc: James Clark <james.clark@xxxxxxx>
> Cc: Wilco Dijkstra <wilco.dijkstra@xxxxxxx>
> ---
>  tools/perf/builtin-report.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c
> index 2a845d6cac09..93661a3eaeb1 100644
> --- a/tools/perf/builtin-report.c
> +++ b/tools/perf/builtin-report.c
> @@ -405,6 +405,10 @@ static int report__setup_sample_type(struct report *rep)
>  
>  	callchain_param_setup(sample_type);
>  
> +	if (callchain_param.record_mode == CALLCHAIN_FP &&
> +			strncmp(rep->session->header.env.arch, "aarch64", 7) == 0)
> +		dwarf_callchain_users = true;
> +

I don't have knowledge for dwarf or FP.

This patch is suspicious for me that since it only fixes the issue for
"perf report" command, but it cannot support "perf script".

I did a quick testing for "perf script" command with the test code from
patch 04, seems to me it cannot fix the fp omitting issue for
"perf script" command:

  arm64_fp_test 11211  2282.355095:     176307 cycles: 
              aaaac2e40740 f2+0x10 (/root/arm64_fp_test)
              aaaac2e4061c main+0xc (/root/arm64_fp_test)
              ffff961fbd24 __libc_start_main+0xe4 (/usr/lib/aarch64-linux-gnu/libc-2.28.so)
              aaaac2e4065c _start+0x34 (/root/arm64_fp_test)

Could you check for this?  Thanks!

Leo

>  	if (rep->stitch_lbr && (callchain_param.record_mode != CALLCHAIN_LBR)) {
>  		ui__warning("Can't find LBR callchain. Switch off --stitch-lbr.\n"
>  			    "Please apply --call-graph lbr when recording.\n");
> -- 
> 2.23.0
>