Re: [PATCH 01/23] perf record: Add --all-user/--all-kernel options

From: Arnaldo Carvalho de Melo
Date: Tue Feb 16 2016 - 11:56:12 EST


Em Mon, Feb 15, 2016 at 09:34:31AM +0100, Jiri Olsa escreveu:
> Allow user to easily switch all events to user or
> kernel space with simple --all-user or --all-kernel
> options.
>
> This will be handy within perf mem/c2c wrappers to
> switch easily monitoring modes.

Humm, some oddities:

# perf record --all-user -e cycles -a
# perf report --tui # And then notice there are some kernel samples, zoom into kernel DSO
Samples: 7 of event 'cycles', Event count (approx.): 3158810, DSO: [kernel.vmlinux]
Overhead Comman Symbol
84.05% chrome [k] page_fault
15.73% chrome [k] entry_SYSCALL_64
0.22% chrome [k] apic_timer_interrupt

Ditto when using --all-kernel, some userspace samples are there, also perhaps we
should show "cycles:u" or "cycles:k" when this --all-user or --all-kernel features
are used.

Need to investigate why there are kernel samples when --all-user is used and
the other way around as well.

Also if I use both it quietly accepts and shows just one of them, I guess we
should bail out in case someone tries both.

- Arnaldo

> Link: http://lkml.kernel.org/n/tip-hdd2u0y1o8cwfpplrpzc1iqd@xxxxxxxxxxxxxx
> Signed-off-by: Jiri Olsa <jolsa@xxxxxxxxxx>
> ---
> tools/perf/Documentation/perf-record.txt | 6 ++++++
> tools/perf/builtin-record.c | 4 ++++
> tools/perf/perf.h | 2 ++
> tools/perf/util/evsel.c | 10 ++++++++++
> 4 files changed, 22 insertions(+)
>
> diff --git a/tools/perf/Documentation/perf-record.txt b/tools/perf/Documentation/perf-record.txt
> index fbceb631387c..19aa17532a16 100644
> --- a/tools/perf/Documentation/perf-record.txt
> +++ b/tools/perf/Documentation/perf-record.txt
> @@ -341,6 +341,12 @@ Specify vmlinux path which has debuginfo.
> --buildid-all::
> Record build-id of all DSOs regardless whether it's actually hit or not.
>
> +--all-kernel::
> +Configure all used events to run in kernel space.
> +
> +--all-user::
> +Configure all used events to run in user space.
> +
> SEE ALSO
> --------
> linkperf:perf-stat[1], linkperf:perf-list[1]
> diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
> index caa8235ed027..b64f31a4c993 100644
> --- a/tools/perf/builtin-record.c
> +++ b/tools/perf/builtin-record.c
> @@ -1151,6 +1151,10 @@ struct option __record_options[] = {
> "per thread proc mmap processing timeout in ms"),
> OPT_BOOLEAN(0, "switch-events", &record.opts.record_switch_events,
> "Record context switch events"),
> + OPT_BOOLEAN(0, "all-kernel", &record.opts.all_kernel,
> + "Configure all used events to run in kernel space."),
> + OPT_BOOLEAN(0, "all-user", &record.opts.all_user,
> + "Configure all used events to run in user space."),
> OPT_STRING(0, "clang-path", &llvm_param.clang_path, "clang path",
> "clang binary to use for compiling BPF scriptlets"),
> OPT_STRING(0, "clang-opt", &llvm_param.clang_opt, "clang options",
> diff --git a/tools/perf/perf.h b/tools/perf/perf.h
> index 90129accffbe..5381a01c0610 100644
> --- a/tools/perf/perf.h
> +++ b/tools/perf/perf.h
> @@ -58,6 +58,8 @@ struct record_opts {
> bool full_auxtrace;
> bool auxtrace_snapshot_mode;
> bool record_switch_events;
> + bool all_kernel;
> + bool all_user;
> unsigned int freq;
> unsigned int mmap_pages;
> unsigned int auxtrace_mmap_pages;
> diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
> index 467808680ee4..6ae20d0056de 100644
> --- a/tools/perf/util/evsel.c
> +++ b/tools/perf/util/evsel.c
> @@ -898,6 +898,16 @@ void perf_evsel__config(struct perf_evsel *evsel, struct record_opts *opts)
> if (evsel->precise_max)
> perf_event_attr__set_max_precise_ip(attr);
>
> + if (opts->all_user) {
> + attr->exclude_kernel = 1;
> + attr->exclude_user = 0;
> + }
> +
> + if (opts->all_kernel) {
> + attr->exclude_kernel = 0;
> + attr->exclude_user = 1;
> + }
> +
> /*
> * Apply event specific term settings,
> * it overloads any global configuration.
> --
> 2.4.3