Re: A concern about overflow ring buffer mode

From: Liang, Kan
Date: Fri Oct 26 2018 - 15:07:47 EST




On 10/26/2018 3:02 PM, Arnaldo Carvalho de Melo wrote:
So, I'm adding the following to my tree to help in diagnosing problems
with this overwrite mode:


Actually, you can use per-event overwrite term to disable overwrite mode for perf top.

/*
* Check per-event overwrite term.
* perf top should support consistent term for all events.
* - All events don't have per-event term
* E.g. "cpu/cpu-cycles/,cpu/instructions/"
* Nothing change, return 0.
* - All events have same per-event term
* E.g. "cpu/cpu-cycles,no-overwrite/,cpu/instructions,no-overwrite/
* Using the per-event setting to replace the opts->overwrite if
* they are different, then return 0.
* - Events have different per-event term
* E.g. "cpu/cpu-cycles,overwrite/,cpu/instructions,no-overwrite/"
* Return -1
* - Some of the event set per-event term, but some not.
* E.g. "cpu/cpu-cycles/,cpu/instructions,no-overwrite/"
* Return -1
*/
static int perf_top__overwrite_check(struct perf_top *top)
{

Thanks,
Kan


From 40feb09001c7cc2ba8aeaa0a8f03b6d28fa4ca95 Mon Sep 17 00:00:00 2001
From: Arnaldo Carvalho de Melo <acme@xxxxxxxxxx>
Date: Fri, 26 Oct 2018 15:55:23 -0300
Subject: [PATCH 1/1] perf top: Allow disabling the overwrite mode

In ebebbf082357 ("perf top: Switch default mode to overwrite mode") we
forgot to leave a way to disable that new default, add a --overwrite
option that can be disabled using --no-overwrite, since the code already
in such a way that we can readily disable this mode.

This is useful when investigating bugs with this mode like the recent
report from David Miller where lots of unknown symbols appear due to
disabling the events while processing them which disables all record
types, not just PERF_RECORD_SAMPLE, which makes it impossible to resolve
maps when we lose PERF_RECORD_MMAP records.

This can be easily seen while building a kernel, when there are lots of
short lived processes.

Cc: Adrian Hunter <adrian.hunter@xxxxxxxxx>
Cc: Andi Kleen <ak@xxxxxxxxxxxxxxx>
Cc: David Ahern <dsahern@xxxxxxxxx>
Cc: David Miller <davem@xxxxxxxxxxxxx>
Cc: Jin Yao <yao.jin@xxxxxxxxxxxxxxx>
Cc: Jiri Olsa <jolsa@xxxxxxxxxx>
Cc: Kan Liang <kan.liang@xxxxxxxxx>
Cc: Namhyung Kim <namhyung@xxxxxxxxxx>
Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
Cc: Wang Nan <wangnan0@xxxxxxxxxx>
Link: https://lkml.kernel.org/n/tip-oqgsz2bq4kgrnnajrafcdhie@xxxxxxxxxxxxxx
Signed-off-by: Arnaldo Carvalho de Melo <acme@xxxxxxxxxx>
---
tools/perf/Documentation/perf-top.txt | 5 +++++
tools/perf/builtin-top.c | 2 ++
2 files changed, 7 insertions(+)

diff --git a/tools/perf/Documentation/perf-top.txt b/tools/perf/Documentation/perf-top.txt
index 114fda12aa49..d4be6061fe1c 100644
--- a/tools/perf/Documentation/perf-top.txt
+++ b/tools/perf/Documentation/perf-top.txt
@@ -242,6 +242,11 @@ Default is to monitor all CPUS.
--hierarchy::
Enable hierarchy output.
+--overwrite::
+ This is the default, but for investigating problems with it or any other strange
+ behaviour like lots of unknown samples, we may want to disable this mode by using
+ --no-overwrite.
+
--force::
Don't do ownership validation.
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index d21d8751e749..214fad747b04 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -1372,6 +1372,8 @@ int cmd_top(int argc, const char **argv)
"Show raw trace event output (do not use print fmt or plugins)"),
OPT_BOOLEAN(0, "hierarchy", &symbol_conf.report_hierarchy,
"Show entries in a hierarchy"),
+ OPT_BOOLEAN(0, "overwrite", &top.record_opts.overwrite,
+ "Use a backward ring buffer, default: yes"),
OPT_BOOLEAN(0, "force", &symbol_conf.force, "don't complain, do it"),
OPT_UINTEGER(0, "num-thread-synthesize", &top.nr_threads_synthesize,
"number of thread to run event synthesize"),