Re: [PATCH 1/2] perf cs-etm: Support sample flags 'insn' and 'insnlen'

From: Mathieu Poirier
Date: Mon Aug 19 2019 - 14:08:39 EST


On Thu, 15 Aug 2019 at 02:30, Leo Yan <leo.yan@xxxxxxxxxx> wrote:
>
> The synthetic branch and instruction samples are missed to set
> instruction related info, thus perf tool fails to display samples with
> flags '-F,+insn,+insnlen'.
>
> CoreSight trace decoder has provided sufficient information to decide
> the instruction size based on the isa type: A64/A32 instruction are
> 32-bit size, but one exception is the T32 instruction size, which might
> be 32-bit or 16-bit.
>
> This patch handles for these cases and it reads the instruction values
> from DSO file; thus can support flags '-F,+insn,+insnlen'.
>
> Before:
>
> # perf script -F,insn,insnlen,ip,sym
> 0 [unknown] ilen: 0
> ffff97174044 _start ilen: 0
> ffff97174938 _dl_start ilen: 0
> ffff97174938 _dl_start ilen: 0
> ffff97174938 _dl_start ilen: 0
> ffff97174938 _dl_start ilen: 0
> ffff97174938 _dl_start ilen: 0
> ffff97174938 _dl_start ilen: 0
> ffff97174938 _dl_start ilen: 0
> ffff97174938 _dl_start ilen: 0
>
> [...]
>
> After:
>
> # perf script -F,insn,insnlen,ip,sym
> 0 [unknown] ilen: 0
> ffff97174044 _start ilen: 4 insn: 2f 02 00 94
> ffff97174938 _dl_start ilen: 4 insn: c1 ff ff 54
> ffff97174938 _dl_start ilen: 4 insn: c1 ff ff 54
> ffff97174938 _dl_start ilen: 4 insn: c1 ff ff 54
> ffff97174938 _dl_start ilen: 4 insn: c1 ff ff 54
> ffff97174938 _dl_start ilen: 4 insn: c1 ff ff 54
> ffff97174938 _dl_start ilen: 4 insn: c1 ff ff 54
> ffff97174938 _dl_start ilen: 4 insn: c1 ff ff 54
> ffff97174938 _dl_start ilen: 4 insn: c1 ff ff 54
>
> [...]
>
> Cc: Mathieu Poirier <mathieu.poirier@xxxxxxxxxx>
> Cc: Suzuki Poulouse <suzuki.poulose@xxxxxxx>
> Cc: Mike Leach <mike.leach@xxxxxxxxxx>
> Cc: Robert Walker <robert.walker@xxxxxxx>
> Cc: coresight@xxxxxxxxxxxxxxxx
> Cc: linux-arm-kernel@xxxxxxxxxxxxxxxxxxx
> Signed-off-by: Leo Yan <leo.yan@xxxxxxxxxx>
> ---
> tools/perf/util/cs-etm.c | 35 ++++++++++++++++++++++++++++++++++-
> 1 file changed, 34 insertions(+), 1 deletion(-)
>
> diff --git a/tools/perf/util/cs-etm.c b/tools/perf/util/cs-etm.c
> index ed6f7fd5b90b..b3a5daaf1a8f 100644
> --- a/tools/perf/util/cs-etm.c
> +++ b/tools/perf/util/cs-etm.c
> @@ -1076,6 +1076,35 @@ bool cs_etm__etmq_is_timeless(struct cs_etm_queue *etmq)
> return !!etmq->etm->timeless_decoding;
> }
>
> +static void cs_etm__copy_insn(struct cs_etm_queue *etmq,
> + u64 trace_chan_id,
> + const struct cs_etm_packet *packet,
> + struct perf_sample *sample)
> +{
> + /*
> + * It's pointless to read instructions for the CS_ETM_DISCONTINUITY
> + * packet, so directly bail out with 'insn_len' = 0.
> + */
> + if (packet->sample_type == CS_ETM_DISCONTINUITY) {
> + sample->insn_len = 0;
> + return;
> + }
> +
> + /*
> + * T32 instruction size might be 32-bit or 16-bit, decide by calling
> + * cs_etm__t32_instr_size().
> + */
> + if (packet->isa == CS_ETM_ISA_T32)
> + sample->insn_len = cs_etm__t32_instr_size(etmq, trace_chan_id,
> + sample->ip);
> + /* Otherwise, A64 and A32 instruction size are always 32-bit. */
> + else
> + sample->insn_len = 4;
> +
> + cs_etm__mem_access(etmq, trace_chan_id, sample->ip,
> + sample->insn_len, (void *)sample->insn);
> +}
> +
> static int cs_etm__synth_instruction_sample(struct cs_etm_queue *etmq,
> struct cs_etm_traceid_queue *tidq,
> u64 addr, u64 period)
> @@ -1097,9 +1126,10 @@ static int cs_etm__synth_instruction_sample(struct cs_etm_queue *etmq,
> sample.period = period;
> sample.cpu = tidq->packet->cpu;
> sample.flags = tidq->prev_packet->flags;
> - sample.insn_len = 1;
> sample.cpumode = event->sample.header.misc;
>
> + cs_etm__copy_insn(etmq, tidq->trace_chan_id, tidq->packet, &sample);
> +
> if (etm->synth_opts.last_branch) {
> cs_etm__copy_last_branch_rb(etmq, tidq);
> sample.branch_stack = tidq->last_branch;
> @@ -1159,6 +1189,9 @@ static int cs_etm__synth_branch_sample(struct cs_etm_queue *etmq,
> sample.flags = tidq->prev_packet->flags;
> sample.cpumode = event->sample.header.misc;
>
> + cs_etm__copy_insn(etmq, tidq->trace_chan_id, tidq->prev_packet,
> + &sample);
> +

The code seems to be correct. I have also tested this patch.

Reviewed-by: Mathieu Poirier <mathieu.poirier@xxxxxxxxxx>

> /*
> * perf report cannot handle events without a branch stack
> */
> --
> 2.17.1
>