Skip to contentSkip to conversation

Python module cleanup

PATCH Superseded Ian Rogers 2 participants 29 messages Nov 9, 2024 – Nov 19, 2024 perf events

v6 of 7 Superseded Latest is v7 →

lore
Share & tools
lore

Ian Rogers

author reviewer
#0 lore raw

v6 cover Python module cleanup

This patch:
- removes workarounds for Python 2 module support due to long deprecation and challenges in developing new code;
- constifies variables and parameters to functions;
- removes python.c stub code which existed due to missing functions that are defined in the builtin-* files, in general the builtin-* code is moved into util;
- remove bench and test perf C code from the python module;
- adds parse_events to the python perf module.
- improves upon some of the existing python perf module functins.

v6. Rebase. When moving find_scripts, separate the switch to using
is_directory_at into a patch before moving the function, Arnaldo/Namhyung. Add acked/tested tags from Arnaldo and Adrian.
v5. Rebase. Fix NO_LIBBPF and NO_AUXTRACE related build failures
caught by Arnaldo and the build-test. Fix NO_AUXTRACE x86 arch_fetch_insn in the process, which was incorrectly using an empty weak symbol stub.
v4. Rebase. Fix the parse events evsel to be embedded in a
pyrf_evsel. Add __str__/__repr__ functions to evlist and evsel. Throw an exception for a bad evlist index.
v3. Move is_directory_at to patch 6 rather than patch 7, respond to
review feedback on the list from Namhyung.
v2. Add the bottom 4 bullet points - 13 more patches.

Ian Rogers (22):
perf python: Remove python 2 scripting support
perf python: Constify variables and parameters
perf python: Remove unused #include
perf script: Move scripting_max_stack out of builtin perf kvm: Move functions used in util out of builtin perf script: Use openat for directory iteration
perf script: Move find_scripts to browser/scripts.c perf stat: Move stat_config into config.c
perf script: Move script_spec code to trace-event-scripting.c perf script: Move script_fetch_insn to trace-event-scripting.c perf script: Move perf_sample__sprintf_flags to
trace-event-scripting.c
perf x86: Define arch_fetch_insn in NO_AUXTRACE builds perf intel-pt: Remove stale build comment
perf env: Move arch errno function to only use in env perf lock: Move common lock contention code to new file perf bench: Remove reference to cmd_inject
perf kwork: Make perf_kwork_add_work a callback
perf build: Remove test library from python shared object perf python: Add parse_events function
perf python: Add __str__ and __repr__ functions to evlist perf python: Add __str__ and __repr__ functions to evsel perf python: Correctly throw IndexError

 tools/perf/Makefile.perf                      |   7 +-
 tools/perf/arch/x86/util/Build                |   2 +-
 tools/perf/bench/inject-buildid.c             |  13 +-
 tools/perf/builtin-kvm.c                      |  61 ----
 tools/perf/builtin-kwork.c                    |   3 +-
 tools/perf/builtin-lock.c                     | 137 +------
 tools/perf/builtin-script.c                   | 304 +---------------
 tools/perf/builtin-stat.c                     |  27 --
 tools/perf/builtin-trace.c                    |   1 -
 tools/perf/builtin.h                          |   6 -
 .../scripts/python/Perf-Trace-Util/Context.c  |  20 +-
 tools/perf/tests/stat.c                       |  16 +-
 tools/perf/trace/beauty/arch_errno_names.sh   |   3 +-
 tools/perf/ui/browsers/scripts.c              | 177 ++++++++-
 tools/perf/util/Build                         |   4 +-
 tools/perf/util/bpf_kwork.c                   |   2 +-
 tools/perf/util/bpf_kwork_top.c               |   2 +-
 tools/perf/util/bpf_lock_contention.c         |   2 +-
 tools/perf/util/cgroup.c                      |   2 +-
 tools/perf/util/config.c                      |  27 ++
 tools/perf/util/dlfilter.c                    |   3 +-
 tools/perf/util/env.c                         |   4 +
 tools/perf/util/env.h                         |   2 -
 tools/perf/util/evsel.c                       |  19 +-
 tools/perf/util/evsel.h                       |   2 +-
 tools/perf/util/intel-pt-decoder/Build        |  18 +-
 .../intel-pt-decoder/intel-pt-insn-decoder.c  |   3 -
 tools/perf/util/kvm-stat.c                    |  70 ++++
 tools/perf/util/kvm-stat.h                    |   3 +
 tools/perf/util/kwork.h                       |   6 +-
 tools/perf/util/lock-contention.c             | 170 +++++++++
 tools/perf/util/lock-contention.h             |  37 +-
 tools/perf/util/path.c                        |  10 +
 tools/perf/util/path.h                        |   1 +
 tools/perf/util/python.c                      | 341 ++++++++----------
 .../scripting-engines/trace-event-python.c    |  63 +---
 tools/perf/util/stat.h                        |   3 +-
 tools/perf/util/trace-event-scripting.c       | 177 +++++++++
 tools/perf/util/trace-event.h                 |   5 +-
 39 files changed, 872 insertions(+), 881 deletions(-)
 create mode 100644 tools/perf/util/kvm-stat.c
 create mode 100644 tools/perf/util/lock-contention.c
Signature · 1 line
2.47.0.277.g8800431eea-goog
author reviewer
#1 lore raw

PATCH v6 1/22 perf python: Remove python 2 scripting support

Python2 was deprecated 4 years ago, remove support and workarounds.

---
 .../scripts/python/Perf-Trace-Util/Context.c  | 18 -----
 tools/perf/util/python.c                      | 73 +++----------------
 .../scripting-engines/trace-event-python.c    | 63 +---------------
 3 files changed, 15 insertions(+), 139 deletions(-)
tools/perf/scripts/python/Perf-Trace-Util/Context.c +0 -18
index 01f54d6724a5..d742daaa5d5a 100644
@@ -24,16 +24,6 @@
 #include "../../../util/srcline.h"
 #include "../../../util/srccode.h"
 
-#if PY_MAJOR_VERSION < 3
-#define _PyCapsule_GetPointer(arg1, arg2) \
-  PyCObject_AsVoidPtr(arg1)
-#define _PyBytes_FromStringAndSize(arg1, arg2) \
-  PyString_FromStringAndSize((arg1), (arg2))
-#define _PyUnicode_AsUTF8(arg) \
-  PyString_AsString(arg)
-
-PyMODINIT_FUNC initperf_trace_context(void);
-#else
 #define _PyCapsule_GetPointer(arg1, arg2) \
   PyCapsule_GetPointer((arg1), (arg2))
 #define _PyBytes_FromStringAndSize(arg1, arg2) \
@@ -42,7 +32,6 @@ PyMODINIT_FUNC initperf_trace_context(void);
   PyUnicode_AsUTF8(arg)
 
 PyMODINIT_FUNC PyInit_perf_trace_context(void);
-#endif
 
 static struct scripting_context *get_args(PyObject *args, const char *name, PyObject **arg2)
 {
@@ -213,12 +202,6 @@ static PyMethodDef ContextMethods[] = {
 	{ NULL, NULL, 0, NULL}
 };
 
-#if PY_MAJOR_VERSION < 3
-PyMODINIT_FUNC initperf_trace_context(void)
-{
-	(void) Py_InitModule("perf_trace_context", ContextMethods);
-}
-#else
 PyMODINIT_FUNC PyInit_perf_trace_context(void)
 {
 	static struct PyModuleDef moduledef = {
@@ -240,4 +223,3 @@ PyMODINIT_FUNC PyInit_perf_trace_context(void)
 
 	return mod;
 }
-#endif
tools/perf/util/python.c +12 -61

193 lines not on this page · raw message

tools/perf/util/scripting-engines/trace-event-python.c +3 -60

113 lines not on this page · raw message

Signature · 1 line
2.47.0.277.g8800431eea-goog
author reviewer
#2 lore raw

PATCH v6 2/22 perf python: Constify variables and parameters

In reply to Ian Rogers

Opportunistically constify variables and parameters when possible.

---
 tools/perf/util/python.c | 55 +++++++++++++++++++++-------------------
 1 file changed, 29 insertions(+), 26 deletions(-)
tools/perf/util/python.c +29 -26
index 6fc283ac5498..945fb7bee0af 100644
@@ -63,7 +63,7 @@ struct pyrf_event {
 	sample_member_def(sample_period, period, T_ULONGLONG, "event period"),		 \
 	sample_member_def(sample_cpu, cpu, T_UINT, "event cpu"),
 
-static char pyrf_mmap_event__doc[] = PyDoc_STR("perf mmap event object.");
+static const char pyrf_mmap_event__doc[] = PyDoc_STR("perf mmap event object.");
 
 static PyMemberDef pyrf_mmap_event__members[] = {
 	sample_members
@@ -78,7 +78,7 @@ static PyMemberDef pyrf_mmap_event__members[] = {
 	{ .name = NULL, },
 };
 
-static PyObject *pyrf_mmap_event__repr(struct pyrf_event *pevent)
+static PyObject *pyrf_mmap_event__repr(const struct pyrf_event *pevent)
 {
 	PyObject *ret;
 	char *s;
@@ -107,7 +107,7 @@ static PyTypeObject pyrf_mmap_event__type = {
 	.tp_repr	= (reprfunc)pyrf_mmap_event__repr,
 };
 
-static char pyrf_task_event__doc[] = PyDoc_STR("perf task (fork/exit) event object.");
+static const char pyrf_task_event__doc[] = PyDoc_STR("perf task (fork/exit) event object.");
 
 static PyMemberDef pyrf_task_event__members[] = {
 	sample_members
@@ -120,7 +120,7 @@ static PyMemberDef pyrf_task_event__members[] = {
 	{ .name = NULL, },
 };
 
-static PyObject *pyrf_task_event__repr(struct pyrf_event *pevent)
+static PyObject *pyrf_task_event__repr(const struct pyrf_event *pevent)
 {
 	return PyUnicode_FromFormat("{ type: %s, pid: %u, ppid: %u, tid: %u, "
 				   "ptid: %u, time: %" PRI_lu64 "}",
@@ -142,7 +142,7 @@ static PyTypeObject pyrf_task_event__type = {
 	.tp_repr	= (reprfunc)pyrf_task_event__repr,
 };
 
-static char pyrf_comm_event__doc[] = PyDoc_STR("perf comm event object.");
+static const char pyrf_comm_event__doc[] = PyDoc_STR("perf comm event object.");
 
 static PyMemberDef pyrf_comm_event__members[] = {
 	sample_members
@@ -153,7 +153,7 @@ static PyMemberDef pyrf_comm_event__members[] = {
 	{ .name = NULL, },
 };
 
-static PyObject *pyrf_comm_event__repr(struct pyrf_event *pevent)
+static PyObject *pyrf_comm_event__repr(const struct pyrf_event *pevent)
 {
 	return PyUnicode_FromFormat("{ type: comm, pid: %u, tid: %u, comm: %s }",
 				   pevent->event.comm.pid,
@@ -171,7 +171,7 @@ static PyTypeObject pyrf_comm_event__type = {
 	.tp_repr	= (reprfunc)pyrf_comm_event__repr,
 };
 
-static char pyrf_throttle_event__doc[] = PyDoc_STR("perf throttle event object.");
+static const char pyrf_throttle_event__doc[] = PyDoc_STR("perf throttle event object.");
 
 static PyMemberDef pyrf_throttle_event__members[] = {
 	sample_members
@@ -182,9 +182,10 @@ static PyMemberDef pyrf_throttle_event__members[] = {
 	{ .name = NULL, },
 };
 
-static PyObject *pyrf_throttle_event__repr(struct pyrf_event *pevent)
+static PyObject *pyrf_throttle_event__repr(const struct pyrf_event *pevent)
 {
-	struct perf_record_throttle *te = (struct perf_record_throttle *)(&pevent->event.header + 1);
+	const struct perf_record_throttle *te = (const struct perf_record_throttle *)
+		(&pevent->event.header + 1);
 
 	return PyUnicode_FromFormat("{ type: %sthrottle, time: %" PRI_lu64 ", id: %" PRI_lu64
 				   ", stream_id: %" PRI_lu64 " }",
@@ -202,7 +203,7 @@ static PyTypeObject pyrf_throttle_event__type = {
 	.tp_repr	= (reprfunc)pyrf_throttle_event__repr,
 };
 
-static char pyrf_lost_event__doc[] = PyDoc_STR("perf lost event object.");
+static const char pyrf_lost_event__doc[] = PyDoc_STR("perf lost event object.");
 
 static PyMemberDef pyrf_lost_event__members[] = {
 	sample_members
@@ -211,7 +212,7 @@ static PyMemberDef pyrf_lost_event__members[] = {
 	{ .name = NULL, },
 };
 
-static PyObject *pyrf_lost_event__repr(struct pyrf_event *pevent)
+static PyObject *pyrf_lost_event__repr(const struct pyrf_event *pevent)
 {
 	PyObject *ret;
 	char *s;
@@ -237,7 +238,7 @@ static PyTypeObject pyrf_lost_event__type = {
 	.tp_repr	= (reprfunc)pyrf_lost_event__repr,
 };
 
-static char pyrf_read_event__doc[] = PyDoc_STR("perf read event object.");
+static const char pyrf_read_event__doc[] = PyDoc_STR("perf read event object.");
 
 static PyMemberDef pyrf_read_event__members[] = {
 	sample_members
@@ -246,7 +247,7 @@ static PyMemberDef pyrf_read_event__members[] = {
 	{ .name = NULL, },
 };
 
-static PyObject *pyrf_read_event__repr(struct pyrf_event *pevent)
+static PyObject *pyrf_read_event__repr(const struct pyrf_event *pevent)
 {
 	return PyUnicode_FromFormat("{ type: read, pid: %u, tid: %u }",
 				   pevent->event.read.pid,
@@ -267,7 +268,7 @@ static PyTypeObject pyrf_read_event__type = {
 	.tp_repr	= (reprfunc)pyrf_read_event__repr,
 };
 
-static char pyrf_sample_event__doc[] = PyDoc_STR("perf sample event object.");
+static const char pyrf_sample_event__doc[] = PyDoc_STR("perf sample event object.");
 
 static PyMemberDef pyrf_sample_event__members[] = {
 	sample_members
@@ -275,7 +276,7 @@ static PyMemberDef pyrf_sample_event__members[] = {
 	{ .name = NULL, },
 };
 
-static PyObject *pyrf_sample_event__repr(struct pyrf_event *pevent)
+static PyObject *pyrf_sample_event__repr(const struct pyrf_event *pevent)
 {
 	PyObject *ret;
 	char *s;
@@ -290,13 +291,13 @@ static PyObject *pyrf_sample_event__repr(struct pyrf_event *pevent)
 }
 
 #ifdef HAVE_LIBTRACEEVENT
-static bool is_tracepoint(struct pyrf_event *pevent)
+static bool is_tracepoint(const struct pyrf_event *pevent)
 {
 	return pevent->evsel->core.attr.type == PERF_TYPE_TRACEPOINT;
 }
 
 static PyObject*
-tracepoint_field(struct pyrf_event *pe, struct tep_format_field *field)
+tracepoint_field(const struct pyrf_event *pe, struct tep_format_field *field)
 {
 	struct tep_handle *pevent = field->event->tep;
 	void *data = pe->sample.raw_data;
@@ -385,7 +386,7 @@ static PyTypeObject pyrf_sample_event__type = {
 	.tp_getattro	= (getattrofunc) pyrf_sample_event__getattro,
 };
 
-static char pyrf_context_switch_event__doc[] = PyDoc_STR("perf context_switch event object.");
+static const char pyrf_context_switch_event__doc[] = PyDoc_STR("perf context_switch event object.");
 
 static PyMemberDef pyrf_context_switch_event__members[] = {
 	sample_members
@@ -395,7 +396,7 @@ static PyMemberDef pyrf_context_switch_event__members[] = {
 	{ .name = NULL, },
 };
 
-static PyObject *pyrf_context_switch_event__repr(struct pyrf_event *pevent)
+static PyObject *pyrf_context_switch_event__repr(const struct pyrf_event *pevent)
 {
 	PyObject *ret;
 	char *s;
@@ -475,7 +476,7 @@ static PyTypeObject *pyrf_event__type[] = {
 	[PERF_RECORD_SWITCH_CPU_WIDE]  = &pyrf_context_switch_event__type,
 };
 
-static PyObject *pyrf_event__new(union perf_event *event)
+static PyObject *pyrf_event__new(const union perf_event *event)
 {
 	struct pyrf_event *pevent;
 	PyTypeObject *ptype;
@@ -543,7 +544,7 @@ static PySequenceMethods pyrf_cpu_map__sequence_methods = {
 	.sq_item   = pyrf_cpu_map__item,
 };
 
-static char pyrf_cpu_map__doc[] = PyDoc_STR("cpu map object.");
+static const char pyrf_cpu_map__doc[] = PyDoc_STR("cpu map object.");
 
 static PyTypeObject pyrf_cpu_map__type = {
 	PyVarObject_HEAD_INIT(NULL, 0)
@@ -612,7 +613,7 @@ static PySequenceMethods pyrf_thread_map__sequence_methods = {
 	.sq_item   = pyrf_thread_map__item,
 };
 
-static char pyrf_thread_map__doc[] = PyDoc_STR("thread map object.");
+static const char pyrf_thread_map__doc[] = PyDoc_STR("thread map object.");
 
 static PyTypeObject pyrf_thread_map__type = {
 	PyVarObject_HEAD_INIT(NULL, 0)
@@ -796,7 +797,7 @@ static PyMethodDef pyrf_evsel__methods[] = {
 	{ .ml_name = NULL, }
 };
 
-static char pyrf_evsel__doc[] = PyDoc_STR("perf event selector list object.");
+static const char pyrf_evsel__doc[] = PyDoc_STR("perf event selector list object.");
 
 static PyTypeObject pyrf_evsel__type = {
 	PyVarObject_HEAD_INIT(NULL, 0)
@@ -1079,7 +1080,7 @@ static PySequenceMethods pyrf_evlist__sequence_methods = {
 	.sq_item   = pyrf_evlist__item,
 };
 
-static char pyrf_evlist__doc[] = PyDoc_STR("perf event selector list object.");
+static const char pyrf_evlist__doc[] = PyDoc_STR("perf event selector list object.");
 
 static PyTypeObject pyrf_evlist__type = {
 	PyVarObject_HEAD_INIT(NULL, 0)
@@ -1101,10 +1102,12 @@ static int pyrf_evlist__setup_types(void)
 
 #define PERF_CONST(name) { #name, PERF_##name }
 
-static struct {
+struct perf_constant {
 	const char *name;
 	int	    value;
-} perf__constants[] = {
+};
+
+static const struct perf_constant perf__constants[] = {
 	PERF_CONST(TYPE_HARDWARE),
 	PERF_CONST(TYPE_SOFTWARE),
 	PERF_CONST(TYPE_TRACEPOINT),
Signature · 1 line
2.47.0.277.g8800431eea-goog
author reviewer
#3 lore raw

PATCH v6 3/22 perf python: Remove unused #include

In reply to Ian Rogers

Remove unused #include of bpf-filter.h.

---
 tools/perf/util/python.c | 1 -
 1 file changed, 1 deletion(-)
tools/perf/util/python.c +0 -1
index 945fb7bee0af..e2109f6c7ead 100644
@@ -16,7 +16,6 @@
 #include "thread_map.h"
 #include "trace-event.h"
 #include "mmap.h"
-#include "util/bpf-filter.h"
 #include "util/env.h"
 #include "util/kvm-stat.h"
 #include "util/stat.h"
Signature · 1 line
2.47.0.277.g8800431eea-goog
author reviewer
#4 lore raw

PATCH v6 4/22 perf script: Move scripting_max_stack out of builtin

In reply to Ian Rogers

scripting_max_stack is used in util code which is linked into the python module. Move the variable declaration to util/trace-event-scripting.c to avoid conditional compilation.

---
 tools/perf/builtin-script.c             | 2 --
 tools/perf/util/python.c                | 2 --
 tools/perf/util/trace-event-scripting.c | 3 +++
 3 files changed, 3 insertions(+), 4 deletions(-)
tools/perf/builtin-script.c +0 -2
index 6b6d4472db6e..5d5a1a06d8c6 100644
@@ -92,8 +92,6 @@ static struct dlfilter		*dlfilter;
 static int			dlargc;
 static char			**dlargv;
 
-unsigned int scripting_max_stack = PERF_MAX_STACK_DEPTH;
-
 enum perf_output_field {
 	PERF_OUTPUT_COMM            = 1ULL << 0,
 	PERF_OUTPUT_TID             = 1ULL << 1,
tools/perf/util/python.c +0 -2
index e2109f6c7ead..f82e8a968e24 100644
@@ -1307,8 +1307,6 @@ PyMODINIT_FUNC PyInit_perf(void)
 /* The following are stubs to avoid dragging in builtin-* objects. */
 /* TODO: move the code out of the builtin-* file into util. */
 
-unsigned int scripting_max_stack = PERF_MAX_STACK_DEPTH;
-
 #ifdef HAVE_KVM_STAT_SUPPORT
 bool kvm_entry_event(struct evsel *evsel __maybe_unused)
 {
tools/perf/util/trace-event-scripting.c +3 -0
index bd0000300c77..8abb7a7b6888 100644
@@ -16,9 +16,12 @@
 #include "debug.h"
 #include "trace-event.h"
 #include "evsel.h"
+#include <linux/perf_event.h>
 #include <linux/zalloc.h>
 #include "util/sample.h"
 
+unsigned int scripting_max_stack = PERF_MAX_STACK_DEPTH;
+
 struct scripting_context *scripting_context;
 
 void scripting_context__update(struct scripting_context *c,
Signature · 1 line
2.47.0.277.g8800431eea-goog
author reviewer
#5 lore raw

PATCH v6 5/22 perf kvm: Move functions used in util out of builtin

In reply to Ian Rogers

The util library code is used by the python module but doesn't have access to the builtin files. Make a util/kvm-stat.c to match the kvm-stat.h file that declares the functions and move the functions there.

---
 tools/perf/builtin-kvm.c   | 61 ---------------------------------
 tools/perf/util/Build      |  1 +
 tools/perf/util/kvm-stat.c | 70 ++++++++++++++++++++++++++++++++++++++
 tools/perf/util/kvm-stat.h |  3 ++
 tools/perf/util/python.c   | 32 -----------------
 5 files changed, 74 insertions(+), 93 deletions(-)
 create mode 100644 tools/perf/util/kvm-stat.c
tools/perf/builtin-kvm.c +0 -61
index 099ce3ebf67c..e50cd04e3725 100644
@@ -615,67 +615,6 @@ static const char *get_filename_for_perf_kvm(void)
 
 #if defined(HAVE_KVM_STAT_SUPPORT) && defined(HAVE_LIBTRACEEVENT)
 
-void exit_event_get_key(struct evsel *evsel,
-			struct perf_sample *sample,
-			struct event_key *key)
-{
-	key->info = 0;
-	key->key  = evsel__intval(evsel, sample, kvm_exit_reason);
-}
-
-bool kvm_exit_event(struct evsel *evsel)
-{
-	return evsel__name_is(evsel, kvm_exit_trace);
-}
-
-bool exit_event_begin(struct evsel *evsel,
-		      struct perf_sample *sample, struct event_key *key)
-{
-	if (kvm_exit_event(evsel)) {
-		exit_event_get_key(evsel, sample, key);
-		return true;
-	}
-
-	return false;
-}
-
-bool kvm_entry_event(struct evsel *evsel)
-{
-	return evsel__name_is(evsel, kvm_entry_trace);
-}
-
-bool exit_event_end(struct evsel *evsel,
-		    struct perf_sample *sample __maybe_unused,
-		    struct event_key *key __maybe_unused)
-{
-	return kvm_entry_event(evsel);
-}
-
-static const char *get_exit_reason(struct perf_kvm_stat *kvm,
-				   struct exit_reasons_table *tbl,
-				   u64 exit_code)
-{
-	while (tbl->reason != NULL) {
-		if (tbl->exit_code == exit_code)
-			return tbl->reason;
-		tbl++;
-	}
-
-	pr_err("unknown kvm exit code:%lld on %s\n",
-		(unsigned long long)exit_code, kvm->exit_reasons_isa);
-	return "UNKNOWN";
-}
-
-void exit_event_decode_key(struct perf_kvm_stat *kvm,
-			   struct event_key *key,
-			   char *decode)
-{
-	const char *exit_reason = get_exit_reason(kvm, key->exit_reasons,
-						  key->key);
-
-	scnprintf(decode, KVM_EVENT_NAME_LEN, "%s", exit_reason);
-}
-
 static bool register_kvm_events_ops(struct perf_kvm_stat *kvm)
 {
 	struct kvm_reg_events_ops *events_ops = kvm_reg_events_ops;
tools/perf/util/Build +1 -0
index 1eedead5f2f2..650974413849 100644
@@ -120,6 +120,7 @@ perf-util-y += spark.o
 perf-util-y += topdown.o
 perf-util-y += iostat.o
 perf-util-y += stream.o
+perf-util-y += kvm-stat.o
 perf-util-$(CONFIG_AUXTRACE) += auxtrace.o
 perf-util-$(CONFIG_AUXTRACE) += intel-pt-decoder/
 perf-util-$(CONFIG_AUXTRACE) += intel-pt.o
tools/perf/util/kvm-stat.c new file +70 -0

70 lines not on this page · raw message

tools/perf/util/kvm-stat.h +3 -0
index 3e9ac754c3d1..4249542544bb 100644
@@ -115,6 +115,8 @@ struct kvm_reg_events_ops {
 	struct kvm_events_ops *ops;
 };
 
+#if defined(HAVE_KVM_STAT_SUPPORT) && defined(HAVE_LIBTRACEEVENT)
+
 void exit_event_get_key(struct evsel *evsel,
 			struct perf_sample *sample,
 			struct event_key *key);
@@ -127,6 +129,7 @@ bool exit_event_end(struct evsel *evsel,
 void exit_event_decode_key(struct perf_kvm_stat *kvm,
 			   struct event_key *key,
 			   char *decode);
+#endif
 
 bool kvm_exit_event(struct evsel *evsel);
 bool kvm_entry_event(struct evsel *evsel);
tools/perf/util/python.c +0 -32
index f82e8a968e24..94902652e371 100644
@@ -1307,38 +1307,6 @@ PyMODINIT_FUNC PyInit_perf(void)
 /* The following are stubs to avoid dragging in builtin-* objects. */
 /* TODO: move the code out of the builtin-* file into util. */
 
-#ifdef HAVE_KVM_STAT_SUPPORT
-bool kvm_entry_event(struct evsel *evsel __maybe_unused)
-{
-	return false;
-}
-
-bool kvm_exit_event(struct evsel *evsel __maybe_unused)
-{
-	return false;
-}
-
-bool exit_event_begin(struct evsel *evsel __maybe_unused,
-		      struct perf_sample *sample  __maybe_unused,
-		      struct event_key *key  __maybe_unused)
-{
-	return false;
-}
-
-bool exit_event_end(struct evsel *evsel __maybe_unused,
-		    struct perf_sample *sample __maybe_unused,
-		    struct event_key *key __maybe_unused)
-{
-	return false;
-}
-
-void exit_event_decode_key(struct perf_kvm_stat *kvm __maybe_unused,
-			   struct event_key *key __maybe_unused,
-			   char *decode __maybe_unused)
-{
-}
-#endif // HAVE_KVM_STAT_SUPPORT
-
 int find_scripts(char **scripts_array  __maybe_unused, char **scripts_path_array  __maybe_unused,
 		int num  __maybe_unused, int pathlen __maybe_unused)
 {
Signature · 1 line
2.47.0.277.g8800431eea-goog
author reviewer
#6 lore raw

PATCH v6 6/22 perf script: Use openat for directory iteration

In reply to Ian Rogers

Rewrite the directory iteration to use openat so that large character arrays aren't needed. The arrays are warned about potential buffer overflows by GCC when the code exists in a single C file.

---
 tools/perf/builtin-script.c | 87 +++++++++++++++++++++++++------------
 tools/perf/util/path.c      | 10 +++++
 tools/perf/util/path.h      |  1 +
 3 files changed, 71 insertions(+), 27 deletions(-)
tools/perf/builtin-script.c +60 -27
index 5d5a1a06d8c6..e20d55b8a741 100644
@@ -3530,27 +3530,35 @@ static void free_dlarg(void)
  * which is covered well now. And new parsing code should be added to
  * cover the future complex formats like event groups etc.
  */
-static int check_ev_match(char *dir_name, char *scriptname,
-			struct perf_session *session)
+static int check_ev_match(int dir_fd, const char *scriptname, struct perf_session *session)
 {
-	char filename[MAXPATHLEN], evname[128];
-	char line[BUFSIZ], *p;
-	struct evsel *pos;
-	int match, len;
+	char line[BUFSIZ];
 	FILE *fp;
 
-	scnprintf(filename, MAXPATHLEN, "%s/bin/%s-record", dir_name, scriptname);
+	{
+		char filename[FILENAME_MAX + 5];
+		int fd;
 
-	fp = fopen(filename, "r");
-	if (!fp)
-		return -1;
+		scnprintf(filename, sizeof(filename), "bin/%s-record", scriptname);
+		fd = openat(dir_fd, filename, O_RDONLY);
+		if (fd == -1)
+			return -1;
+		fp = fdopen(fd, "r");
+		if (!fp)
+			return -1;
+	}
 
 	while (fgets(line, sizeof(line), fp)) {
-		p = skip_spaces(line);
+		char *p = skip_spaces(line);
+
 		if (*p == '#')
 			continue;
 
 		while (strlen(p)) {
+			int match, len;
+			struct evsel *pos;
+			char evname[128];
+
 			p = strstr(p, "-e");
 			if (!p)
 				break;
@@ -3593,7 +3601,7 @@ int find_scripts(char **scripts_array, char **scripts_path_array, int num,
 		 int pathlen)
 {
 	struct dirent *script_dirent, *lang_dirent;
-	char scripts_path[MAXPATHLEN], lang_path[MAXPATHLEN];
+	int scripts_dir_fd, lang_dir_fd;
 	DIR *scripts_dir, *lang_dir;
 	struct perf_session *session;
 	struct perf_data data = {
@@ -3602,51 +3610,76 @@ int find_scripts(char **scripts_array, char **scripts_path_array, int num,
 	};
 	char *temp;
 	int i = 0;
+	const char *exec_path = get_argv_exec_path();
 
 	session = perf_session__new(&data, NULL);
 	if (IS_ERR(session))
 		return PTR_ERR(session);
 
-	snprintf(scripts_path, MAXPATHLEN, "%s/scripts", get_argv_exec_path());
+	{
+		char scripts_path[PATH_MAX];
 
-	scripts_dir = opendir(scripts_path);
+		snprintf(scripts_path, sizeof(scripts_path), "%s/scripts", exec_path);
+		scripts_dir_fd = open(scripts_path, O_DIRECTORY);
+		pr_err("Failed to open directory '%s'", scripts_path);
+		if (scripts_dir_fd == -1) {
+			perf_session__delete(session);
+			return -1;
+		}
+	}
+	scripts_dir = fdopendir(scripts_dir_fd);
 	if (!scripts_dir) {
+		close(scripts_dir_fd);
 		perf_session__delete(session);
 		return -1;
 	}
 
-	for_each_lang(scripts_path, scripts_dir, lang_dirent) {
-		scnprintf(lang_path, MAXPATHLEN, "%s/%s", scripts_path,
-			  lang_dirent->d_name);
+	while ((lang_dirent = readdir(scripts_dir)) != NULL) {
+		if (lang_dirent->d_type != DT_DIR &&
+		    (lang_dirent->d_type == DT_UNKNOWN &&
+		     !is_directory_at(scripts_dir_fd, lang_dirent->d_name)))
+			continue;
+		if (!strcmp(lang_dirent->d_name, ".") || !strcmp(lang_dirent->d_name, ".."))
+			continue;
+
 #ifndef HAVE_LIBPERL_SUPPORT
-		if (strstr(lang_path, "perl"))
+		if (strstr(lang_dirent->d_name, "perl"))
 			continue;
 #endif
 #ifndef HAVE_LIBPYTHON_SUPPORT
-		if (strstr(lang_path, "python"))
+		if (strstr(lang_dirent->d_name, "python"))
 			continue;
 #endif
 
-		lang_dir = opendir(lang_path);
-		if (!lang_dir)
+		lang_dir_fd = openat(scripts_dir_fd, lang_dirent->d_name, O_DIRECTORY);
+		if (lang_dir_fd == -1)
 			continue;
-
-		for_each_script(lang_path, lang_dir, script_dirent) {
+		lang_dir = fdopendir(lang_dir_fd);
+		if (!lang_dir) {
+			close(lang_dir_fd);
+			continue;
+		}
+		while ((script_dirent = readdir(lang_dir)) != NULL) {
+			if (script_dirent->d_type == DT_DIR)
+				continue;
+			if (script_dirent->d_type == DT_UNKNOWN &&
+			    is_directory_at(lang_dir_fd, script_dirent->d_name))
+				continue;
 			/* Skip those real time scripts: xxxtop.p[yl] */
 			if (strstr(script_dirent->d_name, "top."))
 				continue;
 			if (i >= num)
 				break;
-			snprintf(scripts_path_array[i], pathlen, "%s/%s",
-				lang_path,
+			scnprintf(scripts_path_array[i], pathlen, "%s/scripts/%s/%s",
+				exec_path,
+				lang_dirent->d_name,
 				script_dirent->d_name);
 			temp = strchr(script_dirent->d_name, '.');
 			snprintf(scripts_array[i],
 				(temp - script_dirent->d_name) + 1,
 				"%s", script_dirent->d_name);
 
-			if (check_ev_match(lang_path,
-					scripts_array[i], session))
+			if (check_ev_match(lang_dir_fd, scripts_array[i], session))
 				continue;
 
 			i++;
tools/perf/util/path.c +10 -0
index 00adf872bf00..9712466c51e2 100644
@@ -68,6 +68,16 @@ bool is_directory(const char *base_path, const struct dirent *dent)
 	return S_ISDIR(st.st_mode);
 }
 
+bool is_directory_at(int dir_fd, const char *path)
+{
+	struct stat st;
+
+	if (fstatat(dir_fd, path, &st, /*flags=*/0))
+		return false;
+
+	return S_ISDIR(st.st_mode);
+}
+
 bool is_executable_file(const char *base_path, const struct dirent *dent)
 {
 	char path[PATH_MAX];
tools/perf/util/path.h +1 -0
index d94902c22222..fbafbe7015dd 100644
@@ -12,6 +12,7 @@ int path__join3(char *bf, size_t size, const char *path1, const char *path2, con
 
 bool is_regular_file(const char *file);
 bool is_directory(const char *base_path, const struct dirent *dent);
+bool is_directory_at(int dir_fd, const char *path);
 bool is_executable_file(const char *base_path, const struct dirent *dent);
 
 #endif /* _PERF_PATH_H */
Signature · 1 line
2.47.0.277.g8800431eea-goog
maintainer Acked-by on v6 6/22
#7 lore raw

Re: [PATCH v6 06/22] perf script: Use openat for directory iteration

Quoted 5 lines · Ian Rogers

Rewrite the directory iteration to use openat so that large character arrays aren't needed. The arrays are warned about potential buffer overflows by GCC when the code exists in a single C file.

Thanks for doing this!
Namhyung

Quoted 194 lines
---
 tools/perf/builtin-script.c | 87 +++++++++++++++++++++++++------------
 tools/perf/util/path.c      | 10 +++++
 tools/perf/util/path.h      |  1 +
 3 files changed, 71 insertions(+), 27 deletions(-)
tools/perf/builtin-script.c +60 -27
index 5d5a1a06d8c6..e20d55b8a741 100644
@@ -3530,27 +3530,35 @@ static void free_dlarg(void)
  * which is covered well now. And new parsing code should be added to
  * cover the future complex formats like event groups etc.
  */
-static int check_ev_match(char *dir_name, char *scriptname,
-			struct perf_session *session)
+static int check_ev_match(int dir_fd, const char *scriptname, struct perf_session *session)
 {
-	char filename[MAXPATHLEN], evname[128];
-	char line[BUFSIZ], *p;
-	struct evsel *pos;
-	int match, len;
+	char line[BUFSIZ];
 	FILE *fp;
 
-	scnprintf(filename, MAXPATHLEN, "%s/bin/%s-record", dir_name, scriptname);
+	{
+		char filename[FILENAME_MAX + 5];
+		int fd;
 
-	fp = fopen(filename, "r");
-	if (!fp)
-		return -1;
+		scnprintf(filename, sizeof(filename), "bin/%s-record", scriptname);
+		fd = openat(dir_fd, filename, O_RDONLY);
+		if (fd == -1)
+			return -1;
+		fp = fdopen(fd, "r");
+		if (!fp)
+			return -1;
+	}
 
 	while (fgets(line, sizeof(line), fp)) {
-		p = skip_spaces(line);
+		char *p = skip_spaces(line);
+
 		if (*p == '#')
 			continue;
 
 		while (strlen(p)) {
+			int match, len;
+			struct evsel *pos;
+			char evname[128];
+
 			p = strstr(p, "-e");
 			if (!p)
 				break;
@@ -3593,7 +3601,7 @@ int find_scripts(char **scripts_array, char **scripts_path_array, int num,
 		 int pathlen)
 {
 	struct dirent *script_dirent, *lang_dirent;
-	char scripts_path[MAXPATHLEN], lang_path[MAXPATHLEN];
+	int scripts_dir_fd, lang_dir_fd;
 	DIR *scripts_dir, *lang_dir;
 	struct perf_session *session;
 	struct perf_data data = {
@@ -3602,51 +3610,76 @@ int find_scripts(char **scripts_array, char **scripts_path_array, int num,
 	};
 	char *temp;
 	int i = 0;
+	const char *exec_path = get_argv_exec_path();
 
 	session = perf_session__new(&data, NULL);
 	if (IS_ERR(session))
 		return PTR_ERR(session);
 
-	snprintf(scripts_path, MAXPATHLEN, "%s/scripts", get_argv_exec_path());
+	{
+		char scripts_path[PATH_MAX];
 
-	scripts_dir = opendir(scripts_path);
+		snprintf(scripts_path, sizeof(scripts_path), "%s/scripts", exec_path);
+		scripts_dir_fd = open(scripts_path, O_DIRECTORY);
+		pr_err("Failed to open directory '%s'", scripts_path);
+		if (scripts_dir_fd == -1) {
+			perf_session__delete(session);
+			return -1;
+		}
+	}
+	scripts_dir = fdopendir(scripts_dir_fd);
 	if (!scripts_dir) {
+		close(scripts_dir_fd);
 		perf_session__delete(session);
 		return -1;
 	}
 
-	for_each_lang(scripts_path, scripts_dir, lang_dirent) {
-		scnprintf(lang_path, MAXPATHLEN, "%s/%s", scripts_path,
-			  lang_dirent->d_name);
+	while ((lang_dirent = readdir(scripts_dir)) != NULL) {
+		if (lang_dirent->d_type != DT_DIR &&
+		    (lang_dirent->d_type == DT_UNKNOWN &&
+		     !is_directory_at(scripts_dir_fd, lang_dirent->d_name)))
+			continue;
+		if (!strcmp(lang_dirent->d_name, ".") || !strcmp(lang_dirent->d_name, ".."))
+			continue;
+
 #ifndef HAVE_LIBPERL_SUPPORT
-		if (strstr(lang_path, "perl"))
+		if (strstr(lang_dirent->d_name, "perl"))
 			continue;
 #endif
 #ifndef HAVE_LIBPYTHON_SUPPORT
-		if (strstr(lang_path, "python"))
+		if (strstr(lang_dirent->d_name, "python"))
 			continue;
 #endif
 
-		lang_dir = opendir(lang_path);
-		if (!lang_dir)
+		lang_dir_fd = openat(scripts_dir_fd, lang_dirent->d_name, O_DIRECTORY);
+		if (lang_dir_fd == -1)
 			continue;
-
-		for_each_script(lang_path, lang_dir, script_dirent) {
+		lang_dir = fdopendir(lang_dir_fd);
+		if (!lang_dir) {
+			close(lang_dir_fd);
+			continue;
+		}
+		while ((script_dirent = readdir(lang_dir)) != NULL) {
+			if (script_dirent->d_type == DT_DIR)
+				continue;
+			if (script_dirent->d_type == DT_UNKNOWN &&
+			    is_directory_at(lang_dir_fd, script_dirent->d_name))
+				continue;
 			/* Skip those real time scripts: xxxtop.p[yl] */
 			if (strstr(script_dirent->d_name, "top."))
 				continue;
 			if (i >= num)
 				break;
-			snprintf(scripts_path_array[i], pathlen, "%s/%s",
-				lang_path,
+			scnprintf(scripts_path_array[i], pathlen, "%s/scripts/%s/%s",
+				exec_path,
+				lang_dirent->d_name,
 				script_dirent->d_name);
 			temp = strchr(script_dirent->d_name, '.');
 			snprintf(scripts_array[i],
 				(temp - script_dirent->d_name) + 1,
 				"%s", script_dirent->d_name);
 
-			if (check_ev_match(lang_path,
-					scripts_array[i], session))
+			if (check_ev_match(lang_dir_fd, scripts_array[i], session))
 				continue;
 
 			i++;
tools/perf/util/path.c +10 -0
index 00adf872bf00..9712466c51e2 100644
@@ -68,6 +68,16 @@ bool is_directory(const char *base_path, const struct dirent *dent)
 	return S_ISDIR(st.st_mode);
 }
 
+bool is_directory_at(int dir_fd, const char *path)
+{
+	struct stat st;
+
+	if (fstatat(dir_fd, path, &st, /*flags=*/0))
+		return false;
+
+	return S_ISDIR(st.st_mode);
+}
+
 bool is_executable_file(const char *base_path, const struct dirent *dent)
 {
 	char path[PATH_MAX];
tools/perf/util/path.h +1 -0
index d94902c22222..fbafbe7015dd 100644
@@ -12,6 +12,7 @@ int path__join3(char *bf, size_t size, const char *path1, const char *path2, con
 
 bool is_regular_file(const char *file);
 bool is_directory(const char *base_path, const struct dirent *dent);
+bool is_directory_at(int dir_fd, const char *path);
 bool is_executable_file(const char *base_path, const struct dirent *dent);
 
 #endif /* _PERF_PATH_H */
Signature · 1 line
2.47.0.277.g8800431eea-goog
author reviewer
#8 lore raw

PATCH v6 7/22 perf script: Move find_scripts to browser/scripts.c

In reply to Ian Rogers

The only use of find_scripts is in browser/scripts.c but the definition in builtin causes linking problems requiring a stub in python.c. Move the function to allow the stub to be removed.

---
 tools/perf/builtin-script.c      | 171 -----------------------------
 tools/perf/builtin.h             |   6 --
 tools/perf/ui/browsers/scripts.c | 177 ++++++++++++++++++++++++++++++-
 tools/perf/util/python.c         |   6 --
 4 files changed, 175 insertions(+), 185 deletions(-)
tools/perf/builtin-script.c +0 -171
index e20d55b8a741..e9ec74056f71 100644
@@ -3521,177 +3521,6 @@ static void free_dlarg(void)
 	free(dlargv);
 }
 
-/*
- * Some scripts specify the required events in their "xxx-record" file,
- * this function will check if the events in perf.data match those
- * mentioned in the "xxx-record".
- *
- * Fixme: All existing "xxx-record" are all in good formats "-e event ",
- * which is covered well now. And new parsing code should be added to
- * cover the future complex formats like event groups etc.
- */
-static int check_ev_match(int dir_fd, const char *scriptname, struct perf_session *session)
-{
-	char line[BUFSIZ];
-	FILE *fp;
-
-	{
-		char filename[FILENAME_MAX + 5];
-		int fd;
-
-		scnprintf(filename, sizeof(filename), "bin/%s-record", scriptname);
-		fd = openat(dir_fd, filename, O_RDONLY);
-		if (fd == -1)
-			return -1;
-		fp = fdopen(fd, "r");
-		if (!fp)
-			return -1;
-	}
-
-	while (fgets(line, sizeof(line), fp)) {
-		char *p = skip_spaces(line);
-
-		if (*p == '#')
-			continue;
-
-		while (strlen(p)) {
-			int match, len;
-			struct evsel *pos;
-			char evname[128];
-
-			p = strstr(p, "-e");
-			if (!p)
-				break;
-
-			p += 2;
-			p = skip_spaces(p);
-			len = strcspn(p, " \t");
-			if (!len)
-				break;
-
-			snprintf(evname, len + 1, "%s", p);
-
-			match = 0;
-			evlist__for_each_entry(session->evlist, pos) {
-				if (evsel__name_is(pos, evname)) {
-					match = 1;
-					break;
-				}
-			}
-
-			if (!match) {
-				fclose(fp);
-				return -1;
-			}
-		}
-	}
-
-	fclose(fp);
-	return 0;
-}
-
-/*
- * Return -1 if none is found, otherwise the actual scripts number.
- *
- * Currently the only user of this function is the script browser, which
- * will list all statically runnable scripts, select one, execute it and
- * show the output in a perf browser.
- */
-int find_scripts(char **scripts_array, char **scripts_path_array, int num,
-		 int pathlen)
-{
-	struct dirent *script_dirent, *lang_dirent;
-	int scripts_dir_fd, lang_dir_fd;
-	DIR *scripts_dir, *lang_dir;
-	struct perf_session *session;
-	struct perf_data data = {
-		.path = input_name,
-		.mode = PERF_DATA_MODE_READ,
-	};
-	char *temp;
-	int i = 0;
-	const char *exec_path = get_argv_exec_path();
-
-	session = perf_session__new(&data, NULL);
-	if (IS_ERR(session))
-		return PTR_ERR(session);
-
-	{
-		char scripts_path[PATH_MAX];
-
-		snprintf(scripts_path, sizeof(scripts_path), "%s/scripts", exec_path);
-		scripts_dir_fd = open(scripts_path, O_DIRECTORY);
-		pr_err("Failed to open directory '%s'", scripts_path);
-		if (scripts_dir_fd == -1) {
-			perf_session__delete(session);
-			return -1;
-		}
-	}
-	scripts_dir = fdopendir(scripts_dir_fd);
-	if (!scripts_dir) {
-		close(scripts_dir_fd);
-		perf_session__delete(session);
-		return -1;
-	}
-
-	while ((lang_dirent = readdir(scripts_dir)) != NULL) {
-		if (lang_dirent->d_type != DT_DIR &&
-		    (lang_dirent->d_type == DT_UNKNOWN &&
-		     !is_directory_at(scripts_dir_fd, lang_dirent->d_name)))
-			continue;
-		if (!strcmp(lang_dirent->d_name, ".") || !strcmp(lang_dirent->d_name, ".."))
-			continue;
-
-#ifndef HAVE_LIBPERL_SUPPORT
-		if (strstr(lang_dirent->d_name, "perl"))
-			continue;
-#endif
-#ifndef HAVE_LIBPYTHON_SUPPORT
-		if (strstr(lang_dirent->d_name, "python"))
-			continue;
-#endif
-
-		lang_dir_fd = openat(scripts_dir_fd, lang_dirent->d_name, O_DIRECTORY);
-		if (lang_dir_fd == -1)
-			continue;
-		lang_dir = fdopendir(lang_dir_fd);
-		if (!lang_dir) {
-			close(lang_dir_fd);
-			continue;
-		}
-		while ((script_dirent = readdir(lang_dir)) != NULL) {
-			if (script_dirent->d_type == DT_DIR)
-				continue;
-			if (script_dirent->d_type == DT_UNKNOWN &&
-			    is_directory_at(lang_dir_fd, script_dirent->d_name))
-				continue;
-			/* Skip those real time scripts: xxxtop.p[yl] */
-			if (strstr(script_dirent->d_name, "top."))
-				continue;
-			if (i >= num)
-				break;
-			scnprintf(scripts_path_array[i], pathlen, "%s/scripts/%s/%s",
-				exec_path,
-				lang_dirent->d_name,
-				script_dirent->d_name);
-			temp = strchr(script_dirent->d_name, '.');
-			snprintf(scripts_array[i],
-				(temp - script_dirent->d_name) + 1,
-				"%s", script_dirent->d_name);
-
-			if (check_ev_match(lang_dir_fd, scripts_array[i], session))
-				continue;
-
-			i++;
-		}
-		closedir(lang_dir);
-	}
-
-	closedir(scripts_dir);
-	perf_session__delete(session);
-	return i;
-}
-
 static char *get_script_path(const char *script_root, const char *suffix)
 {
 	struct dirent *script_dirent, *lang_dirent;
tools/perf/builtin.h +0 -6
index 94f4b3769bf7..a07e93c53848 100644
@@ -2,10 +2,6 @@
 #ifndef BUILTIN_H
 #define BUILTIN_H
 
-#include <stddef.h>
-#include <linux/compiler.h>
-#include <tools/config.h>
-
 struct feature_status {
 	const char *name;
 	const char *macro;
@@ -56,6 +52,4 @@ int cmd_ftrace(int argc, const char **argv);
 int cmd_daemon(int argc, const char **argv);
 int cmd_kwork(int argc, const char **argv);
 
-int find_scripts(char **scripts_array, char **scripts_path_array, int num,
-		 int pathlen);
 #endif
tools/perf/ui/browsers/scripts.c +175 -2

197 lines not on this page · raw message

tools/perf/util/python.c +0 -6
index 94902652e371..eb15f3b6c4f5 100644
@@ -1307,12 +1307,6 @@ PyMODINIT_FUNC PyInit_perf(void)
 /* The following are stubs to avoid dragging in builtin-* objects. */
 /* TODO: move the code out of the builtin-* file into util. */
 
-int find_scripts(char **scripts_array  __maybe_unused, char **scripts_path_array  __maybe_unused,
-		int num  __maybe_unused, int pathlen __maybe_unused)
-{
-	return -1;
-}
-
 void perf_stat__set_no_csv_summary(int set __maybe_unused)
 {
 }
Signature · 1 line
2.47.0.277.g8800431eea-goog
maintainer Acked-by on v6 7/22
#9 lore raw

Re: [PATCH v6 07/22] perf script: Move find_scripts to browser/scripts.c

Quoted 5 lines · Ian Rogers

The only use of find_scripts is in browser/scripts.c but the definition in builtin causes linking problems requiring a stub in python.c. Move the function to allow the stub to be removed.

Thanks,
Namhyung

Quoted 434 lines
---
 tools/perf/builtin-script.c      | 171 -----------------------------
 tools/perf/builtin.h             |   6 --
 tools/perf/ui/browsers/scripts.c | 177 ++++++++++++++++++++++++++++++-
 tools/perf/util/python.c         |   6 --
 4 files changed, 175 insertions(+), 185 deletions(-)
tools/perf/builtin-script.c +0 -171
index e20d55b8a741..e9ec74056f71 100644
@@ -3521,177 +3521,6 @@ static void free_dlarg(void)
 	free(dlargv);
 }
 
-/*
- * Some scripts specify the required events in their "xxx-record" file,
- * this function will check if the events in perf.data match those
- * mentioned in the "xxx-record".
- *
- * Fixme: All existing "xxx-record" are all in good formats "-e event ",
- * which is covered well now. And new parsing code should be added to
- * cover the future complex formats like event groups etc.
- */
-static int check_ev_match(int dir_fd, const char *scriptname, struct perf_session *session)
-{
-	char line[BUFSIZ];
-	FILE *fp;
-
-	{
-		char filename[FILENAME_MAX + 5];
-		int fd;
-
-		scnprintf(filename, sizeof(filename), "bin/%s-record", scriptname);
-		fd = openat(dir_fd, filename, O_RDONLY);
-		if (fd == -1)
-			return -1;
-		fp = fdopen(fd, "r");
-		if (!fp)
-			return -1;
-	}
-
-	while (fgets(line, sizeof(line), fp)) {
-		char *p = skip_spaces(line);
-
-		if (*p == '#')
-			continue;
-
-		while (strlen(p)) {
-			int match, len;
-			struct evsel *pos;
-			char evname[128];
-
-			p = strstr(p, "-e");
-			if (!p)
-				break;
-
-			p += 2;
-			p = skip_spaces(p);
-			len = strcspn(p, " \t");
-			if (!len)
-				break;
-
-			snprintf(evname, len + 1, "%s", p);
-
-			match = 0;
-			evlist__for_each_entry(session->evlist, pos) {
-				if (evsel__name_is(pos, evname)) {
-					match = 1;
-					break;
-				}
-			}
-
-			if (!match) {
-				fclose(fp);
-				return -1;
-			}
-		}
-	}
-
-	fclose(fp);
-	return 0;
-}
-
-/*
- * Return -1 if none is found, otherwise the actual scripts number.
- *
- * Currently the only user of this function is the script browser, which
- * will list all statically runnable scripts, select one, execute it and
- * show the output in a perf browser.
- */
-int find_scripts(char **scripts_array, char **scripts_path_array, int num,
-		 int pathlen)
-{
-	struct dirent *script_dirent, *lang_dirent;
-	int scripts_dir_fd, lang_dir_fd;
-	DIR *scripts_dir, *lang_dir;
-	struct perf_session *session;
-	struct perf_data data = {
-		.path = input_name,
-		.mode = PERF_DATA_MODE_READ,
-	};
-	char *temp;
-	int i = 0;
-	const char *exec_path = get_argv_exec_path();
-
-	session = perf_session__new(&data, NULL);
-	if (IS_ERR(session))
-		return PTR_ERR(session);
-
-	{
-		char scripts_path[PATH_MAX];
-
-		snprintf(scripts_path, sizeof(scripts_path), "%s/scripts", exec_path);
-		scripts_dir_fd = open(scripts_path, O_DIRECTORY);
-		pr_err("Failed to open directory '%s'", scripts_path);
-		if (scripts_dir_fd == -1) {
-			perf_session__delete(session);
-			return -1;
-		}
-	}
-	scripts_dir = fdopendir(scripts_dir_fd);
-	if (!scripts_dir) {
-		close(scripts_dir_fd);
-		perf_session__delete(session);
-		return -1;
-	}
-
-	while ((lang_dirent = readdir(scripts_dir)) != NULL) {
-		if (lang_dirent->d_type != DT_DIR &&
-		    (lang_dirent->d_type == DT_UNKNOWN &&
-		     !is_directory_at(scripts_dir_fd, lang_dirent->d_name)))
-			continue;
-		if (!strcmp(lang_dirent->d_name, ".") || !strcmp(lang_dirent->d_name, ".."))
-			continue;
-
-#ifndef HAVE_LIBPERL_SUPPORT
-		if (strstr(lang_dirent->d_name, "perl"))
-			continue;
-#endif
-#ifndef HAVE_LIBPYTHON_SUPPORT
-		if (strstr(lang_dirent->d_name, "python"))
-			continue;
-#endif
-
-		lang_dir_fd = openat(scripts_dir_fd, lang_dirent->d_name, O_DIRECTORY);
-		if (lang_dir_fd == -1)
-			continue;
-		lang_dir = fdopendir(lang_dir_fd);
-		if (!lang_dir) {
-			close(lang_dir_fd);
-			continue;
-		}
-		while ((script_dirent = readdir(lang_dir)) != NULL) {
-			if (script_dirent->d_type == DT_DIR)
-				continue;
-			if (script_dirent->d_type == DT_UNKNOWN &&
-			    is_directory_at(lang_dir_fd, script_dirent->d_name))
-				continue;
-			/* Skip those real time scripts: xxxtop.p[yl] */
-			if (strstr(script_dirent->d_name, "top."))
-				continue;
-			if (i >= num)
-				break;
-			scnprintf(scripts_path_array[i], pathlen, "%s/scripts/%s/%s",
-				exec_path,
-				lang_dirent->d_name,
-				script_dirent->d_name);
-			temp = strchr(script_dirent->d_name, '.');
-			snprintf(scripts_array[i],
-				(temp - script_dirent->d_name) + 1,
-				"%s", script_dirent->d_name);
-
-			if (check_ev_match(lang_dir_fd, scripts_array[i], session))
-				continue;
-
-			i++;
-		}
-		closedir(lang_dir);
-	}
-
-	closedir(scripts_dir);
-	perf_session__delete(session);
-	return i;
-}
-
 static char *get_script_path(const char *script_root, const char *suffix)
 {
 	struct dirent *script_dirent, *lang_dirent;
tools/perf/builtin.h +0 -6
index 94f4b3769bf7..a07e93c53848 100644
@@ -2,10 +2,6 @@
 #ifndef BUILTIN_H
 #define BUILTIN_H
 
-#include <stddef.h>
-#include <linux/compiler.h>
-#include <tools/config.h>
-
 struct feature_status {
 	const char *name;
 	const char *macro;
@@ -56,6 +52,4 @@ int cmd_ftrace(int argc, const char **argv);
 int cmd_daemon(int argc, const char **argv);
 int cmd_kwork(int argc, const char **argv);
 
-int find_scripts(char **scripts_array, char **scripts_path_array, int num,
-		 int pathlen);
 #endif
tools/perf/ui/browsers/scripts.c +175 -2
index e437d7889de6..2d04ece833aa 100644
@@ -1,16 +1,18 @@
 // SPDX-License-Identifier: GPL-2.0
-#include "../../builtin.h"
-#include "../../perf.h"
 #include "../../util/util.h" // perf_exe()
 #include "../util.h"
+#include "../../util/evlist.h"
 #include "../../util/hist.h"
 #include "../../util/debug.h"
+#include "../../util/session.h"
 #include "../../util/symbol.h"
 #include "../browser.h"
 #include "../libslang.h"
 #include "config.h"
+#include <linux/err.h>
 #include <linux/string.h>
 #include <linux/zalloc.h>
+#include <subcmd/exec-cmd.h>
 #include <stdlib.h>
 
 #define SCRIPT_NAMELEN	128
@@ -77,6 +79,177 @@ static int scripts_config(const char *var, const char *value, void *data)
 	return 0;
 }
 
+/*
+ * Some scripts specify the required events in their "xxx-record" file,
+ * this function will check if the events in perf.data match those
+ * mentioned in the "xxx-record".
+ *
+ * Fixme: All existing "xxx-record" are all in good formats "-e event ",
+ * which is covered well now. And new parsing code should be added to
+ * cover the future complex formats like event groups etc.
+ */
+static int check_ev_match(int dir_fd, const char *scriptname, struct perf_session *session)
+{
+	char line[BUFSIZ];
+	FILE *fp;
+
+	{
+		char filename[FILENAME_MAX + 5];
+		int fd;
+
+		scnprintf(filename, sizeof(filename), "bin/%s-record", scriptname);
+		fd = openat(dir_fd, filename, O_RDONLY);
+		if (fd == -1)
+			return -1;
+		fp = fdopen(fd, "r");
+		if (!fp)
+			return -1;
+	}
+
+	while (fgets(line, sizeof(line), fp)) {
+		char *p = skip_spaces(line);
+
+		if (*p == '#')
+			continue;
+
+		while (strlen(p)) {
+			int match, len;
+			struct evsel *pos;
+			char evname[128];
+
+			p = strstr(p, "-e");
+			if (!p)
+				break;
+
+			p += 2;
+			p = skip_spaces(p);
+			len = strcspn(p, " \t");
+			if (!len)
+				break;
+
+			snprintf(evname, len + 1, "%s", p);
+
+			match = 0;
+			evlist__for_each_entry(session->evlist, pos) {
+				if (evsel__name_is(pos, evname)) {
+					match = 1;
+					break;
+				}
+			}
+
+			if (!match) {
+				fclose(fp);
+				return -1;
+			}
+		}
+	}
+
+	fclose(fp);
+	return 0;
+}
+
+/*
+ * Return -1 if none is found, otherwise the actual scripts number.
+ *
+ * Currently the only user of this function is the script browser, which
+ * will list all statically runnable scripts, select one, execute it and
+ * show the output in a perf browser.
+ */
+static int find_scripts(char **scripts_array, char **scripts_path_array, int num,
+		 int pathlen)
+{
+	struct dirent *script_dirent, *lang_dirent;
+	int scripts_dir_fd, lang_dir_fd;
+	DIR *scripts_dir, *lang_dir;
+	struct perf_session *session;
+	struct perf_data data = {
+		.path = input_name,
+		.mode = PERF_DATA_MODE_READ,
+	};
+	char *temp;
+	int i = 0;
+	const char *exec_path = get_argv_exec_path();
+
+	session = perf_session__new(&data, NULL);
+	if (IS_ERR(session))
+		return PTR_ERR(session);
+
+	{
+		char scripts_path[PATH_MAX];
+
+		snprintf(scripts_path, sizeof(scripts_path), "%s/scripts", exec_path);
+		scripts_dir_fd = open(scripts_path, O_DIRECTORY);
+		pr_err("Failed to open directory '%s'", scripts_path);
+		if (scripts_dir_fd == -1) {
+			perf_session__delete(session);
+			return -1;
+		}
+	}
+	scripts_dir = fdopendir(scripts_dir_fd);
+	if (!scripts_dir) {
+		close(scripts_dir_fd);
+		perf_session__delete(session);
+		return -1;
+	}
+
+	while ((lang_dirent = readdir(scripts_dir)) != NULL) {
+		if (lang_dirent->d_type != DT_DIR &&
+		    (lang_dirent->d_type == DT_UNKNOWN &&
+		     !is_directory_at(scripts_dir_fd, lang_dirent->d_name)))
+			continue;
+		if (!strcmp(lang_dirent->d_name, ".") || !strcmp(lang_dirent->d_name, ".."))
+			continue;
+
+#ifndef HAVE_LIBPERL_SUPPORT
+		if (strstr(lang_dirent->d_name, "perl"))
+			continue;
+#endif
+#ifndef HAVE_LIBPYTHON_SUPPORT
+		if (strstr(lang_dirent->d_name, "python"))
+			continue;
+#endif
+
+		lang_dir_fd = openat(scripts_dir_fd, lang_dirent->d_name, O_DIRECTORY);
+		if (lang_dir_fd == -1)
+			continue;
+		lang_dir = fdopendir(lang_dir_fd);
+		if (!lang_dir) {
+			close(lang_dir_fd);
+			continue;
+		}
+		while ((script_dirent = readdir(lang_dir)) != NULL) {
+			if (script_dirent->d_type == DT_DIR)
+				continue;
+			if (script_dirent->d_type == DT_UNKNOWN &&
+			    is_directory_at(lang_dir_fd, script_dirent->d_name))
+				continue;
+			/* Skip those real time scripts: xxxtop.p[yl] */
+			if (strstr(script_dirent->d_name, "top."))
+				continue;
+			if (i >= num)
+				break;
+			scnprintf(scripts_path_array[i], pathlen, "%s/scripts/%s/%s",
+				exec_path,
+				lang_dirent->d_name,
+				script_dirent->d_name);
+			temp = strchr(script_dirent->d_name, '.');
+			snprintf(scripts_array[i],
+				(temp - script_dirent->d_name) + 1,
+				"%s", script_dirent->d_name);
+
+			if (check_ev_match(lang_dir_fd, scripts_array[i], session))
+				continue;
+
+			i++;
+		}
+		closedir(lang_dir);
+	}
+
+	closedir(scripts_dir);
+	perf_session__delete(session);
+	return i;
+}
+
 /*
  * When success, will copy the full path of the selected script
  * into  the buffer pointed by script_name, and return 0.
tools/perf/util/python.c +0 -6
index 94902652e371..eb15f3b6c4f5 100644
@@ -1307,12 +1307,6 @@ PyMODINIT_FUNC PyInit_perf(void)
 /* The following are stubs to avoid dragging in builtin-* objects. */
 /* TODO: move the code out of the builtin-* file into util. */
 
-int find_scripts(char **scripts_array  __maybe_unused, char **scripts_path_array  __maybe_unused,
-		int num  __maybe_unused, int pathlen __maybe_unused)
-{
-	return -1;
-}
-
 void perf_stat__set_no_csv_summary(int set __maybe_unused)
 {
 }
Signature · 1 line
2.47.0.277.g8800431eea-goog
author reviewer
#10 lore raw

PATCH v6 8/22 perf stat: Move stat_config into config.c

In reply to Ian Rogers

stat_config is accessed by config.c via helper functions, but declared in builtin-stat. Move to util/config.c so that stub functions aren't needed in python.c which doesn't link against the builtin files.

To avoid name conflicts change builtin-script to use the same stat_config as builtin-stat. Rename local variables in tests to avoid shadow declaration warnings.

---
 tools/perf/builtin-script.c |  1 -
 tools/perf/builtin-stat.c   | 27 ---------------------------
 tools/perf/tests/stat.c     | 16 +++++++++-------
 tools/perf/util/config.c    | 27 +++++++++++++++++++++++++++
 tools/perf/util/python.c    | 10 ----------
 tools/perf/util/stat.h      |  3 ++-
 6 files changed, 38 insertions(+), 46 deletions(-)
tools/perf/builtin-script.c +0 -1
index e9ec74056f71..62e851fdf5ca 100644
@@ -85,7 +85,6 @@ static bool			system_wide;
 static bool			print_flags;
 static const char		*cpu_list;
 static DECLARE_BITMAP(cpu_bitmap, MAX_NR_CPUS);
-static struct perf_stat_config	stat_config;
 static int			max_blocks;
 static bool			native_arch;
 static struct dlfilter		*dlfilter;
tools/perf/builtin-stat.c +0 -27
index fdf5172646a5..77e327d4a9a7 100644
@@ -112,8 +112,6 @@ static struct target target = {
 	.uid	= UINT_MAX,
 };
 
-#define METRIC_ONLY_LEN 20
-
 static volatile sig_atomic_t	child_pid			= -1;
 static int			detailed_run			=  0;
 static bool			transaction_run;
@@ -151,21 +149,6 @@ static struct perf_stat		perf_stat;
 
 static volatile sig_atomic_t done = 0;
 
-static struct perf_stat_config stat_config = {
-	.aggr_mode		= AGGR_GLOBAL,
-	.aggr_level		= MAX_CACHE_LVL + 1,
-	.scale			= true,
-	.unit_width		= 4, /* strlen("unit") */
-	.run_count		= 1,
-	.metric_only_len	= METRIC_ONLY_LEN,
-	.walltime_nsecs_stats	= &walltime_nsecs_stats,
-	.ru_stats		= &ru_stats,
-	.big_num		= true,
-	.ctl_fd			= -1,
-	.ctl_fd_ack		= -1,
-	.iostat_run		= false,
-};
-
 /* Options set from the command line. */
 struct opt_aggr_mode {
 	bool node, socket, die, cluster, cache, core, thread, no_aggr;
@@ -1071,16 +1054,6 @@ static void sig_atexit(void)
 	kill(getpid(), signr);
 }
 
-void perf_stat__set_big_num(int set)
-{
-	stat_config.big_num = (set != 0);
-}
-
-void perf_stat__set_no_csv_summary(int set)
-{
-	stat_config.no_csv_summary = (set != 0);
-}
-
 static int stat__set_big_num(const struct option *opt __maybe_unused,
 			     const char *s __maybe_unused, int unset)
 {
tools/perf/tests/stat.c +9 -7
index 6468cc0d0204..d60983657bad 100644
@@ -27,7 +27,7 @@ static int process_stat_config_event(const struct perf_tool *tool __maybe_unused
 				     struct machine *machine __maybe_unused)
 {
 	struct perf_record_stat_config *config = &event->stat_config;
-	struct perf_stat_config stat_config = {};
+	struct perf_stat_config test_stat_config = {};
 
 #define HAS(term, val) \
 	has_term(config, PERF_STAT_CONFIG_TERM__##term, val)
@@ -39,25 +39,27 @@ static int process_stat_config_event(const struct perf_tool *tool __maybe_unused
 
 #undef HAS
 
-	perf_event__read_stat_config(&stat_config, config);
+	perf_event__read_stat_config(&test_stat_config, config);
 
-	TEST_ASSERT_VAL("wrong aggr_mode", stat_config.aggr_mode == AGGR_CORE);
-	TEST_ASSERT_VAL("wrong scale",     stat_config.scale == 1);
-	TEST_ASSERT_VAL("wrong interval",  stat_config.interval == 1);
+	TEST_ASSERT_VAL("wrong aggr_mode", test_stat_config.aggr_mode == AGGR_CORE);
+	TEST_ASSERT_VAL("wrong scale",     test_stat_config.scale == 1);
+	TEST_ASSERT_VAL("wrong interval",  test_stat_config.interval == 1);
 	return 0;
 }
 
 static int test__synthesize_stat_config(struct test_suite *test __maybe_unused,
 					int subtest __maybe_unused)
 {
-	struct perf_stat_config stat_config = {
+	struct perf_stat_config test_stat_config = {
 		.aggr_mode	= AGGR_CORE,
 		.scale		= 1,
 		.interval	= 1,
 	};
 
 	TEST_ASSERT_VAL("failed to synthesize stat_config",
-		!perf_event__synthesize_stat_config(NULL, &stat_config, process_stat_config_event, NULL));
+		!perf_event__synthesize_stat_config(NULL, &test_stat_config,
+						    process_stat_config_event,
+						    NULL));
 
 	return 0;
 }
tools/perf/util/config.c +27 -0
index 68f9407ca74b..2d07c9257a1a 100644
@@ -13,6 +13,7 @@
 #include <sys/param.h>
 #include "cache.h"
 #include "callchain.h"
+#include "header.h"
 #include <subcmd/exec-cmd.h>
 #include "util/event.h"  /* proc_map_timeout */
 #include "util/hist.h"  /* perf_hist_config */
@@ -34,6 +35,22 @@
 
 #define DEBUG_CACHE_DIR ".debug"
 
+#define METRIC_ONLY_LEN 20
+
+struct perf_stat_config stat_config = {
+	.aggr_mode		= AGGR_GLOBAL,
+	.aggr_level		= MAX_CACHE_LVL + 1,
+	.scale			= true,
+	.unit_width		= 4, /* strlen("unit") */
+	.run_count		= 1,
+	.metric_only_len	= METRIC_ONLY_LEN,
+	.walltime_nsecs_stats	= &walltime_nsecs_stats,
+	.ru_stats		= &ru_stats,
+	.big_num		= true,
+	.ctl_fd			= -1,
+	.ctl_fd_ack		= -1,
+	.iostat_run		= false,
+};
 
 char buildid_dir[MAXPATHLEN]; /* root dir for buildid, binary cache */
 
@@ -455,6 +472,16 @@ static int perf_ui_config(const char *var, const char *value)
 	return 0;
 }
 
+void perf_stat__set_big_num(int set)
+{
+	stat_config.big_num = (set != 0);
+}
+
+static void perf_stat__set_no_csv_summary(int set)
+{
+	stat_config.no_csv_summary = (set != 0);
+}
+
 static int perf_stat_config(const char *var, const char *value)
 {
 	if (!strcmp(var, "stat.big-num"))
tools/perf/util/python.c +0 -10
index eb15f3b6c4f5..0fa8e27769be 100644
@@ -17,8 +17,6 @@
 #include "trace-event.h"
 #include "mmap.h"
 #include "util/env.h"
-#include "util/kvm-stat.h"
-#include "util/stat.h"
 #include "util/kwork.h"
 #include "util/sample.h"
 #include "util/lock-contention.h"
@@ -1307,14 +1305,6 @@ PyMODINIT_FUNC PyInit_perf(void)
 /* The following are stubs to avoid dragging in builtin-* objects. */
 /* TODO: move the code out of the builtin-* file into util. */
 
-void perf_stat__set_no_csv_summary(int set __maybe_unused)
-{
-}
-
-void perf_stat__set_big_num(int set __maybe_unused)
-{
-}
-
 int script_spec_register(const char *spec __maybe_unused, struct scripting_ops *ops __maybe_unused)
 {
 	return -1;
tools/perf/util/stat.h +2 -1
index 6f8cff3cd39a..2fda9acd7374 100644
@@ -117,8 +117,9 @@ struct perf_stat_config {
 	unsigned int		topdown_level;
 };
 
+extern struct perf_stat_config stat_config;
+
 void perf_stat__set_big_num(int set);
-void perf_stat__set_no_csv_summary(int set);
 
 void update_stats(struct stats *stats, u64 val);
 double avg_stats(struct stats *stats);
Signature · 1 line
2.47.0.277.g8800431eea-goog
author reviewer
#11 lore raw

PATCH v6 9/22 perf script: Move script_spec code to trace-event-scripting.c

In reply to Ian Rogers

The script_spec code is referenced in util/trace-event-scripting but the list was in builtin-script, accessed via a function that required a stub function in python.c. Move all the logic to trace-event-scripting, with lookup and foreach functions exposed for builtin-script's benefit.

---
 tools/perf/builtin-script.c             | 67 +---------------------
 tools/perf/util/python.c                |  5 --
 tools/perf/util/trace-event-scripting.c | 75 +++++++++++++++++++++++++
 tools/perf/util/trace-event.h           |  3 +-
 4 files changed, 80 insertions(+), 70 deletions(-)
tools/perf/builtin-script.c +3 -64
index 62e851fdf5ca..11c0ee8c1afc 100644
@@ -2956,79 +2956,18 @@ static int __cmd_script(struct perf_script *script)
 	return ret;
 }
 
-struct script_spec {
-	struct list_head	node;
-	struct scripting_ops	*ops;
-	char			spec[];
-};
-
-static LIST_HEAD(script_specs);
-
-static struct script_spec *script_spec__new(const char *spec,
-					    struct scripting_ops *ops)
+static int list_available_languages_cb(struct scripting_ops *ops, const char *spec)
 {
-	struct script_spec *s = malloc(sizeof(*s) + strlen(spec) + 1);
-
-	if (s != NULL) {
-		strcpy(s->spec, spec);
-		s->ops = ops;
-	}
-
-	return s;
-}
-
-static void script_spec__add(struct script_spec *s)
-{
-	list_add_tail(&s->node, &script_specs);
-}
-
-static struct script_spec *script_spec__find(const char *spec)
-{
-	struct script_spec *s;
-
-	list_for_each_entry(s, &script_specs, node)
-		if (strcasecmp(s->spec, spec) == 0)
-			return s;
-	return NULL;
-}
-
-int script_spec_register(const char *spec, struct scripting_ops *ops)
-{
-	struct script_spec *s;
-
-	s = script_spec__find(spec);
-	if (s)
-		return -1;
-
-	s = script_spec__new(spec, ops);
-	if (!s)
-		return -1;
-	else
-		script_spec__add(s);
-
+	fprintf(stderr, "  %-42s [%s]\n", spec, ops->name);
 	return 0;
 }
 
-static struct scripting_ops *script_spec__lookup(const char *spec)
-{
-	struct script_spec *s = script_spec__find(spec);
-	if (!s)
-		return NULL;
-
-	return s->ops;
-}
-
 static void list_available_languages(void)
 {
-	struct script_spec *s;
-
 	fprintf(stderr, "\n");
 	fprintf(stderr, "Scripting language extensions (used in "
 		"perf script -s [spec:]script.[spec]):\n\n");
-
-	list_for_each_entry(s, &script_specs, node)
-		fprintf(stderr, "  %-42s [%s]\n", s->spec, s->ops->name);
-
+	script_spec__for_each(&list_available_languages_cb);
 	fprintf(stderr, "\n");
 }
 
tools/perf/util/python.c +0 -5
index 0fa8e27769be..3d938fe2de6a 100644
@@ -1305,11 +1305,6 @@ PyMODINIT_FUNC PyInit_perf(void)
 /* The following are stubs to avoid dragging in builtin-* objects. */
 /* TODO: move the code out of the builtin-* file into util. */
 
-int script_spec_register(const char *spec __maybe_unused, struct scripting_ops *ops __maybe_unused)
-{
-	return -1;
-}
-
 arch_syscalls__strerrno_t *arch_syscalls__strerrno_function(const char *arch __maybe_unused)
 {
 	return NULL;
tools/perf/util/trace-event-scripting.c +75 -0

81 lines not on this page · raw message

tools/perf/util/trace-event.h +2 -1
index 0e5133f1b910..0520dd269374 100644
@@ -113,7 +113,8 @@ struct scripting_ops {
 
 extern unsigned int scripting_max_stack;
 
-int script_spec_register(const char *spec, struct scripting_ops *ops);
+struct scripting_ops *script_spec__lookup(const char *spec);
+int script_spec__for_each(int (*cb)(struct scripting_ops *ops, const char *spec));
 
 void script_fetch_insn(struct perf_sample *sample, struct thread *thread,
 		       struct machine *machine);
Signature · 1 line
2.47.0.277.g8800431eea-goog
author reviewer
#12 lore raw

PATCH v6 10/22 perf script: Move script_fetch_insn to trace-event-scripting.c

In reply to Ian Rogers

Add native_arch as a parameter to script_fetch_insn rather than relying on the builtin-script value that won't be initialized for the dlfilter and python Context use cases. Assume both of those cases are running natively.

---
 tools/perf/builtin-script.c                       | 15 +--------------
 .../perf/scripts/python/Perf-Trace-Util/Context.c |  2 +-
 tools/perf/util/dlfilter.c                        |  3 ++-
 tools/perf/util/python.c                          |  6 ------
 tools/perf/util/trace-event-scripting.c           | 14 ++++++++++++++
 tools/perf/util/trace-event.h                     |  2 +-
 6 files changed, 19 insertions(+), 23 deletions(-)
tools/perf/builtin-script.c +1 -14
index 11c0ee8c1afc..22d78a9d8f27 100644
@@ -1586,19 +1586,6 @@ static int perf_sample__fprintf_callindent(struct perf_sample *sample,
 	return len + dlen;
 }
 
-__weak void arch_fetch_insn(struct perf_sample *sample __maybe_unused,
-			    struct thread *thread __maybe_unused,
-			    struct machine *machine __maybe_unused)
-{
-}
-
-void script_fetch_insn(struct perf_sample *sample, struct thread *thread,
-		       struct machine *machine)
-{
-	if (sample->insn_len == 0 && native_arch)
-		arch_fetch_insn(sample, thread, machine);
-}
-
 static int perf_sample__fprintf_insn(struct perf_sample *sample,
 				     struct evsel *evsel,
 				     struct perf_event_attr *attr,
@@ -1608,7 +1595,7 @@ static int perf_sample__fprintf_insn(struct perf_sample *sample,
 {
 	int printed = 0;
 
-	script_fetch_insn(sample, thread, machine);
+	script_fetch_insn(sample, thread, machine, native_arch);
 
 	if (PRINT_FIELD(INSNLEN))
 		printed += fprintf(fp, " ilen: %d", sample->insn_len);
tools/perf/scripts/python/Perf-Trace-Util/Context.c +1 -1
index d742daaa5d5a..60dcfe56d4d9 100644
@@ -93,7 +93,7 @@ static PyObject *perf_sample_insn(PyObject *obj, PyObject *args)
 	if (c->sample->ip && !c->sample->insn_len && thread__maps(c->al->thread)) {
 		struct machine *machine =  maps__machine(thread__maps(c->al->thread));
 
-		script_fetch_insn(c->sample, c->al->thread, machine);
+		script_fetch_insn(c->sample, c->al->thread, machine, /*native_arch=*/true);
 	}
 	if (!c->sample->insn_len)
 		Py_RETURN_NONE; /* N.B. This is a return statement */
tools/perf/util/dlfilter.c +2 -1
index 7d180bdaedbc..ddacef881af2 100644
@@ -234,7 +234,8 @@ static const __u8 *dlfilter__insn(void *ctx, __u32 *len)
 			struct machine *machine = maps__machine(thread__maps(al->thread));
 
 			if (machine)
-				script_fetch_insn(d->sample, al->thread, machine);
+				script_fetch_insn(d->sample, al->thread, machine,
+						  /*native_arch=*/true);
 		}
 	}
 
tools/perf/util/python.c +0 -6
index 3d938fe2de6a..22edadd64e5f 100644
@@ -1317,12 +1317,6 @@ struct kwork_work *perf_kwork_add_work(struct perf_kwork *kwork __maybe_unused,
 	return NULL;
 }
 
-void script_fetch_insn(struct perf_sample *sample __maybe_unused,
-		struct thread *thread __maybe_unused,
-		struct machine *machine __maybe_unused)
-{
-}
-
 int perf_sample__sprintf_flags(u32 flags __maybe_unused, char *str __maybe_unused,
 			size_t sz __maybe_unused)
 {
tools/perf/util/trace-event-scripting.c +14 -0
index ad62d8e5a368..beac456260ae 100644
@@ -13,6 +13,7 @@
 #include <traceevent/event-parse.h>
 #endif
 
+#include "archinsn.h"
 #include "debug.h"
 #include "trace-event.h"
 #include "evsel.h"
@@ -269,3 +270,16 @@ void setup_perl_scripting(void)
 }
 #endif
 #endif
+
+__weak void arch_fetch_insn(struct perf_sample *sample __maybe_unused,
+		     struct thread *thread __maybe_unused,
+		     struct machine *machine __maybe_unused)
+{
+}
+
+void script_fetch_insn(struct perf_sample *sample, struct thread *thread,
+		       struct machine *machine, bool native_arch)
+{
+	if (sample->insn_len == 0 && native_arch)
+		arch_fetch_insn(sample, thread, machine);
+}
tools/perf/util/trace-event.h +1 -1
index 0520dd269374..0c5becc822f9 100644
@@ -117,7 +117,7 @@ struct scripting_ops *script_spec__lookup(const char *spec);
 int script_spec__for_each(int (*cb)(struct scripting_ops *ops, const char *spec));
 
 void script_fetch_insn(struct perf_sample *sample, struct thread *thread,
-		       struct machine *machine);
+		       struct machine *machine, bool native_arch);
 
 void setup_perl_scripting(void);
 void setup_python_scripting(void);
Signature · 1 line
2.47.0.277.g8800431eea-goog
author reviewer
#13 lore raw

PATCH v6 11/22 perf script: Move perf_sample__sprintf_flags to trace-event-scripting.c

In reply to Ian Rogers

perf_sample__sprintf_flags is used in the python C code and so needs to be in the util library rather than a builtin.

---
 tools/perf/builtin-script.c             | 81 ------------------------
 tools/perf/util/python.c                |  6 --
 tools/perf/util/trace-event-scripting.c | 83 +++++++++++++++++++++++++
 3 files changed, 83 insertions(+), 87 deletions(-)
tools/perf/builtin-script.c +0 -81
index 22d78a9d8f27..1eda065dcb2b 100644
@@ -1693,87 +1693,6 @@ static int perf_sample__fprintf_bts(struct perf_sample *sample,
 	return printed;
 }
 
-static struct {
-	u32 flags;
-	const char *name;
-} sample_flags[] = {
-	{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_CALL, "call"},
-	{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_RETURN, "return"},
-	{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_CONDITIONAL, "jcc"},
-	{PERF_IP_FLAG_BRANCH, "jmp"},
-	{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_CALL | PERF_IP_FLAG_INTERRUPT, "int"},
-	{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_RETURN | PERF_IP_FLAG_INTERRUPT, "iret"},
-	{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_CALL | PERF_IP_FLAG_SYSCALLRET, "syscall"},
-	{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_RETURN | PERF_IP_FLAG_SYSCALLRET, "sysret"},
-	{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_ASYNC, "async"},
-	{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_CALL | PERF_IP_FLAG_ASYNC |	PERF_IP_FLAG_INTERRUPT, "hw int"},
-	{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_TX_ABORT, "tx abrt"},
-	{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_TRACE_BEGIN, "tr strt"},
-	{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_TRACE_END, "tr end"},
-	{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_CALL | PERF_IP_FLAG_VMENTRY, "vmentry"},
-	{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_CALL | PERF_IP_FLAG_VMEXIT, "vmexit"},
-	{PERF_IP_FLAG_BRANCH | PERF_IP_FLAG_BRANCH_MISS, "br miss"},
-	{0, NULL}
-};
-
-static const char *sample_flags_to_name(u32 flags)
-{
-	int i;
-
-	for (i = 0; sample_flags[i].name ; i++) {
-		if (sample_flags[i].flags == flags)
-			return sample_flags[i].name;
-	}
-
-	return NULL;
-}
-
-int perf_sample__sprintf_flags(u32 flags, char *str, size_t sz)
-{
-	u32 xf = PERF_IP_FLAG_IN_TX | PERF_IP_FLAG_INTR_DISABLE |
-		 PERF_IP_FLAG_INTR_TOGGLE;
-	const char *chars = PERF_IP_FLAG_CHARS;
-	const size_t n = strlen(PERF_IP_FLAG_CHARS);
-	const char *name = NULL;
-	size_t i, pos = 0;
-	char xs[16] = {0};
-
-	if (flags & xf)
-		snprintf(xs, sizeof(xs), "(%s%s%s)",
-			 flags & PERF_IP_FLAG_IN_TX ? "x" : "",
-			 flags & PERF_IP_FLAG_INTR_DISABLE ? "D" : "",
-			 flags & PERF_IP_FLAG_INTR_TOGGLE ? "t" : "");
-
-	name = sample_flags_to_name(flags & ~xf);
-	if (name)
-		return snprintf(str, sz, "%-15s%6s", name, xs);
-
-	if (flags & PERF_IP_FLAG_TRACE_BEGIN) {
-		name = sample_flags_to_name(flags & ~(xf | PERF_IP_FLAG_TRACE_BEGIN));
-		if (name)
-			return snprintf(str, sz, "tr strt %-7s%6s", name, xs);
-	}
-
-	if (flags & PERF_IP_FLAG_TRACE_END) {
-		name = sample_flags_to_name(flags & ~(xf | PERF_IP_FLAG_TRACE_END));
-		if (name)
-			return snprintf(str, sz, "tr end  %-7s%6s", name, xs);
-	}
-
-	for (i = 0; i < n; i++, flags >>= 1) {
-		if ((flags & 1) && pos < sz)
-			str[pos++] = chars[i];
-	}
-	for (; i < 32; i++, flags >>= 1) {
-		if ((flags & 1) && pos < sz)
-			str[pos++] = '?';
-	}
-	if (pos < sz)
-		str[pos] = 0;
-
-	return pos;
-}
-
 static int perf_sample__fprintf_flags(u32 flags, FILE *fp)
 {
 	char str[SAMPLE_FLAGS_BUF_SIZE];
tools/perf/util/python.c +0 -6
index 22edadd64e5f..77e02fcc51dd 100644
@@ -1317,12 +1317,6 @@ struct kwork_work *perf_kwork_add_work(struct perf_kwork *kwork __maybe_unused,
 	return NULL;
 }
 
-int perf_sample__sprintf_flags(u32 flags __maybe_unused, char *str __maybe_unused,
-			size_t sz __maybe_unused)
-{
-	return -1;
-}
-
 bool match_callstack_filter(struct machine *machine __maybe_unused, u64 *callstack __maybe_unused)
 {
 	return false;
tools/perf/util/trace-event-scripting.c +83 -0

92 lines not on this page · raw message

Signature · 1 line
2.47.0.277.g8800431eea-goog
author reviewer
#14 lore raw

PATCH v6 12/22 perf x86: Define arch_fetch_insn in NO_AUXTRACE builds

In reply to Ian Rogers

archinsn.c containing arch_fetch_insn was only enabled with CONFIG_AUXTRACE, but this meant that a NO_AUXTRACE build on x86 would use the empty weak version of arch_fetch_insn - weak symbols are a frequent source of errors like this and are outside of the C specification. Change it so that archinsn.c is always built on x86 and make the weak symbol empty version of arch_fetch_insn a strong one guarded by ifdefs.

arch_fetch_insn on x86 depends on insn_decode which is a function included then built into
intel-pt-insn-decoder.c. intel-pt-insn-decoder.c isn't built in a NO_AUXTRACE=1 build. Separate the insn_decode function from intel-pt-insn-decoder.c by just directly compiling the relevant file. Guard this compilation to be for either always on x86 (because of the use in arch_fetch_insn) or when auxtrace is enabled. Apply the CFLAGS overrides as necessary, reducing the amount of code where warnings are disabled.

---
 tools/perf/arch/x86/util/Build                 |  2 +-
 tools/perf/util/Build                          |  2 +-
 tools/perf/util/intel-pt-decoder/Build         | 18 ++++++++++++++----
 .../intel-pt-decoder/intel-pt-insn-decoder.c   |  3 ---
 tools/perf/util/trace-event-scripting.c        |  4 +++-
 5 files changed, 19 insertions(+), 10 deletions(-)
tools/perf/arch/x86/util/Build +1 -1
index bc56a8e70f34..c5df4a2cf180 100644
@@ -18,6 +18,6 @@ perf-util-$(CONFIG_LOCAL_LIBUNWIND)    += unwind-libunwind.o
 perf-util-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o
 
 perf-util-$(CONFIG_AUXTRACE) += auxtrace.o
-perf-util-$(CONFIG_AUXTRACE) += archinsn.o
+perf-util-y += archinsn.o
 perf-util-$(CONFIG_AUXTRACE) += intel-pt.o
 perf-util-$(CONFIG_AUXTRACE) += intel-bts.o
tools/perf/util/Build +1 -1
index 650974413849..340544a6f5ec 100644
@@ -122,7 +122,7 @@ perf-util-y += iostat.o
 perf-util-y += stream.o
 perf-util-y += kvm-stat.o
 perf-util-$(CONFIG_AUXTRACE) += auxtrace.o
-perf-util-$(CONFIG_AUXTRACE) += intel-pt-decoder/
+perf-util-y += intel-pt-decoder/
 perf-util-$(CONFIG_AUXTRACE) += intel-pt.o
 perf-util-$(CONFIG_AUXTRACE) += intel-bts.o
 perf-util-$(CONFIG_AUXTRACE) += arm-spe.o
tools/perf/util/intel-pt-decoder/Build +14 -4
index 30793d08c6d4..f99d150059b9 100644
@@ -9,14 +9,24 @@ $(OUTPUT)util/intel-pt-decoder/inat-tables.c: $(inat_tables_script) $(inat_table
 
 # Busybox's diff doesn't have -I, avoid warning in the case
 
-$(OUTPUT)util/intel-pt-decoder/intel-pt-insn-decoder.o: util/intel-pt-decoder/intel-pt-insn-decoder.c $(OUTPUT)util/intel-pt-decoder/inat-tables.c
+ifeq ($(SRCARCH),x86)
+  perf-util-y += inat.o insn.o
+else
+  perf-util-$(CONFIG_AUXTRACE) += inat.o insn.o
+endif
+
+$(OUTPUT)util/intel-pt-decoder/inat.o: $(srctree)/tools/arch/x86/lib/inat.c $(OUTPUT)util/intel-pt-decoder/inat-tables.c
 	$(call rule_mkdir)
 	$(call if_changed_dep,cc_o_c)
 
-CFLAGS_intel-pt-insn-decoder.o += -I$(OUTPUT)util/intel-pt-decoder
+CFLAGS_inat.o += -I$(OUTPUT)util/intel-pt-decoder
+
+$(OUTPUT)util/intel-pt-decoder/insn.o: $(srctree)/tools/arch/x86/lib/insn.c
+	$(call rule_mkdir)
+	$(call if_changed_dep,cc_o_c)
 
 ifeq ($(CC_NO_CLANG), 1)
-  CFLAGS_intel-pt-insn-decoder.o += -Wno-override-init
+  CFLAGS_insn.o += -Wno-override-init
 endif
 
-CFLAGS_intel-pt-insn-decoder.o += -Wno-packed
+CFLAGS_insn.o += -Wno-packed
tools/perf/util/intel-pt-decoder/intel-pt-insn-decoder.c +0 -3
index 47cf35799a4d..8fabddc1c0da 100644
@@ -11,9 +11,6 @@
 #include <byteswap.h>
 #include "../../../arch/x86/include/asm/insn.h"
 
-#include "../../../arch/x86/lib/inat.c"
-#include "../../../arch/x86/lib/insn.c"
-
 #include "event.h"
 
 #include "intel-pt-insn-decoder.h"
tools/perf/util/trace-event-scripting.c +3 -1
index 2e9da0b089ef..8d71998a9ecd 100644
@@ -272,11 +272,13 @@ void setup_perl_scripting(void)
 #endif
 #endif
 
-__weak void arch_fetch_insn(struct perf_sample *sample __maybe_unused,
+#if !defined(__i386__) && !defined(__x86_64__)
+void arch_fetch_insn(struct perf_sample *sample __maybe_unused,
 		     struct thread *thread __maybe_unused,
 		     struct machine *machine __maybe_unused)
 {
 }
+#endif
 
 void script_fetch_insn(struct perf_sample *sample, struct thread *thread,
 		       struct machine *machine, bool native_arch)
Signature · 1 line
2.47.0.277.g8800431eea-goog
author reviewer
#15 lore raw

PATCH v6 13/22 perf intel-pt: Remove stale build comment

In reply to Ian Rogers

Commit 00a263902ac3 ("perf intel-pt: Use shared x86 insn decoder") removed the use of diff, so remove stale busybox comment.

---
 tools/perf/util/intel-pt-decoder/Build | 2 --
 1 file changed, 2 deletions(-)
tools/perf/util/intel-pt-decoder/Build +0 -2
index f99d150059b9..5b8f0149167d 100644
@@ -7,8 +7,6 @@ $(OUTPUT)util/intel-pt-decoder/inat-tables.c: $(inat_tables_script) $(inat_table
 	$(call rule_mkdir)
 	@$(call echo-cmd,gen)$(AWK) -f $(inat_tables_script) $(inat_tables_maps) > $@ || rm -f $@
 
-# Busybox's diff doesn't have -I, avoid warning in the case
-
 ifeq ($(SRCARCH),x86)
   perf-util-y += inat.o insn.o
 else
Signature · 1 line
2.47.0.277.g8800431eea-goog
author reviewer
#16 lore raw

PATCH v6 14/22 perf env: Move arch errno function to only use in env

In reply to Ian Rogers

Move arch_syscalls__strerrno_function out of builtin-trace.c to env.c so that there isn't a util to builtin function call. This allows the python.c stub to be removed. Also, remove declaration/prototype from env.h and make static to reduce scope. The include is moved inside ifdefs to avoid, "defined but unused warnings".

---
 tools/perf/builtin-trace.c                  | 1 -
 tools/perf/trace/beauty/arch_errno_names.sh | 3 ++-
 tools/perf/util/env.c                       | 4 ++++
 tools/perf/util/env.h                       | 2 --
 tools/perf/util/python.c                    | 6 ------
 5 files changed, 6 insertions(+), 10 deletions(-)
tools/perf/builtin-trace.c +0 -1
index a661fbd870e7..29f64a4352dd 100644
@@ -1108,7 +1108,6 @@ static bool syscall_arg__strtoul_btf_type(char *bf __maybe_unused, size_t size _
 	    .strtoul	= STUL_STRARRAY_FLAGS, \
 	    .parm	= &strarray__##array, }
 
-#include "trace/beauty/arch_errno_names.c"
 #include "trace/beauty/eventfd.c"
 #include "trace/beauty/futex_op.c"
 #include "trace/beauty/futex_val3.c"
tools/perf/trace/beauty/arch_errno_names.sh +2 -1
index 30d3889b2957..b22890b8d272 100755
@@ -57,7 +57,8 @@ create_arch_errno_table_func()
 	archlist="$1"
 	default="$2"
 
-	printf 'arch_syscalls__strerrno_t *arch_syscalls__strerrno_function(const char *arch)\n'
+	printf 'static arch_syscalls__strerrno_t *\n'
+	printf 'arch_syscalls__strerrno_function(const char *arch)\n'
 	printf '{\n'
 	for arch in $archlist; do
 		arch_str=$(arch_string "$arch")
tools/perf/util/env.c +4 -0
index ccb464026642..e58f56412fbb 100644
@@ -470,6 +470,10 @@ const char *perf_env__arch(struct perf_env *env)
 	return normalize_arch(arch_name);
 }
 
+#if defined(HAVE_SYSCALL_TABLE_SUPPORT) && defined(HAVE_LIBTRACEEVENT)
+#include "trace/beauty/arch_errno_names.c"
+#endif
+
 const char *perf_env__arch_strerrno(struct perf_env *env __maybe_unused, int err __maybe_unused)
 {
 #if defined(HAVE_SYSCALL_TABLE_SUPPORT) && defined(HAVE_LIBTRACEEVENT)
tools/perf/util/env.h +0 -2
index ae604c4edbb7..e9db89395dab 100644
@@ -56,8 +56,6 @@ struct pmu_caps {
 
 typedef const char *(arch_syscalls__strerrno_t)(int err);
 
-arch_syscalls__strerrno_t *arch_syscalls__strerrno_function(const char *arch);
-
 struct perf_env {
 	char			*hostname;
 	char			*os_release;
tools/perf/util/python.c +0 -6
index 77e02fcc51dd..35d84a96dbec 100644
@@ -16,7 +16,6 @@
 #include "thread_map.h"
 #include "trace-event.h"
 #include "mmap.h"
-#include "util/env.h"
 #include "util/kwork.h"
 #include "util/sample.h"
 #include "util/lock-contention.h"
@@ -1305,11 +1304,6 @@ PyMODINIT_FUNC PyInit_perf(void)
 /* The following are stubs to avoid dragging in builtin-* objects. */
 /* TODO: move the code out of the builtin-* file into util. */
 
-arch_syscalls__strerrno_t *arch_syscalls__strerrno_function(const char *arch __maybe_unused)
-{
-	return NULL;
-}
-
 struct kwork_work *perf_kwork_add_work(struct perf_kwork *kwork __maybe_unused,
 				       struct kwork_class *class __maybe_unused,
 				       struct kwork_work *key  __maybe_unused)
Signature · 1 line
2.47.0.277.g8800431eea-goog
author reviewer
#17 lore raw

PATCH v6 15/22 perf lock: Move common lock contention code to new file

In reply to Ian Rogers

Avoid references from util code to builtin-lock that require python stubs. Move the functions and related variables to util/lock-contention.c. Add max_stack_depth parameter to match_callstack_filter to avoid sharing a global variable.

---
 tools/perf/builtin-lock.c             | 137 +--------------------
 tools/perf/util/Build                 |   1 +
 tools/perf/util/bpf_lock_contention.c |   2 +-
 tools/perf/util/lock-contention.c     | 170 ++++++++++++++++++++++++++
 tools/perf/util/lock-contention.h     |  37 ++----
 tools/perf/util/python.c              |  17 ---
 6 files changed, 185 insertions(+), 179 deletions(-)
 create mode 100644 tools/perf/util/lock-contention.c
tools/perf/builtin-lock.c +1 -136
index 062e2b56a2ab..f66948b1fbed 100644
@@ -46,15 +46,6 @@
 static struct perf_session *session;
 static struct target target;
 
-/* based on kernel/lockdep.c */
-#define LOCKHASH_BITS		12
-#define LOCKHASH_SIZE		(1UL << LOCKHASH_BITS)
-
-static struct hlist_head *lockhash_table;
-
-#define __lockhashfn(key)	hash_long((unsigned long)key, LOCKHASH_BITS)
-#define lockhashentry(key)	(lockhash_table + __lockhashfn((key)))
-
 static struct rb_root		thread_stats;
 
 static bool combine_locks;
@@ -67,24 +58,13 @@ static unsigned long bpf_map_entries = MAX_ENTRIES;
 static int max_stack_depth = CONTENTION_STACK_DEPTH;
 static int stack_skip = CONTENTION_STACK_SKIP;
 static int print_nr_entries = INT_MAX / 2;
-static LIST_HEAD(callstack_filters);
 static const char *output_name = NULL;
 static FILE *lock_output;
 
-struct callstack_filter {
-	struct list_head list;
-	char name[];
-};
-
 static struct lock_filter filters;
 
 static enum lock_aggr_mode aggr_mode = LOCK_AGGR_ADDR;
 
-static bool needs_callstack(void)
-{
-	return !list_empty(&callstack_filters);
-}
-
 static struct thread_stat *thread_stat_find(u32 tid)
 {
 	struct rb_node *node;
@@ -477,93 +457,6 @@ static struct lock_stat *pop_from_result(void)
 	return container_of(node, struct lock_stat, rb);
 }
 
-struct lock_stat *lock_stat_find(u64 addr)
-{
-	struct hlist_head *entry = lockhashentry(addr);
-	struct lock_stat *ret;
-
-	hlist_for_each_entry(ret, entry, hash_entry) {
-		if (ret->addr == addr)
-			return ret;
-	}
-	return NULL;
-}
-
-struct lock_stat *lock_stat_findnew(u64 addr, const char *name, int flags)
-{
-	struct hlist_head *entry = lockhashentry(addr);
-	struct lock_stat *ret, *new;
-
-	hlist_for_each_entry(ret, entry, hash_entry) {
-		if (ret->addr == addr)
-			return ret;
-	}
-
-	new = zalloc(sizeof(struct lock_stat));
-	if (!new)
-		goto alloc_failed;
-
-	new->addr = addr;
-	new->name = strdup(name);
-	if (!new->name) {
-		free(new);
-		goto alloc_failed;
-	}
-
-	new->flags = flags;
-	new->wait_time_min = ULLONG_MAX;
-
-	hlist_add_head(&new->hash_entry, entry);
-	return new;
-
-alloc_failed:
-	pr_err("memory allocation failed\n");
-	return NULL;
-}
-
-bool match_callstack_filter(struct machine *machine, u64 *callstack)
-{
-	struct map *kmap;
-	struct symbol *sym;
-	u64 ip;
-	const char *arch = perf_env__arch(machine->env);
-
-	if (list_empty(&callstack_filters))
-		return true;
-
-	for (int i = 0; i < max_stack_depth; i++) {
-		struct callstack_filter *filter;
-
-		/*
-		 * In powerpc, the callchain saved by kernel always includes
-		 * first three entries as the NIP (next instruction pointer),
-		 * LR (link register), and the contents of LR save area in the
-		 * second stack frame. In certain scenarios its possible to have
-		 * invalid kernel instruction addresses in either LR or the second
-		 * stack frame's LR. In that case, kernel will store that address as
-		 * zero.
-		 *
-		 * The below check will continue to look into callstack,
-		 * incase first or second callstack index entry has 0
-		 * address for powerpc.
-		 */
-		if (!callstack || (!callstack[i] && (strcmp(arch, "powerpc") ||
-						(i != 1 && i != 2))))
-			break;
-
-		ip = callstack[i];
-		sym = machine__find_kernel_symbol(machine, ip, &kmap);
-		if (sym == NULL)
-			continue;
-
-		list_for_each_entry(filter, &callstack_filters, list) {
-			if (strstr(sym->name, filter->name))
-				return true;
-		}
-	}
-	return false;
-}
-
 struct trace_lock_handler {
 	/* it's used on CONFIG_LOCKDEP */
 	int (*acquire_event)(struct evsel *evsel,
@@ -1165,7 +1058,7 @@ static int report_lock_contention_begin_event(struct evsel *evsel,
 		if (callstack == NULL)
 			return -ENOMEM;
 
-		if (!match_callstack_filter(machine, callstack)) {
+		if (!match_callstack_filter(machine, callstack, max_stack_depth)) {
 			free(callstack);
 			return 0;
 		}
@@ -2449,34 +2342,6 @@ static int parse_lock_addr(const struct option *opt __maybe_unused, const char *
 	return ret;
 }
 
-static int parse_call_stack(const struct option *opt __maybe_unused, const char *str,
-			   int unset __maybe_unused)
-{
-	char *s, *tmp, *tok;
-	int ret = 0;
-
-	s = strdup(str);
-	if (s == NULL)
-		return -1;
-
-	for (tok = strtok_r(s, ", ", &tmp); tok; tok = strtok_r(NULL, ", ", &tmp)) {
-		struct callstack_filter *entry;
-
-		entry = malloc(sizeof(*entry) + strlen(tok) + 1);
-		if (entry == NULL) {
-			pr_err("Memory allocation failure\n");
-			free(s);
-			return -1;
-		}
-
-		strcpy(entry->name, tok);
-		list_add_tail(&entry->list, &callstack_filters);
-	}
-
-	free(s);
-	return ret;
-}
-
 static int parse_output(const struct option *opt __maybe_unused, const char *str,
 			int unset __maybe_unused)
 {
tools/perf/util/Build +1 -0
index 340544a6f5ec..3c6cd8d81d88 100644
@@ -121,6 +121,7 @@ perf-util-y += topdown.o
 perf-util-y += iostat.o
 perf-util-y += stream.o
 perf-util-y += kvm-stat.o
+perf-util-y += lock-contention.o
 perf-util-$(CONFIG_AUXTRACE) += auxtrace.o
 perf-util-y += intel-pt-decoder/
 perf-util-$(CONFIG_AUXTRACE) += intel-pt.o
tools/perf/util/bpf_lock_contention.c +1 -1
index 41a1ad087895..37e17c56f106 100644
@@ -458,7 +458,7 @@ int lock_contention_read(struct lock_contention *con)
 		if (con->save_callstack) {
 			bpf_map_lookup_elem(stack, &key.stack_id, stack_trace);
 
-			if (!match_callstack_filter(machine, stack_trace)) {
+			if (!match_callstack_filter(machine, stack_trace, con->max_stack)) {
 				con->nr_filtered += data.count;
 				goto next;
 			}
tools/perf/util/lock-contention.c new file +170 -0

170 lines not on this page · raw message

tools/perf/util/lock-contention.h +12 -25
index 1a7248ff3889..bfa5c7db0a5d 100644
@@ -67,10 +67,11 @@ struct lock_stat {
  */
 #define MAX_LOCK_DEPTH 48
 
-struct lock_stat *lock_stat_find(u64 addr);
-struct lock_stat *lock_stat_findnew(u64 addr, const char *name, int flags);
+/* based on kernel/lockdep.c */
+#define LOCKHASH_BITS		12
+#define LOCKHASH_SIZE		(1UL << LOCKHASH_BITS)
 
-bool match_callstack_filter(struct machine *machine, u64 *callstack);
+extern struct hlist_head *lockhash_table;
 
 /*
  * struct lock_seq_stat:
@@ -148,7 +149,14 @@ struct lock_contention {
 	bool save_callstack;
 };
 
-#ifdef HAVE_BPF_SKEL
+struct option;
+int parse_call_stack(const struct option *opt, const char *str, int unset);
+bool needs_callstack(void);
+
+struct lock_stat *lock_stat_find(u64 addr);
+struct lock_stat *lock_stat_findnew(u64 addr, const char *name, int flags);
+
+bool match_callstack_filter(struct machine *machine, u64 *callstack, int max_stack_depth);
 
 int lock_contention_prepare(struct lock_contention *con);
 int lock_contention_start(void);
@@ -156,25 +164,4 @@ int lock_contention_stop(void);
 int lock_contention_read(struct lock_contention *con);
 int lock_contention_finish(struct lock_contention *con);
 
-#else  /* !HAVE_BPF_SKEL */
-
-static inline int lock_contention_prepare(struct lock_contention *con __maybe_unused)
-{
-	return 0;
-}
-
-static inline int lock_contention_start(void) { return 0; }
-static inline int lock_contention_stop(void) { return 0; }
-static inline int lock_contention_finish(struct lock_contention *con __maybe_unused)
-{
-	return 0;
-}
-
-static inline int lock_contention_read(struct lock_contention *con __maybe_unused)
-{
-	return 0;
-}
-
-#endif  /* HAVE_BPF_SKEL */
-
 #endif  /* PERF_LOCK_CONTENTION_H */
tools/perf/util/python.c +0 -17
index 35d84a96dbec..91fd444615cd 100644
@@ -18,7 +18,6 @@
 #include "mmap.h"
 #include "util/kwork.h"
 #include "util/sample.h"
-#include "util/lock-contention.h"
 #include <internal/lib.h>
 #include "../builtin.h"
 
@@ -1311,22 +1310,6 @@ struct kwork_work *perf_kwork_add_work(struct perf_kwork *kwork __maybe_unused,
 	return NULL;
 }
 
-bool match_callstack_filter(struct machine *machine __maybe_unused, u64 *callstack __maybe_unused)
-{
-	return false;
-}
-
-struct lock_stat *lock_stat_find(u64 addr __maybe_unused)
-{
-	return NULL;
-}
-
-struct lock_stat *lock_stat_findnew(u64 addr __maybe_unused, const char *name __maybe_unused,
-				int flags __maybe_unused)
-{
-	return NULL;
-}
-
 int cmd_inject(int argc __maybe_unused, const char *argv[] __maybe_unused)
 {
 	return -1;
Signature · 1 line
2.47.0.277.g8800431eea-goog
maintainer
#18 lore raw

Re: [PATCH v6 15/22] perf lock: Move common lock contention code to new file

Quoted 391 lines · Ian Rogers

Avoid references from util code to builtin-lock that require python stubs. Move the functions and related variables to util/lock-contention.c. Add max_stack_depth parameter to match_callstack_filter to avoid sharing a global variable.

---
 tools/perf/builtin-lock.c             | 137 +--------------------
 tools/perf/util/Build                 |   1 +
 tools/perf/util/bpf_lock_contention.c |   2 +-
 tools/perf/util/lock-contention.c     | 170 ++++++++++++++++++++++++++
 tools/perf/util/lock-contention.h     |  37 ++----
 tools/perf/util/python.c              |  17 ---
 6 files changed, 185 insertions(+), 179 deletions(-)
 create mode 100644 tools/perf/util/lock-contention.c
tools/perf/builtin-lock.c +1 -136
index 062e2b56a2ab..f66948b1fbed 100644
@@ -46,15 +46,6 @@
 static struct perf_session *session;
 static struct target target;
 
-/* based on kernel/lockdep.c */
-#define LOCKHASH_BITS		12
-#define LOCKHASH_SIZE		(1UL << LOCKHASH_BITS)
-
-static struct hlist_head *lockhash_table;
-
-#define __lockhashfn(key)	hash_long((unsigned long)key, LOCKHASH_BITS)
-#define lockhashentry(key)	(lockhash_table + __lockhashfn((key)))
-
 static struct rb_root		thread_stats;
 
 static bool combine_locks;
@@ -67,24 +58,13 @@ static unsigned long bpf_map_entries = MAX_ENTRIES;
 static int max_stack_depth = CONTENTION_STACK_DEPTH;
 static int stack_skip = CONTENTION_STACK_SKIP;
 static int print_nr_entries = INT_MAX / 2;
-static LIST_HEAD(callstack_filters);
 static const char *output_name = NULL;
 static FILE *lock_output;
 
-struct callstack_filter {
-	struct list_head list;
-	char name[];
-};
-
 static struct lock_filter filters;
 
 static enum lock_aggr_mode aggr_mode = LOCK_AGGR_ADDR;
 
-static bool needs_callstack(void)
-{
-	return !list_empty(&callstack_filters);
-}
-
 static struct thread_stat *thread_stat_find(u32 tid)
 {
 	struct rb_node *node;
@@ -477,93 +457,6 @@ static struct lock_stat *pop_from_result(void)
 	return container_of(node, struct lock_stat, rb);
 }
 
-struct lock_stat *lock_stat_find(u64 addr)
-{
-	struct hlist_head *entry = lockhashentry(addr);
-	struct lock_stat *ret;
-
-	hlist_for_each_entry(ret, entry, hash_entry) {
-		if (ret->addr == addr)
-			return ret;
-	}
-	return NULL;
-}
-
-struct lock_stat *lock_stat_findnew(u64 addr, const char *name, int flags)
-{
-	struct hlist_head *entry = lockhashentry(addr);
-	struct lock_stat *ret, *new;
-
-	hlist_for_each_entry(ret, entry, hash_entry) {
-		if (ret->addr == addr)
-			return ret;
-	}
-
-	new = zalloc(sizeof(struct lock_stat));
-	if (!new)
-		goto alloc_failed;
-
-	new->addr = addr;
-	new->name = strdup(name);
-	if (!new->name) {
-		free(new);
-		goto alloc_failed;
-	}
-
-	new->flags = flags;
-	new->wait_time_min = ULLONG_MAX;
-
-	hlist_add_head(&new->hash_entry, entry);
-	return new;
-
-alloc_failed:
-	pr_err("memory allocation failed\n");
-	return NULL;
-}
-
-bool match_callstack_filter(struct machine *machine, u64 *callstack)
-{
-	struct map *kmap;
-	struct symbol *sym;
-	u64 ip;
-	const char *arch = perf_env__arch(machine->env);
-
-	if (list_empty(&callstack_filters))
-		return true;
-
-	for (int i = 0; i < max_stack_depth; i++) {
-		struct callstack_filter *filter;
-
-		/*
-		 * In powerpc, the callchain saved by kernel always includes
-		 * first three entries as the NIP (next instruction pointer),
-		 * LR (link register), and the contents of LR save area in the
-		 * second stack frame. In certain scenarios its possible to have
-		 * invalid kernel instruction addresses in either LR or the second
-		 * stack frame's LR. In that case, kernel will store that address as
-		 * zero.
-		 *
-		 * The below check will continue to look into callstack,
-		 * incase first or second callstack index entry has 0
-		 * address for powerpc.
-		 */
-		if (!callstack || (!callstack[i] && (strcmp(arch, "powerpc") ||
-						(i != 1 && i != 2))))
-			break;
-
-		ip = callstack[i];
-		sym = machine__find_kernel_symbol(machine, ip, &kmap);
-		if (sym == NULL)
-			continue;
-
-		list_for_each_entry(filter, &callstack_filters, list) {
-			if (strstr(sym->name, filter->name))
-				return true;
-		}
-	}
-	return false;
-}
-
 struct trace_lock_handler {
 	/* it's used on CONFIG_LOCKDEP */
 	int (*acquire_event)(struct evsel *evsel,
@@ -1165,7 +1058,7 @@ static int report_lock_contention_begin_event(struct evsel *evsel,
 		if (callstack == NULL)
 			return -ENOMEM;
 
-		if (!match_callstack_filter(machine, callstack)) {
+		if (!match_callstack_filter(machine, callstack, max_stack_depth)) {
 			free(callstack);
 			return 0;
 		}
@@ -2449,34 +2342,6 @@ static int parse_lock_addr(const struct option *opt __maybe_unused, const char *
 	return ret;
 }
 
-static int parse_call_stack(const struct option *opt __maybe_unused, const char *str,
-			   int unset __maybe_unused)
-{
-	char *s, *tmp, *tok;
-	int ret = 0;
-
-	s = strdup(str);
-	if (s == NULL)
-		return -1;
-
-	for (tok = strtok_r(s, ", ", &tmp); tok; tok = strtok_r(NULL, ", ", &tmp)) {
-		struct callstack_filter *entry;
-
-		entry = malloc(sizeof(*entry) + strlen(tok) + 1);
-		if (entry == NULL) {
-			pr_err("Memory allocation failure\n");
-			free(s);
-			return -1;
-		}
-
-		strcpy(entry->name, tok);
-		list_add_tail(&entry->list, &callstack_filters);
-	}
-
-	free(s);
-	return ret;
-}
-
 static int parse_output(const struct option *opt __maybe_unused, const char *str,
 			int unset __maybe_unused)
 {
tools/perf/util/Build +1 -0
index 340544a6f5ec..3c6cd8d81d88 100644
@@ -121,6 +121,7 @@ perf-util-y += topdown.o
 perf-util-y += iostat.o
 perf-util-y += stream.o
 perf-util-y += kvm-stat.o
+perf-util-y += lock-contention.o
 perf-util-$(CONFIG_AUXTRACE) += auxtrace.o
 perf-util-y += intel-pt-decoder/
 perf-util-$(CONFIG_AUXTRACE) += intel-pt.o
tools/perf/util/bpf_lock_contention.c +1 -1
index 41a1ad087895..37e17c56f106 100644
@@ -458,7 +458,7 @@ int lock_contention_read(struct lock_contention *con)
 		if (con->save_callstack) {
 			bpf_map_lookup_elem(stack, &key.stack_id, stack_trace);
 
-			if (!match_callstack_filter(machine, stack_trace)) {
+			if (!match_callstack_filter(machine, stack_trace, con->max_stack)) {
 				con->nr_filtered += data.count;
 				goto next;
 			}
tools/perf/util/lock-contention.c new file +160 -0
new file mode 100644
index 000000000000..841bb18b1f06
@@ -0,0 +1,170 @@
+// SPDX-License-Identifier: GPL-2.0
+#include "debug.h"
+#include "env.h"
+#include "lock-contention.h"
+#include "machine.h"
+#include "symbol.h"
+
+#include <limits.h>
+#include <string.h>
+
+#include <linux/hash.h>
+#include <linux/zalloc.h>
+
+#define __lockhashfn(key)	hash_long((unsigned long)key, LOCKHASH_BITS)
+#define lockhashentry(key)	(lockhash_table + __lockhashfn((key)))
+
+struct callstack_filter {
+	struct list_head list;
+	char name[];
+};
+
+static LIST_HEAD(callstack_filters);
+struct hlist_head *lockhash_table;
+
+int parse_call_stack(const struct option *opt __maybe_unused, const char *str,
+		     int unset __maybe_unused)
+{
+	char *s, *tmp, *tok;
+	int ret = 0;
+
+	s = strdup(str);
+	if (s == NULL)
+		return -1;
+
+	for (tok = strtok_r(s, ", ", &tmp); tok; tok = strtok_r(NULL, ", ", &tmp)) {
+		struct callstack_filter *entry;
+
+		entry = malloc(sizeof(*entry) + strlen(tok) + 1);
+		if (entry == NULL) {
+			pr_err("Memory allocation failure\n");
+			free(s);
+			return -1;
+		}
+
+		strcpy(entry->name, tok);
+		list_add_tail(&entry->list, &callstack_filters);
+	}
+
+	free(s);
+	return ret;
+}
+
+bool needs_callstack(void)
+{
+	return !list_empty(&callstack_filters);
+}
+
+struct lock_stat *lock_stat_find(u64 addr)
+{
+	struct hlist_head *entry = lockhashentry(addr);
+	struct lock_stat *ret;
+
+	hlist_for_each_entry(ret, entry, hash_entry) {
+		if (ret->addr == addr)
+			return ret;
+	}
+	return NULL;
+}
+
+struct lock_stat *lock_stat_findnew(u64 addr, const char *name, int flags)
+{
+	struct hlist_head *entry = lockhashentry(addr);
+	struct lock_stat *ret, *new;
+
+	hlist_for_each_entry(ret, entry, hash_entry) {
+		if (ret->addr == addr)
+			return ret;
+	}
+
+	new = zalloc(sizeof(struct lock_stat));
+	if (!new)
+		goto alloc_failed;
+
+	new->addr = addr;
+	new->name = strdup(name);
+	if (!new->name) {
+		free(new);
+		goto alloc_failed;
+	}
+
+	new->flags = flags;
+	new->wait_time_min = ULLONG_MAX;
+
+	hlist_add_head(&new->hash_entry, entry);
+	return new;
+
+alloc_failed:
+	pr_err("memory allocation failed\n");
+	return NULL;
+}
+
+bool match_callstack_filter(struct machine *machine, u64 *callstack, int max_stack_depth)
+{
+	struct map *kmap;
+	struct symbol *sym;
+	u64 ip;
+	const char *arch = perf_env__arch(machine->env);
+
+	if (list_empty(&callstack_filters))
+		return true;
+
+	for (int i = 0; i < max_stack_depth; i++) {
+		struct callstack_filter *filter;
+
+		/*
+		 * In powerpc, the callchain saved by kernel always includes
+		 * first three entries as the NIP (next instruction pointer),
+		 * LR (link register), and the contents of LR save area in the
+		 * second stack frame. In certain scenarios its possible to have
+		 * invalid kernel instruction addresses in either LR or the second
+		 * stack frame's LR. In that case, kernel will store that address as
+		 * zero.
+		 *
+		 * The below check will continue to look into callstack,
+		 * incase first or second callstack index entry has 0
+		 * address for powerpc.
+		 */
+		if (!callstack || (!callstack[i] && (strcmp(arch, "powerpc") ||
+						(i != 1 && i != 2))))
+			break;
+
+		ip = callstack[i];
+		sym = machine__find_kernel_symbol(machine, ip, &kmap);
+		if (sym == NULL)
+			continue;
+
+		list_for_each_entry(filter, &callstack_filters, list) {
+			if (strstr(sym->name, filter->name))
+				return true;
+		}
+	}
+	return false;
+}
+
+#ifndef HAVE_BPF_SKEL
+int lock_contention_prepare(struct lock_contention *con __maybe_unused)
+{
+	return 0;
+}
+
+int lock_contention_start(void)
+{
+	return 0;
+}
+
+int lock_contention_stop(void)
+{
+	return 0;
+}
+

tools/perf/util/lock-contention.c line 170

Quoted 10 lines
tools/perf/util/lock-contention.c new file +10 -0
+int lock_contention_finish(struct lock_contention *con __maybe_unused)
+{
+	return 0;
+}
+
+int lock_contention_read(struct lock_contention *con __maybe_unused)
+{
+	return 0;
+}
+#endif  /* !HAVE_BPF_SKEL */

I still think it's the convention to have them in a header file as static inline functions and reduce the #ifdef in the .c file.

Thanks,
Namhyung

Quoted 99 lines
tools/perf/util/lock-contention.h +12 -25
index 1a7248ff3889..bfa5c7db0a5d 100644
@@ -67,10 +67,11 @@ struct lock_stat {
  */
 #define MAX_LOCK_DEPTH 48
 
-struct lock_stat *lock_stat_find(u64 addr);
-struct lock_stat *lock_stat_findnew(u64 addr, const char *name, int flags);
+/* based on kernel/lockdep.c */
+#define LOCKHASH_BITS		12
+#define LOCKHASH_SIZE		(1UL << LOCKHASH_BITS)
 
-bool match_callstack_filter(struct machine *machine, u64 *callstack);
+extern struct hlist_head *lockhash_table;
 
 /*
  * struct lock_seq_stat:
@@ -148,7 +149,14 @@ struct lock_contention {
 	bool save_callstack;
 };
 
-#ifdef HAVE_BPF_SKEL
+struct option;
+int parse_call_stack(const struct option *opt, const char *str, int unset);
+bool needs_callstack(void);
+
+struct lock_stat *lock_stat_find(u64 addr);
+struct lock_stat *lock_stat_findnew(u64 addr, const char *name, int flags);
+
+bool match_callstack_filter(struct machine *machine, u64 *callstack, int max_stack_depth);
 
 int lock_contention_prepare(struct lock_contention *con);
 int lock_contention_start(void);
@@ -156,25 +164,4 @@ int lock_contention_stop(void);
 int lock_contention_read(struct lock_contention *con);
 int lock_contention_finish(struct lock_contention *con);
 
-#else  /* !HAVE_BPF_SKEL */
-
-static inline int lock_contention_prepare(struct lock_contention *con __maybe_unused)
-{
-	return 0;
-}
-
-static inline int lock_contention_start(void) { return 0; }
-static inline int lock_contention_stop(void) { return 0; }
-static inline int lock_contention_finish(struct lock_contention *con __maybe_unused)
-{
-	return 0;
-}
-
-static inline int lock_contention_read(struct lock_contention *con __maybe_unused)
-{
-	return 0;
-}
-
-#endif  /* HAVE_BPF_SKEL */
-
 #endif  /* PERF_LOCK_CONTENTION_H */
tools/perf/util/python.c +0 -17
index 35d84a96dbec..91fd444615cd 100644
@@ -18,7 +18,6 @@
 #include "mmap.h"
 #include "util/kwork.h"
 #include "util/sample.h"
-#include "util/lock-contention.h"
 #include <internal/lib.h>
 #include "../builtin.h"
 
@@ -1311,22 +1310,6 @@ struct kwork_work *perf_kwork_add_work(struct perf_kwork *kwork __maybe_unused,
 	return NULL;
 }
 
-bool match_callstack_filter(struct machine *machine __maybe_unused, u64 *callstack __maybe_unused)
-{
-	return false;
-}
-
-struct lock_stat *lock_stat_find(u64 addr __maybe_unused)
-{
-	return NULL;
-}
-
-struct lock_stat *lock_stat_findnew(u64 addr __maybe_unused, const char *name __maybe_unused,
-				int flags __maybe_unused)
-{
-	return NULL;
-}
-
 int cmd_inject(int argc __maybe_unused, const char *argv[] __maybe_unused)
 {
 	return -1;
Signature · 1 line
2.47.0.277.g8800431eea-goog
author reviewer
#19 lore raw

Re: [PATCH v6 15/22] perf lock: Move common lock contention code to new file

Quoted 398 lines · Namhyung Kim
Quoted 396 lines · Ian Rogers

Avoid references from util code to builtin-lock that require python stubs. Move the functions and related variables to util/lock-contention.c. Add max_stack_depth parameter to match_callstack_filter to avoid sharing a global variable.

---
 tools/perf/builtin-lock.c             | 137 +--------------------
 tools/perf/util/Build                 |   1 +
 tools/perf/util/bpf_lock_contention.c |   2 +-
 tools/perf/util/lock-contention.c     | 170 ++++++++++++++++++++++++++
 tools/perf/util/lock-contention.h     |  37 ++----
 tools/perf/util/python.c              |  17 ---
 6 files changed, 185 insertions(+), 179 deletions(-)
 create mode 100644 tools/perf/util/lock-contention.c
tools/perf/builtin-lock.c +1 -136
index 062e2b56a2ab..f66948b1fbed 100644
@@ -46,15 +46,6 @@
 static struct perf_session *session;
 static struct target target;
 
-/* based on kernel/lockdep.c */
-#define LOCKHASH_BITS                12
-#define LOCKHASH_SIZE                (1UL << LOCKHASH_BITS)
-
-static struct hlist_head *lockhash_table;
-
-#define __lockhashfn(key)    hash_long((unsigned long)key, LOCKHASH_BITS)
-#define lockhashentry(key)   (lockhash_table + __lockhashfn((key)))
-
 static struct rb_root                thread_stats;
 
 static bool combine_locks;
@@ -67,24 +58,13 @@ static unsigned long bpf_map_entries = MAX_ENTRIES;
 static int max_stack_depth = CONTENTION_STACK_DEPTH;
 static int stack_skip = CONTENTION_STACK_SKIP;
 static int print_nr_entries = INT_MAX / 2;
-static LIST_HEAD(callstack_filters);
 static const char *output_name = NULL;
 static FILE *lock_output;
 
-struct callstack_filter {
-     struct list_head list;
-     char name[];
-};
-
 static struct lock_filter filters;
 
 static enum lock_aggr_mode aggr_mode = LOCK_AGGR_ADDR;
 
-static bool needs_callstack(void)
-{
-     return !list_empty(&callstack_filters);
-}
-
 static struct thread_stat *thread_stat_find(u32 tid)
 {
      struct rb_node *node;
@@ -477,93 +457,6 @@ static struct lock_stat *pop_from_result(void)
      return container_of(node, struct lock_stat, rb);
 }
 
-struct lock_stat *lock_stat_find(u64 addr)
-{
-     struct hlist_head *entry = lockhashentry(addr);
-     struct lock_stat *ret;
-
-     hlist_for_each_entry(ret, entry, hash_entry) {
-             if (ret->addr == addr)
-                     return ret;
-     }
-     return NULL;
-}
-
-struct lock_stat *lock_stat_findnew(u64 addr, const char *name, int flags)
-{
-     struct hlist_head *entry = lockhashentry(addr);
-     struct lock_stat *ret, *new;
-
-     hlist_for_each_entry(ret, entry, hash_entry) {
-             if (ret->addr == addr)
-                     return ret;
-     }
-
-     new = zalloc(sizeof(struct lock_stat));
-     if (!new)
-             goto alloc_failed;
-
-     new->addr = addr;
-     new->name = strdup(name);
-     if (!new->name) {
-             free(new);
-             goto alloc_failed;
-     }
-
-     new->flags = flags;
-     new->wait_time_min = ULLONG_MAX;
-
-     hlist_add_head(&new->hash_entry, entry);
-     return new;
-
-alloc_failed:
-     pr_err("memory allocation failed\n");
-     return NULL;
-}
-
-bool match_callstack_filter(struct machine *machine, u64 *callstack)
-{
-     struct map *kmap;
-     struct symbol *sym;
-     u64 ip;
-     const char *arch = perf_env__arch(machine->env);
-
-     if (list_empty(&callstack_filters))
-             return true;
-
-     for (int i = 0; i < max_stack_depth; i++) {
-             struct callstack_filter *filter;
-
-             /*
-              * In powerpc, the callchain saved by kernel always includes
-              * first three entries as the NIP (next instruction pointer),
-              * LR (link register), and the contents of LR save area in the
-              * second stack frame. In certain scenarios its possible to have
-              * invalid kernel instruction addresses in either LR or the second
-              * stack frame's LR. In that case, kernel will store that address as
-              * zero.
-              *
-              * The below check will continue to look into callstack,
-              * incase first or second callstack index entry has 0
-              * address for powerpc.
-              */
-             if (!callstack || (!callstack[i] && (strcmp(arch, "powerpc") ||
-                                             (i != 1 && i != 2))))
-                     break;
-
-             ip = callstack[i];
-             sym = machine__find_kernel_symbol(machine, ip, &kmap);
-             if (sym == NULL)
-                     continue;
-
-             list_for_each_entry(filter, &callstack_filters, list) {
-                     if (strstr(sym->name, filter->name))
-                             return true;
-             }
-     }
-     return false;
-}
-
 struct trace_lock_handler {
      /* it's used on CONFIG_LOCKDEP */
      int (*acquire_event)(struct evsel *evsel,
@@ -1165,7 +1058,7 @@ static int report_lock_contention_begin_event(struct evsel *evsel,
              if (callstack == NULL)
                      return -ENOMEM;
 
-             if (!match_callstack_filter(machine, callstack)) {
+             if (!match_callstack_filter(machine, callstack, max_stack_depth)) {
                      free(callstack);
                      return 0;
              }
@@ -2449,34 +2342,6 @@ static int parse_lock_addr(const struct option *opt __maybe_unused, const char *
      return ret;
 }
 
-static int parse_call_stack(const struct option *opt __maybe_unused, const char *str,
-                        int unset __maybe_unused)
-{
-     char *s, *tmp, *tok;
-     int ret = 0;
-
-     s = strdup(str);
-     if (s == NULL)
-             return -1;
-
-     for (tok = strtok_r(s, ", ", &tmp); tok; tok = strtok_r(NULL, ", ", &tmp)) {
-             struct callstack_filter *entry;
-
-             entry = malloc(sizeof(*entry) + strlen(tok) + 1);
-             if (entry == NULL) {
-                     pr_err("Memory allocation failure\n");
-                     free(s);
-                     return -1;
-             }
-
-             strcpy(entry->name, tok);
-             list_add_tail(&entry->list, &callstack_filters);
-     }
-
-     free(s);
-     return ret;
-}
-
 static int parse_output(const struct option *opt __maybe_unused, const char *str,
                      int unset __maybe_unused)
 {
tools/perf/util/Build +1 -0
index 340544a6f5ec..3c6cd8d81d88 100644
@@ -121,6 +121,7 @@ perf-util-y += topdown.o
 perf-util-y += iostat.o
 perf-util-y += stream.o
 perf-util-y += kvm-stat.o
+perf-util-y += lock-contention.o
 perf-util-$(CONFIG_AUXTRACE) += auxtrace.o
 perf-util-y += intel-pt-decoder/
 perf-util-$(CONFIG_AUXTRACE) += intel-pt.o
tools/perf/util/bpf_lock_contention.c +1 -1
index 41a1ad087895..37e17c56f106 100644
@@ -458,7 +458,7 @@ int lock_contention_read(struct lock_contention *con)
              if (con->save_callstack) {
                      bpf_map_lookup_elem(stack, &key.stack_id, stack_trace);
 
-                     if (!match_callstack_filter(machine, stack_trace)) {
+                     if (!match_callstack_filter(machine, stack_trace, con->max_stack)) {
                              con->nr_filtered += data.count;
                              goto next;
                      }
tools/perf/util/lock-contention.c new file +165 -0
new file mode 100644
index 000000000000..841bb18b1f06
@@ -0,0 +1,170 @@
+// SPDX-License-Identifier: GPL-2.0
+#include "debug.h"
+#include "env.h"
+#include "lock-contention.h"
+#include "machine.h"
+#include "symbol.h"
+
+#include <limits.h>
+#include <string.h>
+
+#include <linux/hash.h>
+#include <linux/zalloc.h>
+
+#define __lockhashfn(key)    hash_long((unsigned long)key, LOCKHASH_BITS)
+#define lockhashentry(key)   (lockhash_table + __lockhashfn((key)))
+
+struct callstack_filter {
+     struct list_head list;
+     char name[];
+};
+
+static LIST_HEAD(callstack_filters);
+struct hlist_head *lockhash_table;
+
+int parse_call_stack(const struct option *opt __maybe_unused, const char *str,
+                  int unset __maybe_unused)
+{
+     char *s, *tmp, *tok;
+     int ret = 0;
+
+     s = strdup(str);
+     if (s == NULL)
+             return -1;
+
+     for (tok = strtok_r(s, ", ", &tmp); tok; tok = strtok_r(NULL, ", ", &tmp)) {
+             struct callstack_filter *entry;
+
+             entry = malloc(sizeof(*entry) + strlen(tok) + 1);
+             if (entry == NULL) {
+                     pr_err("Memory allocation failure\n");
+                     free(s);
+                     return -1;
+             }
+
+             strcpy(entry->name, tok);
+             list_add_tail(&entry->list, &callstack_filters);
+     }
+
+     free(s);
+     return ret;
+}
+
+bool needs_callstack(void)
+{
+     return !list_empty(&callstack_filters);
+}
+
+struct lock_stat *lock_stat_find(u64 addr)
+{
+     struct hlist_head *entry = lockhashentry(addr);
+     struct lock_stat *ret;
+
+     hlist_for_each_entry(ret, entry, hash_entry) {
+             if (ret->addr == addr)
+                     return ret;
+     }
+     return NULL;
+}
+
+struct lock_stat *lock_stat_findnew(u64 addr, const char *name, int flags)
+{
+     struct hlist_head *entry = lockhashentry(addr);
+     struct lock_stat *ret, *new;
+
+     hlist_for_each_entry(ret, entry, hash_entry) {
+             if (ret->addr == addr)
+                     return ret;
+     }
+
+     new = zalloc(sizeof(struct lock_stat));
+     if (!new)
+             goto alloc_failed;
+
+     new->addr = addr;
+     new->name = strdup(name);
+     if (!new->name) {
+             free(new);
+             goto alloc_failed;
+     }
+
+     new->flags = flags;
+     new->wait_time_min = ULLONG_MAX;
+
+     hlist_add_head(&new->hash_entry, entry);
+     return new;
+
+alloc_failed:
+     pr_err("memory allocation failed\n");
+     return NULL;
+}
+
+bool match_callstack_filter(struct machine *machine, u64 *callstack, int max_stack_depth)
+{
+     struct map *kmap;
+     struct symbol *sym;
+     u64 ip;
+     const char *arch = perf_env__arch(machine->env);
+
+     if (list_empty(&callstack_filters))
+             return true;
+
+     for (int i = 0; i < max_stack_depth; i++) {
+             struct callstack_filter *filter;
+
+             /*
+              * In powerpc, the callchain saved by kernel always includes
+              * first three entries as the NIP (next instruction pointer),
+              * LR (link register), and the contents of LR save area in the
+              * second stack frame. In certain scenarios its possible to have
+              * invalid kernel instruction addresses in either LR or the second
+              * stack frame's LR. In that case, kernel will store that address as
+              * zero.
+              *
+              * The below check will continue to look into callstack,
+              * incase first or second callstack index entry has 0
+              * address for powerpc.
+              */
+             if (!callstack || (!callstack[i] && (strcmp(arch, "powerpc") ||
+                                             (i != 1 && i != 2))))
+                     break;
+
+             ip = callstack[i];
+             sym = machine__find_kernel_symbol(machine, ip, &kmap);
+             if (sym == NULL)
+                     continue;
+
+             list_for_each_entry(filter, &callstack_filters, list) {
+                     if (strstr(sym->name, filter->name))
+                             return true;
+             }
+     }
+     return false;
+}
+
+#ifndef HAVE_BPF_SKEL
+int lock_contention_prepare(struct lock_contention *con __maybe_unused)
+{
+     return 0;
+}
+
+int lock_contention_start(void)
+{
+     return 0;
+}
+
+int lock_contention_stop(void)
+{
+     return 0;
+}
+
+int lock_contention_finish(struct lock_contention *con __maybe_unused)
+{
+     return 0;
+}
+
Quoted 8 lines

tools/perf/util/lock-contention.c line 170

Quoted 5 lines
tools/perf/util/lock-contention.c new file +5 -0
+int lock_contention_read(struct lock_contention *con __maybe_unused)
+{
+     return 0;
+}
+#endif  /* !HAVE_BPF_SKEL */

I still think it's the convention to have them in a header file as static inline functions and reduce the #ifdef in the .c file.

Shouldn't minimizing ifdefs, and associated cognitive load, in header files be the priority given they are #included many times while the .c file is only compiled once?
Shouldn't a goal of the header file be to abstract away things like HAVE_BPF_SKEL?
I'm not clear what the goal of having the functions in the header files is, performance? The code isn't going to run anyway. I feel lock_contention.h is smaller and easier to read like this but I also don't care enough to fight. I did this change here as lock_contention.h was being brought into python.c for the sake of stubbing out functions that the header file was also subbing out for !BPF_HAVE_SKEL. A single stub felt like progress.

Thanks,
Ian

Quoted 99 lines
Quoted 99 lines
tools/perf/util/lock-contention.h +12 -25
index 1a7248ff3889..bfa5c7db0a5d 100644
@@ -67,10 +67,11 @@ struct lock_stat {
  */
 #define MAX_LOCK_DEPTH 48
 
-struct lock_stat *lock_stat_find(u64 addr);
-struct lock_stat *lock_stat_findnew(u64 addr, const char *name, int flags);
+/* based on kernel/lockdep.c */
+#define LOCKHASH_BITS                12
+#define LOCKHASH_SIZE                (1UL << LOCKHASH_BITS)
 
-bool match_callstack_filter(struct machine *machine, u64 *callstack);
+extern struct hlist_head *lockhash_table;
 
 /*
  * struct lock_seq_stat:
@@ -148,7 +149,14 @@ struct lock_contention {
      bool save_callstack;
 };
 
-#ifdef HAVE_BPF_SKEL
+struct option;
+int parse_call_stack(const struct option *opt, const char *str, int unset);
+bool needs_callstack(void);
+
+struct lock_stat *lock_stat_find(u64 addr);
+struct lock_stat *lock_stat_findnew(u64 addr, const char *name, int flags);
+
+bool match_callstack_filter(struct machine *machine, u64 *callstack, int max_stack_depth);
 
 int lock_contention_prepare(struct lock_contention *con);
 int lock_contention_start(void);
@@ -156,25 +164,4 @@ int lock_contention_stop(void);
 int lock_contention_read(struct lock_contention *con);
 int lock_contention_finish(struct lock_contention *con);
 
-#else  /* !HAVE_BPF_SKEL */
-
-static inline int lock_contention_prepare(struct lock_contention *con __maybe_unused)
-{
-     return 0;
-}
-
-static inline int lock_contention_start(void) { return 0; }
-static inline int lock_contention_stop(void) { return 0; }
-static inline int lock_contention_finish(struct lock_contention *con __maybe_unused)
-{
-     return 0;
-}
-
-static inline int lock_contention_read(struct lock_contention *con __maybe_unused)
-{
-     return 0;
-}
-
-#endif  /* HAVE_BPF_SKEL */
-
 #endif  /* PERF_LOCK_CONTENTION_H */
tools/perf/util/python.c +0 -18
index 35d84a96dbec..91fd444615cd 100644
@@ -18,7 +18,6 @@
 #include "mmap.h"
 #include "util/kwork.h"
 #include "util/sample.h"
-#include "util/lock-contention.h"
 #include <internal/lib.h>
 #include "../builtin.h"
 
@@ -1311,22 +1310,6 @@ struct kwork_work *perf_kwork_add_work(struct perf_kwork *kwork __maybe_unused,
      return NULL;
 }
 
-bool match_callstack_filter(struct machine *machine __maybe_unused, u64 *callstack __maybe_unused)
-{
-     return false;
-}
-
-struct lock_stat *lock_stat_find(u64 addr __maybe_unused)
-{
-     return NULL;
-}
-
-struct lock_stat *lock_stat_findnew(u64 addr __maybe_unused, const char *name __maybe_unused,
-                             int flags __maybe_unused)
-{
-     return NULL;
-}
-
 int cmd_inject(int argc __maybe_unused, const char *argv[] __maybe_unused)
 {
      return -1;
--

2.47.0.277.g8800431eea-goog

maintainer
#20 lore raw

Re: [PATCH v6 15/22] perf lock: Move common lock contention code to new file

Quoted 1 line · Ian Rogers

On Mon, Nov 18, 2024 at 4:23 PM Namhyung Kim <[email protected]> wrote:

[SNIP]

Quoted 31 lines
Quoted 22 lines · Ian Rogers
Quoted 21 lines
tools/perf/util/lock-contention.c new file +21 -0
+#ifndef HAVE_BPF_SKEL
+int lock_contention_prepare(struct lock_contention *con __maybe_unused)
+{
+     return 0;
+}
+
+int lock_contention_start(void)
+{
+     return 0;
+}
+
+int lock_contention_stop(void)
+{
+     return 0;
+}
+
+int lock_contention_finish(struct lock_contention *con __maybe_unused)
+{
+     return 0;
+}
+
Quoted 8 lines

tools/perf/util/lock-contention.c line 170

Quoted 5 lines
tools/perf/util/lock-contention.c new file +5 -0
+int lock_contention_read(struct lock_contention *con __maybe_unused)
+{
+     return 0;
+}
+#endif  /* !HAVE_BPF_SKEL */

I still think it's the convention to have them in a header file as static inline functions and reduce the #ifdef in the .c file.

Quoted 12 lines

Shouldn't minimizing ifdefs, and associated cognitive load, in header files be the priority given they are #included many times while the .c file is only compiled once?
Shouldn't a goal of the header file be to abstract away things like HAVE_BPF_SKEL?
I'm not clear what the goal of having the functions in the header files is, performance? The code isn't going to run anyway. I feel lock_contention.h is smaller and easier to read like this but I also don't care enough to fight. I did this change here as lock_contention.h was being brought into python.c for the sake of stubbing out functions that the header file was also subbing out for !BPF_HAVE_SKEL. A single stub felt like progress.

I think it may have the empty functions in the binary if we keep the functions in the .c file whereas compilers would optimize away them if they are static inline functions.

Thanks,
Namhyung

author reviewer
#21 lore raw

Re: [PATCH v6 15/22] perf lock: Move common lock contention code to new file

Quoted 48 lines · Namhyung Kim
Quoted 1 line · Ian Rogers

On Mon, Nov 18, 2024 at 4:23 PM Namhyung Kim <[email protected]> wrote:

[SNIP]

Quoted 31 lines
Quoted 22 lines · Ian Rogers
Quoted 21 lines
tools/perf/util/lock-contention.c new file +21 -0
+#ifndef HAVE_BPF_SKEL
+int lock_contention_prepare(struct lock_contention *con __maybe_unused)
+{
+     return 0;
+}
+
+int lock_contention_start(void)
+{
+     return 0;
+}
+
+int lock_contention_stop(void)
+{
+     return 0;
+}
+
+int lock_contention_finish(struct lock_contention *con __maybe_unused)
+{
+     return 0;
+}
+
Quoted 8 lines

tools/perf/util/lock-contention.c line 170

Quoted 5 lines
tools/perf/util/lock-contention.c new file +5 -0
+int lock_contention_read(struct lock_contention *con __maybe_unused)
+{
+     return 0;
+}
+#endif  /* !HAVE_BPF_SKEL */

I still think it's the convention to have them in a header file as static inline functions and reduce the #ifdef in the .c file.

Quoted 12 lines

Shouldn't minimizing ifdefs, and associated cognitive load, in header files be the priority given they are #included many times while the .c file is only compiled once?
Shouldn't a goal of the header file be to abstract away things like HAVE_BPF_SKEL?
I'm not clear what the goal of having the functions in the header files is, performance? The code isn't going to run anyway. I feel lock_contention.h is smaller and easier to read like this but I also don't care enough to fight. I did this change here as lock_contention.h was being brought into python.c for the sake of stubbing out functions that the header file was also subbing out for !BPF_HAVE_SKEL. A single stub felt like progress.

Quoted 3 lines

I think it may have the empty functions in the binary if we keep the functions in the .c file whereas compilers would optimize away them if they are static inline functions.

The functions will be 1 or 2 bytes and can all be deduplicated, the linker can also garbage collect them. It is better to optimize code for readability rather than for wins like this.

Thanks,
Ian

author reviewer
#22 lore raw

PATCH v6 16/22 perf bench: Remove reference to cmd_inject

In reply to Ian Rogers

Avoid perf bench internals inject-build-id referencing the cmd_inject sub-command that requires perf-bench to backward reference internals of builtins. Replace the reference to cmd_inject with a call to main. To avoid python.c needing to link with something providing main, drop the libperf-bench library from the python shared object.

---
 tools/perf/Makefile.perf          |  7 +++++--
 tools/perf/bench/inject-buildid.c | 13 +++++++------
 tools/perf/util/python.c          |  6 ------
 3 files changed, 12 insertions(+), 14 deletions(-)
tools/perf/Makefile.perf +5 -2
index d74241a15131..cbcb698cea38 100644
@@ -487,6 +487,9 @@ endif
 EXTLIBS := $(call filter-out,$(EXCLUDE_EXTLIBS),$(EXTLIBS))
 LIBS = -Wl,--whole-archive $(PERFLIBS) $(EXTRA_PERFLIBS) -Wl,--no-whole-archive -Wl,--start-group $(EXTLIBS) -Wl,--end-group
 
+PERFLIBS_PY := $(call filter-out,$(LIBPERF_BENCH),$(PERFLIBS))
+LIBS_PY = -Wl,--whole-archive $(PERFLIBS_PY) $(EXTRA_PERFLIBS) -Wl,--no-whole-archive -Wl,--start-group $(EXTLIBS) -Wl,--end-group
+
 export INSTALL SHELL_PATH
 
 ### Build rules
@@ -735,9 +738,9 @@ all: shell_compatibility_test $(ALL_PROGRAMS) $(LANG_BINDINGS) $(OTHER_PROGRAMS)
 # Create python binding output directory if not already present
 $(shell [ -d '$(OUTPUT)python' ] || mkdir -p '$(OUTPUT)python')
 
-$(OUTPUT)python/perf$(PYTHON_EXTENSION_SUFFIX): util/python.c util/setup.py $(PERFLIBS)
+$(OUTPUT)python/perf$(PYTHON_EXTENSION_SUFFIX): util/python.c util/setup.py $(PERFLIBS_PY)
 	$(QUIET_GEN)LDSHARED="$(CC) -pthread -shared" \
-        CFLAGS='$(CFLAGS)' LDFLAGS='$(LDFLAGS) $(LIBS)' \
+        CFLAGS='$(CFLAGS)' LDFLAGS='$(LDFLAGS) $(LIBS_PY)' \
 	  $(PYTHON_WORD) util/setup.py \
 	  --quiet build_ext; \
 	cp $(PYTHON_EXTBUILD_LIB)perf*.so $(OUTPUT)python/
tools/perf/bench/inject-buildid.c +7 -6
index a759eb2328be..f55c07e4be94 100644
@@ -52,7 +52,7 @@ struct bench_dso {
 static int nr_dsos;
 static struct bench_dso *dsos;
 
-extern int cmd_inject(int argc, const char *argv[]);
+extern int main(int argc, const char **argv);
 
 static const struct option options[] = {
 	OPT_UINTEGER('i', "iterations", &iterations,
@@ -294,7 +294,7 @@ static int setup_injection(struct bench_data *data, bool build_id_all)
 
 	if (data->pid == 0) {
 		const char **inject_argv;
-		int inject_argc = 2;
+		int inject_argc = 3;
 
 		close(data->input_pipe[1]);
 		close(data->output_pipe[0]);
@@ -318,15 +318,16 @@ static int setup_injection(struct bench_data *data, bool build_id_all)
 		if (inject_argv == NULL)
 			exit(1);
 
-		inject_argv[0] = strdup("inject");
-		inject_argv[1] = strdup("-b");
+		inject_argv[0] = strdup("perf");
+		inject_argv[1] = strdup("inject");
+		inject_argv[2] = strdup("-b");
 		if (build_id_all)
-			inject_argv[2] = strdup("--buildid-all");
+			inject_argv[3] = strdup("--buildid-all");
 
 		/* signal that we're ready to go */
 		close(ready_pipe[1]);
 
-		cmd_inject(inject_argc, inject_argv);
+		main(inject_argc, inject_argv);
 
 		exit(0);
 	}
tools/perf/util/python.c +0 -6
index 91fd444615cd..c52da509ae58 100644
@@ -19,7 +19,6 @@
 #include "util/kwork.h"
 #include "util/sample.h"
 #include <internal/lib.h>
-#include "../builtin.h"
 
 #define _PyUnicode_FromString(arg) \
   PyUnicode_FromString(arg)
@@ -1309,8 +1308,3 @@ struct kwork_work *perf_kwork_add_work(struct perf_kwork *kwork __maybe_unused,
 {
 	return NULL;
 }
-
-int cmd_inject(int argc __maybe_unused, const char *argv[] __maybe_unused)
-{
-	return -1;
-}
Signature · 1 line
2.47.0.277.g8800431eea-goog
author reviewer
#23 lore raw

PATCH v6 17/22 perf kwork: Make perf_kwork_add_work a callback

In reply to Ian Rogers

perf_kwork_add_work is declared in builtin-kwork, whereas much kwork code is in util. To avoid needing to stub perf_kwork_add_work in python.c, add a callback to struct perf_kwork and initialize it in builtin-kwork to perf_kwork_add_work - this is the only struct perf_kwork. This removes the need for the stub in python.c.

---
 tools/perf/builtin-kwork.c      |  3 ++-
 tools/perf/util/bpf_kwork.c     |  2 +-
 tools/perf/util/bpf_kwork_top.c |  2 +-
 tools/perf/util/kwork.h         |  6 ++++--
 tools/perf/util/python.c        | 12 ------------
 5 files changed, 8 insertions(+), 17 deletions(-)
tools/perf/builtin-kwork.c +2 -1
index c1daf82c9b92..fabfcb74800b 100644
@@ -1846,7 +1846,7 @@ static void process_skipped_events(struct perf_kwork *kwork,
 	}
 }
 
-struct kwork_work *perf_kwork_add_work(struct perf_kwork *kwork,
+static struct kwork_work *perf_kwork_add_work(struct perf_kwork *kwork,
 				       struct kwork_class *class,
 				       struct kwork_work *key)
 {
@@ -2344,6 +2344,7 @@ int cmd_kwork(int argc, const char **argv)
 		.all_runtime         = 0,
 		.all_count           = 0,
 		.nr_skipped_events   = { 0 },
+		.add_work            = perf_kwork_add_work,
 	};
 	static const char default_report_sort_order[] = "runtime, max, count";
 	static const char default_latency_sort_order[] = "avg, max, count";
tools/perf/util/bpf_kwork.c +1 -1
index 6c7126b7670d..5cff755c71fa 100644
@@ -285,7 +285,7 @@ static int add_work(struct perf_kwork *kwork,
 	    (bpf_trace->get_work_name(key, &tmp.name)))
 		return -1;
 
-	work = perf_kwork_add_work(kwork, tmp.class, &tmp);
+	work = kwork->add_work(kwork, tmp.class, &tmp);
 	if (work == NULL)
 		return -1;
 
tools/perf/util/bpf_kwork_top.c +1 -1
index 7261cad43468..b6f187dd9136 100644
@@ -255,7 +255,7 @@ static int add_work(struct perf_kwork *kwork, struct work_key *key,
 	bpf_trace = kwork_class_bpf_supported_list[type];
 	tmp.class = bpf_trace->class;
 
-	work = perf_kwork_add_work(kwork, tmp.class, &tmp);
+	work = kwork->add_work(kwork, tmp.class, &tmp);
 	if (!work)
 		return -1;
 
tools/perf/util/kwork.h +4 -2
index 76fe2a821bcf..29352ca1d497 100644
@@ -251,12 +251,14 @@ struct perf_kwork {
 	 * perf kwork top data
 	 */
 	struct kwork_top_stat top_stat;
-};
 
-struct kwork_work *perf_kwork_add_work(struct perf_kwork *kwork,
+	/* Add work callback. */
+	struct kwork_work *(*add_work)(struct perf_kwork *kwork,
 				       struct kwork_class *class,
 				       struct kwork_work *key);
 
+};
+
 #ifdef HAVE_BPF_SKEL
 
 int perf_kwork__trace_prepare_bpf(struct perf_kwork *kwork);
tools/perf/util/python.c +0 -12
index c52da509ae58..6e2ff0076daa 100644
@@ -16,7 +16,6 @@
 #include "thread_map.h"
 #include "trace-event.h"
 #include "mmap.h"
-#include "util/kwork.h"
 #include "util/sample.h"
 #include <internal/lib.h>
 
@@ -1297,14 +1296,3 @@ PyMODINIT_FUNC PyInit_perf(void)
 		PyErr_SetString(PyExc_ImportError, "perf: Init failed!");
 	return module;
 }
-
-
-/* The following are stubs to avoid dragging in builtin-* objects. */
-/* TODO: move the code out of the builtin-* file into util. */
-
-struct kwork_work *perf_kwork_add_work(struct perf_kwork *kwork __maybe_unused,
-				       struct kwork_class *class __maybe_unused,
-				       struct kwork_work *key  __maybe_unused)
-{
-	return NULL;
-}
Signature · 1 line
2.47.0.277.g8800431eea-goog
author reviewer
#24 lore raw

PATCH v6 18/22 perf build: Remove test library from python shared object

In reply to Ian Rogers

With the attr.c code moved to a shell test, there is no need to link the test code into the python dso to avoid a missing reference to test_attr__open. Drop the test code from the python library.

With the bench and test code removed from the python library on my x86 debian derived laptop the python library is reduced in size by 508,712 bytes or nearly 5%.

---
 tools/perf/Makefile.perf | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/Makefile.perf +1 -1
index cbcb698cea38..adbc7ab325c7 100644
@@ -487,7 +487,7 @@ endif
 EXTLIBS := $(call filter-out,$(EXCLUDE_EXTLIBS),$(EXTLIBS))
 LIBS = -Wl,--whole-archive $(PERFLIBS) $(EXTRA_PERFLIBS) -Wl,--no-whole-archive -Wl,--start-group $(EXTLIBS) -Wl,--end-group
 
-PERFLIBS_PY := $(call filter-out,$(LIBPERF_BENCH),$(PERFLIBS))
+PERFLIBS_PY := $(call filter-out,$(LIBPERF_BENCH) $(LIBPERF_TEST),$(PERFLIBS))
 LIBS_PY = -Wl,--whole-archive $(PERFLIBS_PY) $(EXTRA_PERFLIBS) -Wl,--no-whole-archive -Wl,--start-group $(EXTLIBS) -Wl,--end-group
 
 export INSTALL SHELL_PATH
Signature · 1 line
2.47.0.277.g8800431eea-goog
author reviewer
#25 lore raw

PATCH v6 19/22 perf python: Add parse_events function

In reply to Ian Rogers

Add basic parse_events function that takes a string and returns an evlist. As the python evlist is embedded in a pyrf_evlist, and the evsels are embedded in pyrf_evsels, copy the parsed data into those structs and update evsel__clone to enable this.

---
 tools/perf/util/cgroup.c |  2 +-
 tools/perf/util/evsel.c  | 19 ++++++++-----
 tools/perf/util/evsel.h  |  2 +-
 tools/perf/util/python.c | 61 ++++++++++++++++++++++++++++++++++++++++
 4 files changed, 75 insertions(+), 9 deletions(-)
tools/perf/util/cgroup.c +1 -1
index 0f759dd96db7..fbcc0626f9ce 100644
@@ -473,7 +473,7 @@ int evlist__expand_cgroup(struct evlist *evlist, const char *str,
 
 		leader = NULL;
 		evlist__for_each_entry(orig_list, pos) {
-			evsel = evsel__clone(pos);
+			evsel = evsel__clone(/*dest=*/NULL, pos);
 			if (evsel == NULL)
 				goto out_err;
 
tools/perf/util/evsel.c +12 -7
index f780e30aa259..4213ffacdc3c 100644
@@ -453,7 +453,7 @@ static int evsel__copy_config_terms(struct evsel *dst, struct evsel *src)
  * The assumption is that @orig is not configured nor opened yet.
  * So we only care about the attributes that can be set while it's parsed.
  */
-struct evsel *evsel__clone(struct evsel *orig)
+struct evsel *evsel__clone(struct evsel *dest, struct evsel *orig)
 {
 	struct evsel *evsel;
 
@@ -466,7 +466,11 @@ struct evsel *evsel__clone(struct evsel *orig)
 	if (orig->bpf_obj)
 		return NULL;
 
-	evsel = evsel__new(&orig->core.attr);
+	if (dest)
+		evsel = dest;
+	else
+		evsel = evsel__new(&orig->core.attr);
+
 	if (evsel == NULL)
 		return NULL;
 
@@ -511,11 +515,12 @@ struct evsel *evsel__clone(struct evsel *orig)
 	evsel->core.leader = orig->core.leader;
 
 	evsel->max_events = orig->max_events;
-	free((char *)evsel->unit);
-	evsel->unit = strdup(orig->unit);
-	if (evsel->unit == NULL)
-		goto out_err;
-
+	zfree(&evsel->unit);
+	if (orig->unit) {
+		evsel->unit = strdup(orig->unit);
+		if (evsel->unit == NULL)
+			goto out_err;
+	}
 	evsel->scale = orig->scale;
 	evsel->snapshot = orig->snapshot;
 	evsel->per_pkg = orig->per_pkg;
tools/perf/util/evsel.h +1 -1
index 04934a7af174..7f41d6f8cf6f 100644
@@ -241,7 +241,7 @@ static inline struct evsel *evsel__new(struct perf_event_attr *attr)
 	return evsel__new_idx(attr, 0);
 }
 
-struct evsel *evsel__clone(struct evsel *orig);
+struct evsel *evsel__clone(struct evsel *dest, struct evsel *orig);
 
 int copy_config_terms(struct list_head *dst, struct list_head *src);
 void free_config_terms(struct list_head *config_terms);
tools/perf/util/python.c +61 -0

79 lines not on this page · raw message

Signature · 1 line
2.47.0.277.g8800431eea-goog
author reviewer
#26 lore raw

PATCH v6 20/22 perf python: Add __str__ and __repr__ functions to evlist

In reply to Ian Rogers

This allows the values in the evlist to be shown in the REPL like:

```
Python 3.11.9 (main, Jun 19 2024, 00:38:48) [GCC 13.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import sys
>>> sys.path.insert(0,'/tmp/perf/python')
>>> import perf
>>> perf.parse_events('cycles,data_read')
evlist([cycles,uncore_imc_free_running_0/data_read/,uncore_imc_free_running_1/data_read/])
```
---
 tools/perf/util/python.c | 26 ++++++++++++++++++++++++++
 1 file changed, 26 insertions(+)
tools/perf/util/python.c +26 -0
index 1a53a97c513d..a71bad3418da 100644
@@ -1069,6 +1069,30 @@ static PyObject *pyrf_evlist__item(PyObject *obj, Py_ssize_t i)
 	return Py_BuildValue("O", container_of(pos, struct pyrf_evsel, evsel));
 }
 
+static PyObject *pyrf_evlist__str(PyObject *self)
+{
+	struct pyrf_evlist *pevlist = (void *)self;
+	struct evsel *pos;
+	struct strbuf sb = STRBUF_INIT;
+	bool first = true;
+	PyObject *result;
+
+	strbuf_addstr(&sb, "evlist([");
+	evlist__for_each_entry(&pevlist->evlist, pos) {
+		if (!first)
+			strbuf_addch(&sb, ',');
+		if (!pos->pmu)
+			strbuf_addstr(&sb, evsel__name(pos));
+		else
+			strbuf_addf(&sb, "%s/%s/", pos->pmu->name, evsel__name(pos));
+		first = false;
+	}
+	strbuf_addstr(&sb, "])");
+	result = PyUnicode_FromString(sb.buf);
+	strbuf_release(&sb);
+	return result;
+}
+
 static PySequenceMethods pyrf_evlist__sequence_methods = {
 	.sq_length = pyrf_evlist__length,
 	.sq_item   = pyrf_evlist__item,
@@ -1086,6 +1110,8 @@ static PyTypeObject pyrf_evlist__type = {
 	.tp_doc		= pyrf_evlist__doc,
 	.tp_methods	= pyrf_evlist__methods,
 	.tp_init	= (initproc)pyrf_evlist__init,
+	.tp_repr        = pyrf_evlist__str,
+	.tp_str         = pyrf_evlist__str,
 };
 
 static int pyrf_evlist__setup_types(void)
Signature · 1 line
2.47.0.277.g8800431eea-goog
author reviewer
#27 lore raw

PATCH v6 21/22 perf python: Add __str__ and __repr__ functions to evsel

In reply to Ian Rogers

This allows evsel to be shown in the REPL like:

```
Python 3.11.9 (main, Jun 19 2024, 00:38:48) [GCC 13.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import sys
>>> sys.path.insert(0, '/tmp/perf/python')
>>> import perf
>>> x=perf.parse_events('cycles,data_read')
>>> print(x)
evlist([cycles,uncore_imc_free_running_0/data_read/,uncore_imc_free_running_1/data_read/])
>>> x[0]
evsel(cycles)
>>> x[1]
evsel(uncore_imc_free_running_0/data_read/)
>>> x[2]
evsel(uncore_imc_free_running_1/data_read/)
```
---
 tools/perf/util/python.c | 13 +++++++++++++
 1 file changed, 13 insertions(+)
tools/perf/util/python.c +13 -0
index a71bad3418da..0d71ec673aa3 100644
@@ -781,6 +781,17 @@ static PyObject *pyrf_evsel__open(struct pyrf_evsel *pevsel,
 	return Py_None;
 }
 
+static PyObject *pyrf_evsel__str(PyObject *self)
+{
+	struct pyrf_evsel *pevsel = (void *)self;
+	struct evsel *evsel = &pevsel->evsel;
+
+	if (!evsel->pmu)
+		return PyUnicode_FromFormat("evsel(%s)", evsel__name(evsel));
+
+	return PyUnicode_FromFormat("evsel(%s/%s/)", evsel->pmu->name, evsel__name(evsel));
+}
+
 static PyMethodDef pyrf_evsel__methods[] = {
 	{
 		.ml_name  = "open",
@@ -802,6 +813,8 @@ static PyTypeObject pyrf_evsel__type = {
 	.tp_doc		= pyrf_evsel__doc,
 	.tp_methods	= pyrf_evsel__methods,
 	.tp_init	= (initproc)pyrf_evsel__init,
+	.tp_str         = pyrf_evsel__str,
+	.tp_repr        = pyrf_evsel__str,
 };
 
 static int pyrf_evsel__setup_types(void)
Signature · 1 line
2.47.0.277.g8800431eea-goog
author reviewer
#28 lore raw

PATCH v6 22/22 perf python: Correctly throw IndexError

In reply to Ian Rogers

Correctly throw IndexError for out-of-bound accesses to evlist:

```
Python 3.11.9 (main, Jun 19 2024, 00:38:48) [GCC 13.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import sys
>>> sys.path.insert(0, '/tmp/perf/python')
>>> import perf
>>> x=perf.parse_events('cycles')
>>> print(x)
evlist([cycles])
>>> x[2]
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
IndexError: Index out of range
```
---
 tools/perf/util/python.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
tools/perf/util/python.c +3 -1
index 0d71ec673aa3..25114dcadd21 100644
@@ -1071,8 +1071,10 @@ static PyObject *pyrf_evlist__item(PyObject *obj, Py_ssize_t i)
 	struct pyrf_evlist *pevlist = (void *)obj;
 	struct evsel *pos;
 
-	if (i >= pevlist->evlist.core.nr_entries)
+	if (i >= pevlist->evlist.core.nr_entries) {
+		PyErr_SetString(PyExc_IndexError, "Index out of range");
 		return NULL;
+	}
 
 	evlist__for_each_entry(&pevlist->evlist, pos) {
 		if (i-- == 0)
Signature · 1 line
2.47.0.277.g8800431eea-goog