[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <C9F248C8-AF8D-40A1-A1AD-BCC39FBA01C7@linux.vnet.ibm.com>
Date: Wed, 7 Dec 2022 19:08:28 +0530
From: Athira Rajeev <atrajeev@...ux.vnet.ibm.com>
To: Arnaldo Carvalho de Melo <acme@...nel.org>,
Ian Rogers <irogers@...gle.com>
Cc: Peter Zijlstra <peterz@...radead.org>,
Ingo Molnar <mingo@...hat.com>,
Mark Rutland <mark.rutland@....com>,
Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
Jiri Olsa <jolsa@...nel.org>,
Namhyung Kim <namhyung@...nel.org>,
Nick Desaulniers <ndesaulniers@...gle.com>,
LKML <linux-kernel@...r.kernel.org>,
linux-perf-users@...r.kernel.org, bpf@...r.kernel.org,
Stephane Eranian <eranian@...gle.com>
Subject: Re: [PATCH 2/3] perf build: Use libtraceevent from the system
> On 06-Dec-2022, at 11:01 PM, Arnaldo Carvalho de Melo <acme@...nel.org> wrote:
>
> Em Tue, Dec 06, 2022 at 02:13:48PM -0300, Arnaldo Carvalho de Melo escreveu:
>> Em Tue, Dec 06, 2022 at 02:07:32PM -0300, Arnaldo Carvalho de Melo escreveu:
>>> Em Tue, Dec 06, 2022 at 02:01:43PM -0300, Arnaldo Carvalho de Melo escreveu:
>>>> Em Tue, Dec 06, 2022 at 01:41:41PM -0300, Arnaldo Carvalho de Melo escreveu:
>>>>> Now to look at the BUILD_BPF_SKEL=1 kaboom:
>>>>>
>>>>> [acme@...co perf]$ alias m
>>>>> alias m='rm -rf ~/libexec/perf-core/ ; make -k NO_LIBTRACEEVENT=1 BUILD_BPF_SKEL=1 O=/tmp/build/perf -C tools/perf install-bin && perf test python'
>>>>> [acme@...co perf]$ m
>>>>> make: Entering directory '/home/acme/git/perf/tools/perf'
>>>>> BUILD: Doing 'make -j8' parallel build
>>>>> <SNIP>
>>>>> /usr/bin/ld: /tmp/build/perf/perf-in.o: in function `add_work':
>>>>> /home/acme/git/perf/tools/perf/util/bpf_kwork.c:285: undefined reference to `perf_kwork_add_work'
>>>>> /usr/bin/ld: /tmp/build/perf/perf-in.o: in function `lock_contention_read':
>>>
>>> For that bpf_kwork.c see below. Now to see why the python binding is not
>>> building, I guess is unrelated and you have some other outstanding
>>> patch?
>>
>> Its related:
>>
>> [acme@...co perf]$ perf test -v python
>> Couldn't bump rlimit(MEMLOCK), failures may take place when creating BPF maps, etc
>> 14: 'import perf' in python :
>> --- start ---
>> test child forked, pid 1669872
>> python usage test: "echo "import sys ; sys.path.append('/tmp/build/perf/python'); import perf" | '/usr/bin/python3' "
>> Traceback (most recent call last):
>> File "<stdin>", line 1, in <module>
>> ImportError: /tmp/build/perf/python/perf.cpython-310-x86_64-linux-gnu.so: undefined symbol: tep_unload_plugins
>> test child finished with -1
>> ---- end ----
>> 'import perf' in python: FAILED!
>> [acme@...co perf]$
>>
>> Now checking why NO_LIBTRACEEVENT=1 fails with this...
>>
>> [acme@...co perf]$ find tools/perf/ -name "*.c" | xargs grep tep_unload_plugins
>> tools/perf/util/trace-event.c: tep_unload_plugins(t->plugin_list, t->pevent);
>> [acme@...co perf]$
>>
>> [acme@...co perf]$ grep trace-event tools/perf/util/python-ext-sources
>> util/trace-event.c
>> [acme@...co perf]$
>>
>> Trying to fix...
>
> I'm missing some detail, this isn't working, util/trace-event.c is still
> being built and linked.
Hi,
Observed similar issue with “builtin-trace.c”
In system without libtraceevent-devel, I could still see builtin-trace trying to get
compiled and hitting error. In my understanding, CONFIG_TRACE will be
disabled when libtraceevent is not present and hence builtin-trace should be disabled
which I am not seeing.
# rpm -qa|grep libtraceevent
libtraceevent-1.2.1-1.el8.ppc64le
# grep -i traceevent FEATURE-DUMP
feature-libtraceevent=0
Snippet from make logs:
builtin-trace.c:266:50: error: dereferencing pointer to incomplete type 'struct tep_format_field'
return __tp_field__init_uint(field, format_field->size, format_field->offset, needs_swap);
^~
builtin-trace.c: In function 'evsel__syscall_arg_fmt':
builtin-trace.c:361:25: error: 'struct evsel' has no member named 'tp_format'
et->fmt = calloc(evsel->tp_format->format.nr_fields, sizeof(struct syscall_arg_fmt));
Thanks
Athira
>
> The python binding should be usable without tracepoints, in fact its
> first usage was just to have access to the perf metaevents, see
> tools/perf/python/twatch.py.
>
> diff --git a/tools/perf/Makefile.perf b/tools/perf/Makefile.perf
> index f0e4daeef8120853..5bd67d54d036f281 100644
> --- a/tools/perf/Makefile.perf
> +++ b/tools/perf/Makefile.perf
> @@ -348,6 +348,10 @@ export PYTHON_EXTBUILD_LIB PYTHON_EXTBUILD_TMP
> python-clean := $(call QUIET_CLEAN, python) $(RM) -r $(PYTHON_EXTBUILD) $(OUTPUT)python/perf*.so
>
> PYTHON_EXT_SRCS := $(shell grep -v ^\# util/python-ext-sources)
> +ifneq ($(CONFIG_TRACEEVENT),y)
> + PYTHON_EXT_SRCS := $(call filter-out,$(PYTHON_EXT_SRCS),util/trace-event.c)
> +endif
> +
> PYTHON_EXT_DEPS := util/python-ext-sources util/setup.py $(LIBAPI)
>
> SCRIPTS = $(patsubst %.sh,%,$(SCRIPT_SH))
> diff --git a/tools/perf/util/python.c b/tools/perf/util/python.c
> index 728fe2f51759eb0d..72088aec2857b316 100644
> --- a/tools/perf/util/python.c
> +++ b/tools/perf/util/python.c
> @@ -1330,6 +1330,9 @@ static struct {
> static PyObject *pyrf__tracepoint(struct pyrf_evsel *pevsel,
> PyObject *args, PyObject *kwargs)
> {
> +#ifndef HAVE_LIBTRACEEVENT
> + return NULL;
> +#else
> struct tep_event *tp_format;
> static char *kwlist[] = { "sys", "name", NULL };
> char *sys = NULL;
> @@ -1344,6 +1347,7 @@ static PyObject *pyrf__tracepoint(struct pyrf_evsel *pevsel,
> return _PyLong_FromLong(-1);
>
> return _PyLong_FromLong(tp_format->id);
> +#endif // HAVE_LIBTRACEEVENT
> }
>
> static PyMethodDef perf__methods[] = {
Powered by blists - more mailing lists