diff mbox series

[v1,3/3] perf cs-etm: Correct the packet usage for instruction sample

Message ID 20190830062421.31275-4-leo.yan@linaro.org (mailing list archive)
State New, archived
Headers show
Series perf cs-etm: Add support for callchain | expand

Commit Message

Leo Yan Aug. 30, 2019, 6:24 a.m. UTC
It uses 'tidq->packet' rather than 'tidq->prev_packet' to generate
instruction sample, comparing against the thread stack and the branch
samples which are using the 'tidp->prev_packet', thus this leads the
instruction sample to use one ahead packet than thread stack and branch
samples.

As result, the instruction's call chain can be wrongly displayed as
below:

  main  1579        100      instructions:
  	ffff000010214854 perf_event_update_userpage+0x4c ([kernel.kallsyms])
  	ffff000010214850 perf_event_update_userpage+0x48 ([kernel.kallsyms])
  	ffff000010219360 perf_swevent_add+0x88 ([kernel.kallsyms])
  	ffff0000102135f4 event_sched_in.isra.57+0xbc ([kernel.kallsyms])
  	ffff0000102137a0 group_sched_in+0x60 ([kernel.kallsyms])
  	ffff000010213b84 flexible_sched_in+0xfc ([kernel.kallsyms])
  	ffff00001020c0b4 visit_groups_merge+0x12c ([kernel.kallsyms])

In this log, the continuous two lines includes two functions, the up
line contains the child function info and the bottom line contains
its parent function, and so forth.  But if review the first two lines:

  perf_event_update_userpage+0x4c  => the sampled instruction
  perf_event_update_userpage+0x48  => the parent function's calling

The child function and parent function is the same function
perf_event_update_userpage(), but perf_event_update_userpage() isn't
a recursive function, thus this calling sequence shouldn't never happen.
This is caused by the instruction sample using the 'tidq->packet', but
this packet is not handled yet by thread stack, the thread stack is
delayed to handle one return packet for stack popping.

To fix this issue, we can simply to use 'tidq->prev_packet' to generate
the instruction sample, this allows the thread stack to push/pop
properly for instruction sample.  Finally, we can get below result:

  main  1579        100      instructions:
	ffff000010214854 perf_event_update_userpage+0x4c ([kernel.kallsyms])
	ffff000010219360 perf_swevent_add+0x88 ([kernel.kallsyms])
	ffff0000102135f4 event_sched_in.isra.57+0xbc ([kernel.kallsyms])
	ffff0000102137a0 group_sched_in+0x60 ([kernel.kallsyms])
	ffff000010213b84 flexible_sched_in+0xfc ([kernel.kallsyms])
	ffff00001020c0b4 visit_groups_merge+0x12c ([kernel.kallsyms])

Signed-off-by: Leo Yan <leo.yan@linaro.org>
---
 tools/perf/util/cs-etm.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
diff mbox series

Patch

diff --git a/tools/perf/util/cs-etm.c b/tools/perf/util/cs-etm.c
index ad573d3bd305..0bd2b3c48394 100644
--- a/tools/perf/util/cs-etm.c
+++ b/tools/perf/util/cs-etm.c
@@ -1414,7 +1414,7 @@  static int cs_etm__sample(struct cs_etm_queue *etmq,
 	struct cs_etm_packet *tmp;
 	int ret;
 	u8 trace_chan_id = tidq->trace_chan_id;
-	u64 instrs_executed = tidq->packet->instr_count;
+	u64 instrs_executed = tidq->prev_packet->instr_count;
 
 	tidq->period_instructions += instrs_executed;
 
@@ -1445,7 +1445,7 @@  static int cs_etm__sample(struct cs_etm_queue *etmq,
 		 */
 		u64 offset = (instrs_executed - instrs_over - 1);
 		u64 addr = cs_etm__instr_addr(etmq, trace_chan_id,
-					      tidq->packet, offset);
+					      tidq->prev_packet, offset);
 
 		ret = cs_etm__synth_instruction_sample(
 			etmq, tidq, addr, etm->instructions_sample_period);