gh-98894: Restore function entry/exit DTrace probes (#142397)

The function__entry and function__return probes stopped working in Python 3.11
when the interpreter was restructured around the new bytecode system. This change
restores these probes by adding DTRACE_FUNCTION_ENTRY() at the start_frame label
in bytecodes.c and DTRACE_FUNCTION_RETURN() in the RETURN_VALUE and YIELD_VALUE
instructions. The helper functions are defined in ceval.c and extract the
filename, function name, and line number from the frame before firing the probe.

This builds on the approach from #125019
but avoids modifying the JIT template since the JIT does not currently support
DTrace. The macros are conditionally compiled with WITH_DTRACE and are no-ops
otherwise. The tests have been updated to use modern opcode names (CALL, CALL_KW,
CALL_FUNCTION_EX) and a new bpftrace backend was added for Linux CI alongside
the existing SystemTap tests. Line probe tests were removed since that probe
was never restored after 3.11.

Author: pablogsal

Lib/test/dtracedata/call_stack.py

@@ -5,16 +5,16 @@ def function_1():
 def function_2():
     function_1()
 
-# CALL_FUNCTION_VAR
+# CALL with positional args
 def function_3(dummy, dummy2):
     pass
 
-# CALL_FUNCTION_KW
+# CALL_KW (keyword arguments)
 def function_4(**dummy):
     return 1
     return 2  # unreachable
 
-# CALL_FUNCTION_VAR_KW
+# CALL_FUNCTION_EX (unpacking)
 def function_5(dummy, dummy2, **dummy3):
     if False:
         return 7

Lib/test/dtracedata/call_stack.stp.expected

@@ -1,8 +1,11 @@
-function__entry:call_stack.py:start:23
 function__entry:call_stack.py:function_1:1
+function__entry:call_stack.py:function_3:9
+function__return:call_stack.py:function_3:10
 function__return:call_stack.py:function_1:2
 function__entry:call_stack.py:function_2:5
 function__entry:call_stack.py:function_1:1
+function__entry:call_stack.py:function_3:9
+function__return:call_stack.py:function_3:10
 function__return:call_stack.py:function_1:2
 function__return:call_stack.py:function_2:6
 function__entry:call_stack.py:function_3:9
@@ -11,4 +14,3 @@ function__entry:call_stack.py:function_4:13
 function__return:call_stack.py:function_4:14
 function__entry:call_stack.py:function_5:18
 function__return:call_stack.py:function_5:21
-function__return:call_stack.py:start:28

Lib/test/dtracedata/line.d

@@ -33,11 +33,17 @@ def normalize_trace_output(output):
         result = [
             row.split("\t")
             for row in output.splitlines()
-            if row and not row.startswith('#')
+            if row and not row.startswith('#') and not row.startswith('@')
         ]
         result.sort(key=lambda row: int(row[0]))
         result = [row[1] for row in result]
-        return "\n".join(result)
+        # Normalize paths to basenames (bpftrace outputs full paths)
+        normalized = []
+        for line in result:
+            # Replace full paths with just the filename
+            line = re.sub(r'/[^:]+/([^/:]+\.py)', r'\1', line)
+            normalized.append(line)
+        return "\n".join(normalized)
     except (IndexError, ValueError):
         raise AssertionError(
             "tracer produced unparsable output:\n{}".format(output)
@@ -96,13 +102,186 @@ def assert_usable(self):
 class DTraceBackend(TraceBackend):
     EXTENSION = ".d"
     COMMAND = ["dtrace", "-q", "-s"]
+    if sys.platform == "sunos5":
+        COMMAND.insert(2, "-Z")
 
 
 class SystemTapBackend(TraceBackend):
     EXTENSION = ".stp"
     COMMAND = ["stap", "-g"]
 
 
+class BPFTraceBackend(TraceBackend):
+    EXTENSION = ".bt"
+    COMMAND = ["bpftrace"]
+
+    # Inline bpftrace programs for each test case
+    PROGRAMS = {
+        "call_stack": """
+            usdt:{python}:python:function__entry {{
+                printf("%lld\\tfunction__entry:%s:%s:%d\\n",
+                       nsecs, str(arg0), str(arg1), arg2);
+            }}
+            usdt:{python}:python:function__return {{
+                printf("%lld\\tfunction__return:%s:%s:%d\\n",
+                       nsecs, str(arg0), str(arg1), arg2);
+            }}
+        """,
+        "gc": """
+            usdt:{python}:python:function__entry {{
+                if (str(arg1) == "start") {{ @tracing = 1; }}
+            }}
+            usdt:{python}:python:function__return {{
+                if (str(arg1) == "start") {{ @tracing = 0; }}
+            }}
+            usdt:{python}:python:gc__start {{
+                if (@tracing) {{
+                    printf("%lld\\tgc__start:%d\\n", nsecs, arg0);
+                }}
+            }}
+            usdt:{python}:python:gc__done {{
+                if (@tracing) {{
+                    printf("%lld\\tgc__done:%lld\\n", nsecs, arg0);
+                }}
+            }}
+            END {{ clear(@tracing); }}
+        """,
+    }
+
+    # Which test scripts to filter by filename (None = use @tracing flag)
+    FILTER_BY_FILENAME = {"call_stack": "call_stack.py"}
+
+    @staticmethod
+    def _filter_probe_rows(output):
+        return "\n".join(
+            line for line in output.splitlines()
+            if line.partition("\t")[0].isdigit()
+        )
+
+    # Expected outputs for each test case
+    # Note: bpftrace captures <module> entry/return and may have slight timing
+    # differences compared to SystemTap due to probe firing order
+    EXPECTED = {
+        "call_stack": """function__entry:call_stack.py:<module>:0
+function__entry:call_stack.py:start:23
+function__entry:call_stack.py:function_1:1
+function__entry:call_stack.py:function_3:9
+function__return:call_stack.py:function_3:10
+function__return:call_stack.py:function_1:2
+function__entry:call_stack.py:function_2:5
+function__entry:call_stack.py:function_1:1
+function__entry:call_stack.py:function_3:9
+function__return:call_stack.py:function_3:10
+function__return:call_stack.py:function_1:2
+function__return:call_stack.py:function_2:6
+function__entry:call_stack.py:function_3:9
+function__return:call_stack.py:function_3:10
+function__entry:call_stack.py:function_4:13
+function__return:call_stack.py:function_4:14
+function__entry:call_stack.py:function_5:18
+function__return:call_stack.py:function_5:21
+function__return:call_stack.py:start:28
+function__return:call_stack.py:<module>:30""",
+        "gc": """gc__start:0
+gc__done:0
+gc__start:1
+gc__done:0
+gc__start:2
+gc__done:0
+gc__start:2
+gc__done:1""",
+    }
+
+    def run_case(self, name, optimize_python=None):
+        if name not in self.PROGRAMS:
+            raise unittest.SkipTest(f"No bpftrace program for {name}")
+
+        python_file = abspath(name + ".py")
+        python_flags = []
+        if optimize_python:
+            python_flags.extend(["-O"] * optimize_python)
+
+        subcommand = [sys.executable] + python_flags + [python_file]
+        program = self.PROGRAMS[name].format(python=sys.executable)
+
+        try:
+            proc = subprocess.Popen(
+                ["bpftrace", "-e", program, "-c", " ".join(subcommand)],
+                stdout=subprocess.PIPE,
+                stderr=subprocess.PIPE,
+                universal_newlines=True,
+            )
+            stdout, stderr = proc.communicate(timeout=60)
+        except subprocess.TimeoutExpired:
+            proc.kill()
+            raise AssertionError("bpftrace timed out")
+        except (FileNotFoundError, PermissionError) as e:
+            raise unittest.SkipTest(f"bpftrace not available: {e}")
+
+        if proc.returncode != 0:
+            raise AssertionError(
+                f"bpftrace failed with code {proc.returncode}:\n{stderr}"
+            )
+
+        stdout = self._filter_probe_rows(stdout)
+
+        # Filter output by filename if specified (bpftrace captures everything)
+        if name in self.FILTER_BY_FILENAME:
+            filter_filename = self.FILTER_BY_FILENAME[name]
+            filtered_lines = [
+                line for line in stdout.splitlines()
+                if filter_filename in line
+            ]
+            stdout = "\n".join(filtered_lines)
+
+        actual_output = normalize_trace_output(stdout)
+        expected_output = self.EXPECTED[name].strip()
+
+        return (expected_output, actual_output)
+
+    def assert_usable(self):
+        # Check if bpftrace is available and can attach to USDT probes
+        program = f'usdt:{sys.executable}:python:function__entry {{ printf("probe: success\\n"); exit(); }}'
+        try:
+            proc = subprocess.Popen(
+                ["bpftrace", "-e", program, "-c", f"{sys.executable} -c pass"],
+                stdout=subprocess.PIPE,
+                stderr=subprocess.PIPE,
+                universal_newlines=True,
+            )
+            stdout, stderr = proc.communicate(timeout=10)
+        except subprocess.TimeoutExpired:
+            proc.kill()
+            proc.communicate()  # Clean up
+            raise unittest.SkipTest("bpftrace timed out during usability check")
+        except OSError as e:
+            raise unittest.SkipTest(f"bpftrace not available: {e}")
+
+        # Check for permission errors (bpftrace usually requires root)
+        if proc.returncode != 0:
+            raise unittest.SkipTest(
+                f"bpftrace(1) failed with code {proc.returncode}: {stderr}"
+            )
+
+        if "probe: success" not in stdout:
+            raise unittest.SkipTest(
+                f"bpftrace(1) failed: stdout={stdout!r} stderr={stderr!r}"
+            )
+
+
+class BPFTraceOutputTests(unittest.TestCase):
+    def test_filter_probe_rows_ignores_warnings(self):
+        output = """stdin:1-19: WARNING: found external warnings
+HINT: include/vmlinux.h:1439:3: warning: declaration does not declare anything
+4623214882928\tgc__start:0
+4623214885575\tgc__done:0
+"""
+        self.assertEqual(
+            BPFTraceBackend._filter_probe_rows(output),
+            "4623214882928\tgc__start:0\n4623214885575\tgc__done:0",
+        )
+
+
 @unittest.skipIf(MS_WINDOWS, "Tests not compliant with trace on Windows.")
 class TraceTests:
     # unittest.TestCase options
@@ -127,7 +306,8 @@ def test_function_entry_return(self):
     def test_verify_call_opcodes(self):
         """Ensure our call stack test hits all function call opcodes"""
 
-        opcodes = set(["CALL_FUNCTION", "CALL_FUNCTION_EX", "CALL_FUNCTION_KW"])
+        # Modern Python uses CALL, CALL_KW, and CALL_FUNCTION_EX
+        opcodes = set(["CALL", "CALL_FUNCTION_EX", "CALL_KW"])
 
         with open(abspath("call_stack.py")) as f:
             code_string = f.read()
@@ -152,9 +332,6 @@ def get_function_instructions(funcname):
     def test_gc(self):
         self.run_case("gc")
 
-    def test_line(self):
-        self.run_case("line")
-
 
 class DTraceNormalTests(TraceTests, unittest.TestCase):
     backend = DTraceBackend()
@@ -175,6 +352,17 @@ class SystemTapOptimizedTests(TraceTests, unittest.TestCase):
     backend = SystemTapBackend()
     optimize_python = 2
 
+
+class BPFTraceNormalTests(TraceTests, unittest.TestCase):
+    backend = BPFTraceBackend()
+    optimize_python = 0
+
+
+class BPFTraceOptimizedTests(TraceTests, unittest.TestCase):
+    backend = BPFTraceBackend()
+    optimize_python = 2
+
+
 class CheckDtraceProbes(unittest.TestCase):
     @classmethod
     def setUpClass(cls):
@@ -235,6 +423,8 @@ def test_check_probes(self):
             "Name: audit",
             "Name: gc__start",
             "Name: gc__done",
+            "Name: function__entry",
+            "Name: function__return",
         ]
 
         for probe_name in available_probe_names:
@@ -247,8 +437,6 @@ def test_missing_probes(self):
 
         # Missing probes will be added in the future.
         missing_probe_names = [
-            "Name: function__entry",
-            "Name: function__return",
             "Name: line",
         ]
 

Lib/test/dtracedata/line.d.expected

@@ -0,0 +1,2 @@
+Restore ``function__entry`` and ``function__return`` DTrace/SystemTap probes
+that were broken since Python 3.11.