DTrace support: function calls, GC activity, line execution

Tested on macOS 10.11 dtrace, Ubuntu 16.04 SystemTap, and libbcc.

Largely based by an initial patch by Jesús Cea Avión, with some
influence from Dave Malcolm's SystemTap patch and Nikhil Benesch's
unification patch.

Things deliberately left out for simplicity:
- ustack helpers, I have no way of testing them at this point since
they are Solaris-specific
- PyFrameObject * in function__entry/function__return, this is
SystemTap-specific
- SPARC support
- dynamic tracing
- sys module dtrace facility introspection

All of those might be added later.

31 files changed, 1305 insertions, 18 deletions

diff --git a/.hgignore b/.hgignore
index a3ffcf4..f5d1414 100644
--- a/.hgignore
+++ b/.hgignore
@@ -56,6 +56,7 @@ libpython*.dylib
 *.profclang?
 *.profraw
 *.dyn
+Include/pydtrace_probes.h
 Lib/distutils/command/*.pdb
 Lib/lib2to3/*.pickle
 Lib/test/data/*
diff --git a/Doc/howto/index.rst b/Doc/howto/index.rst
index de65950..593341c 100644
--- a/Doc/howto/index.rst
+++ b/Doc/howto/index.rst
@@ -28,4 +28,5 @@ Currently, the HOWTOs are:
    argparse.rst
    ipaddress.rst
    clinic.rst
+   instrumentation.rst
 
diff --git a/Doc/howto/instrumentation.rst b/Doc/howto/instrumentation.rst
new file mode 100644
index 0000000..2b21224
--- /dev/null
+++ b/Doc/howto/instrumentation.rst
@@ -0,0 +1,411 @@
+.. _instrumentation:
+
+===============================================
+Instrumenting CPython with DTrace and SystemTap
+===============================================
+
+:author: David Malcolm
+:author: Łukasz Langa
+
+DTrace and SystemTap are monitoring tools, each providing a way to inspect
+what the processes on a computer system are doing.  They both use
+domain-specific languages allowing a user to write scripts which:
+
+  - filter which processes are to be observed
+  - gather data from the processes of interest
+  - generate reports on the data
+
+As of Python 3.6, CPython can be built with embedded "markers", also
+known as "probes", that can be observed by a DTrace or SystemTap script,
+making it easier to monitor what the CPython processes on a system are
+doing.
+
+.. I'm using ".. code-block:: c" for SystemTap scripts, as "c" is syntactically
+   the closest match that Sphinx supports
+
+.. impl-detail::
+
+   DTrace markers are implementation details of the CPython interpreter.
+   No guarantees are made about probe compatibility between versions of
+   CPython. DTrace scripts can stop working or work incorrectly without
+   warning when changing CPython versions.
+
+
+Enabling the static markers
+---------------------------
+
+macOS comes with built-in support for DTrace.  On Linux, in order to
+build CPython with the embedded markers for SystemTap, the SystemTap
+development tools must be installed.
+
+On a Linux machine, this can be done via::
+
+   yum install systemtap-sdt-devel
+
+or::
+
+   sudo apt-get install systemtap-sdt-dev
+
+
+CPython must then be configured `--with-dtrace`::
+
+   checking for --with-dtrace... yes
+
+On macOS, you can list available DTrace probes by running a Python
+process in the background and listing all probes made available by the
+Python provider::
+
+   $ python3.6 -q &
+   $ sudo dtrace -l -P python$!  # or: dtrace -l -m python3.6
+
+      ID   PROVIDER            MODULE                          FUNCTION NAME
+   29564 python18035        python3.6          _PyEval_EvalFrameDefault function-entry
+   29565 python18035        python3.6             dtrace_function_entry function-entry
+   29566 python18035        python3.6          _PyEval_EvalFrameDefault function-return
+   29567 python18035        python3.6            dtrace_function_return function-return
+   29568 python18035        python3.6                           collect gc-done
+   29569 python18035        python3.6                           collect gc-start
+   29570 python18035        python3.6          _PyEval_EvalFrameDefault line
+   29571 python18035        python3.6                 maybe_dtrace_line line
+
+On Linux, you can verify if the SystemTap static markers are present in
+the built binary by seeing if it contains a ".note.stapsdt" section.
+
+.. code-block:: bash
+
+   $ readelf -S ./python | grep .note.stapsdt
+   [30] .note.stapsdt        NOTE         0000000000000000 00308d78
+
+If you've built Python as a shared library (with --enable-shared), you
+need to look instead within the shared library.  For example:
+
+.. code-block:: bash
+
+   $ readelf -S libpython3.3dm.so.1.0 | grep .note.stapsdt
+   [29] .note.stapsdt        NOTE         0000000000000000 00365b68
+
+Sufficiently modern readelf can print the metadata:
+
+.. code-block:: bash
+
+    $ readelf -n ./python
+
+    Displaying notes found at file offset 0x00000254 with length 0x00000020:
+        Owner                 Data size          Description
+        GNU                  0x00000010          NT_GNU_ABI_TAG (ABI version tag)
+            OS: Linux, ABI: 2.6.32
+
+    Displaying notes found at file offset 0x00000274 with length 0x00000024:
+        Owner                 Data size          Description
+        GNU                  0x00000014          NT_GNU_BUILD_ID (unique build ID bitstring)
+            Build ID: df924a2b08a7e89f6e11251d4602022977af2670
+
+    Displaying notes found at file offset 0x002d6c30 with length 0x00000144:
+        Owner                 Data size          Description
+        stapsdt              0x00000031          NT_STAPSDT (SystemTap probe descriptors)
+            Provider: python
+            Name: gc__start
+            Location: 0x00000000004371c3, Base: 0x0000000000630ce2, Semaphore: 0x00000000008d6bf6
+            Arguments: -4@%ebx
+        stapsdt              0x00000030          NT_STAPSDT (SystemTap probe descriptors)
+            Provider: python
+            Name: gc__done
+            Location: 0x00000000004374e1, Base: 0x0000000000630ce2, Semaphore: 0x00000000008d6bf8
+            Arguments: -8@%rax
+        stapsdt              0x00000045          NT_STAPSDT (SystemTap probe descriptors)
+            Provider: python
+            Name: function__entry
+            Location: 0x000000000053db6c, Base: 0x0000000000630ce2, Semaphore: 0x00000000008d6be8
+            Arguments: 8@%rbp 8@%r12 -4@%eax
+        stapsdt              0x00000046          NT_STAPSDT (SystemTap probe descriptors)
+            Provider: python
+            Name: function__return
+            Location: 0x000000000053dba8, Base: 0x0000000000630ce2, Semaphore: 0x00000000008d6bea
+            Arguments: 8@%rbp 8@%r12 -4@%eax
+
+The above metadata contains information for SystemTap describing how it
+can patch strategically-placed machine code instructions to enable the
+tracing hooks used by a SystemTap script.
+
+
+Static DTrace probes
+--------------------
+
+The following example DTrace script can be used to show the call/return
+hierarchy of a Python script, only tracing within the invocation of
+a function called "start". In other words, import-time function
+invocations are not going to be listed:
+
+.. code-block:: c
+
+    self int indent;
+
+    python$target:::function-entry
+    /copyinstr(arg1) == "start"/
+    {
+            self->trace = 1;
+    }
+
+    python$target:::function-entry
+    /self->trace/
+    {
+            printf("%d\t%*s:", timestamp, 15, probename);
+            printf("%*s", self->indent, "");
+            printf("%s:%s:%d\n", basename(copyinstr(arg0)), copyinstr(arg1), arg2);
+            self->indent++;
+    }
+
+    python$target:::function-return
+    /self->trace/
+    {
+            self->indent--;
+            printf("%d\t%*s:", timestamp, 15, probename);
+            printf("%*s", self->indent, "");
+            printf("%s:%s:%d\n", basename(copyinstr(arg0)), copyinstr(arg1), arg2);
+    }
+
+    python$target:::function-return
+    /copyinstr(arg1) == "start"/
+    {
+            self->trace = 0;
+    }
+
+It can be invoked like this:
+
+.. code-block:: bash
+
+  $ sudo dtrace -q -s call_stack.d -c "python3.6 script.py"
+
+The output looks like this::
+
+    156641360502280  function-entry:call_stack.py:start:23
+    156641360518804  function-entry: call_stack.py:function_1:1
+    156641360532797  function-entry:  call_stack.py:function_3:9
+    156641360546807 function-return:  call_stack.py:function_3:10
+    156641360563367 function-return: call_stack.py:function_1:2
+    156641360578365  function-entry: call_stack.py:function_2:5
+    156641360591757  function-entry:  call_stack.py:function_1:1
+    156641360605556  function-entry:   call_stack.py:function_3:9
+    156641360617482 function-return:   call_stack.py:function_3:10
+    156641360629814 function-return:  call_stack.py:function_1:2
+    156641360642285 function-return: call_stack.py:function_2:6
+    156641360656770  function-entry: call_stack.py:function_3:9
+    156641360669707 function-return: call_stack.py:function_3:10
+    156641360687853  function-entry: call_stack.py:function_4:13
+    156641360700719 function-return: call_stack.py:function_4:14
+    156641360719640  function-entry: call_stack.py:function_5:18
+    156641360732567 function-return: call_stack.py:function_5:21
+    156641360747370 function-return:call_stack.py:start:28
+
+
+Static SystemTap markers
+------------------------
+
+The low-level way to use the SystemTap integration is to use the static
+markers directly.  This requires you to explicitly state the binary file
+containing them.
+
+For example, this SystemTap script can be used to show the call/return
+hierarchy of a Python script:
+
+.. code-block:: c
+
+   probe process('python').mark("function__entry") {
+        filename = user_string($arg1);
+        funcname = user_string($arg2);
+        lineno = $arg3;
+
+        printf("%s => %s in %s:%d\\n",
+               thread_indent(1), funcname, filename, lineno);
+   }
+
+   probe process('python').mark("function__return") {
+       filename = user_string($arg1);
+       funcname = user_string($arg2);
+       lineno = $arg3;
+
+       printf("%s <= %s in %s:%d\\n",
+              thread_indent(-1), funcname, filename, lineno);
+   }
+
+It can be invoked like this:
+
+.. code-block:: bash
+
+   $ stap \
+     show-call-hierarchy.stp \
+     -c ./python test.py
+
+The output looks like this::
+
+   11408 python(8274):        => __contains__ in Lib/_abcoll.py:362
+   11414 python(8274):         => __getitem__ in Lib/os.py:425
+   11418 python(8274):          => encode in Lib/os.py:490
+   11424 python(8274):          <= encode in Lib/os.py:493
+   11428 python(8274):         <= __getitem__ in Lib/os.py:426
+   11433 python(8274):        <= __contains__ in Lib/_abcoll.py:366
+
+where the columns are:
+
+  - time in microseconds since start of script
+
+  - name of executable
+
+  - PID of process
+
+and the remainder indicates the call/return hierarchy as the script executes.
+
+For a `--enable-shared` build of CPython, the markers are contained within the
+libpython shared library, and the probe's dotted path needs to reflect this. For
+example, this line from the above example::
+
+   probe process('python').mark("function__entry") {
+
+should instead read::
+
+   probe process('python').library("libpython3.6dm.so.1.0").mark("function__entry") {
+
+(assuming a debug build of CPython 3.6)
+
+
+Available static markers
+------------------------
+
+.. I'm reusing the "c:function" type for markers
+
+.. c:function:: function__entry(str filename, str funcname, int lineno)
+
+   This marker indicates that execution of a Python function has begun.
+   It is only triggered for pure-Python (bytecode) functions.
+
+   The filename, function name, and line number are provided back to the
+   tracing script as positional arguments, which must be accessed using
+   `$arg1`, `$arg2`, `$arg3`:
+
+       * `$arg1` : `(const char *)` filename, accessible using `user_string($arg1)`
+
+       * `$arg2` : `(const char *)` function name, accessible using
+         `user_string($arg2)`
+
+       * `$arg3` : `int` line number
+
+.. c:function:: function__return(str filename, str funcname, int lineno)
+
+   This marker is the converse of `function__entry`, and indicates that
+   execution of a Python function has ended (either via ``return``, or
+   via an exception).  It is only triggered for pure-Python (bytecode)
+   functions.
+
+   The arguments are the same as for `function__entry`
+
+.. c:function:: line(str filename, str funcname, int lineno)
+
+   This marker indicates a Python line is about to be executed.  It is
+   the equivalent of line-by-line tracing with a Python profiler.  It is
+   not triggered within C functions.
+
+   The arguments are the same as for `function__entry`.
+
+.. c:function:: gc__start(int generation)
+
+   Fires when the Python interpreter starts a garbage collection cycle.
+   `arg0` is the generation to scan, like :func:`gc.collect()`.
+
+.. c:function:: gc__done(long collected)
+
+   Fires when the Python interpreter finishes a garbage collection
+   cycle. `arg0` is the number of collected objects.
+
+
+SystemTap Tapsets
+-----------------
+
+The higher-level way to use the SystemTap integration is to use a "tapset":
+SystemTap's equivalent of a library, which hides some of the lower-level
+details of the static markers.
+
+Here is a tapset file, based on a non-shared build of CPython:
+
+.. code-block:: c
+
+    /*
+       Provide a higher-level wrapping around the function__entry and
+       function__return markers:
+     \*/
+    probe python.function.entry = process("python").mark("function__entry")
+    {
+        filename = user_string($arg1);
+        funcname = user_string($arg2);
+        lineno = $arg3;
+        frameptr = $arg4
+    }
+    probe python.function.return = process("python").mark("function__return")
+    {
+        filename = user_string($arg1);
+        funcname = user_string($arg2);
+        lineno = $arg3;
+        frameptr = $arg4
+    }
+
+If this file is installed in SystemTap's tapset directory (e.g.
+`/usr/share/systemtap/tapset`), then these additional probepoints become
+available:
+
+.. c:function:: python.function.entry(str filename, str funcname, int lineno, frameptr)
+
+   This probe point indicates that execution of a Python function has begun.
+   It is only triggered for pure-python (bytecode) functions.
+
+.. c:function:: python.function.return(str filename, str funcname, int lineno, frameptr)
+
+   This probe point is the converse of `python.function.return`, and indicates
+   that execution of a Python function has ended (either via ``return``, or
+   via an exception).  It is only triggered for pure-python (bytecode) functions.
+
+
+Examples
+--------
+This SystemTap script uses the tapset above to more cleanly implement the
+example given above of tracing the Python function-call hierarchy, without
+needing to directly name the static markers:
+
+.. code-block:: c
+
+    probe python.function.entry
+    {
+      printf("%s => %s in %s:%d\n",
+             thread_indent(1), funcname, filename, lineno);
+    }
+
+    probe python.function.return
+    {
+      printf("%s <= %s in %s:%d\n",
+             thread_indent(-1), funcname, filename, lineno);
+    }
+
+
+The following script uses the tapset above to provide a top-like view of all
+running CPython code, showing the top 20 most frequently-entered bytecode
+frames, each second, across the whole system:
+
+.. code-block:: c
+
+    global fn_calls;
+
+    probe python.function.entry
+    {
+        fn_calls[pid(), filename, funcname, lineno] += 1;
+    }
+
+    probe timer.ms(1000) {
+        printf("\033[2J\033[1;1H") /* clear screen \*/
+        printf("%6s %80s %6s %30s %6s\n",
+               "PID", "FILENAME", "LINE", "FUNCTION", "CALLS")
+        foreach ([pid, filename, funcname, lineno] in fn_calls- limit 20) {
+            printf("%6d %80s %6d %30s %6d\n",
+                pid, filename, lineno, funcname,
+                fn_calls[pid, filename, funcname, lineno]);
+        }
+        delete fn_calls;
+    }
+
diff --git a/Doc/whatsnew/3.6.rst b/Doc/whatsnew/3.6.rst
index d3dd337..7b4d0b3 100644
--- a/Doc/whatsnew/3.6.rst
+++ b/Doc/whatsnew/3.6.rst
@@ -407,6 +407,31 @@ Example of fatal error on buffer overflow using
 (Contributed by Victor Stinner in :issue:`26516` and :issue:`26564`.)
 
 
+DTrace and SystemTap probing support
+------------------------------------
+
+Python can now be built ``--with-dtrace`` which enables static markers
+for the following events in the interpreter:
+
+* function call/return
+
+* garbage collection started/finished
+
+* line of code executed.
+
+This can be used to instrument running interpreters in production,
+without the need to recompile specific debug builds or providing
+application-specific profiling/debugging code.
+
+More details in:ref:`instrumentation`.
+
+The current implementation is tested on Linux and macOS.  Additional
+markers may be added in the future.
+
+(Contributed by Łukasz Langa in :issue:`21590`, based on patches by
+Jesús Cea Avión, David Malcolm, and Nikhil Benesch.)
+
+
 .. _whatsnew-deforder:
 
 PEP 520: Preserving Class Attribute Definition Order
diff --git a/Include/pydtrace.d b/Include/pydtrace.d
new file mode 100644
index 0000000..8836055
--- /dev/null
+++ b/Include/pydtrace.d
@@ -0,0 +1,19 @@
+/* Python DTrace provider */
+
+provider python {
+    probe function__entry(const char *, const char *, int);
+    probe function__return(const char *, const char *, int);
+    probe instance__new__start(const char *, const char *);
+    probe instance__new__done(const char *, const char *);
+    probe instance__delete__start(const char *, const char *);
+    probe instance__delete__done(const char *, const char *);
+    probe line(const char *, const char *, int);
+    probe gc__start(int);
+    probe gc__done(long);
+};
+
+#pragma D attributes Evolving/Evolving/Common provider python provider
+#pragma D attributes Evolving/Evolving/Common provider python module
+#pragma D attributes Evolving/Evolving/Common provider python function
+#pragma D attributes Evolving/Evolving/Common provider python name
+#pragma D attributes Evolving/Evolving/Common provider python args
diff --git a/Include/pydtrace.h b/Include/pydtrace.h
new file mode 100644
index 0000000..4c06d0e
--- /dev/null
+++ b/Include/pydtrace.h
@@ -0,0 +1,47 @@
+/* Static DTrace probes interface */
+
+#ifndef Py_DTRACE_H
+#define Py_DTRACE_H
+
+#ifdef WITH_DTRACE
+
+#include "pydtrace_probes.h"
+
+/* pydtrace_probes.h, on systems with DTrace, is auto-generated to include
+   `PyDTrace_{PROBE}` and `PyDTrace_{PROBE}_ENABLED()` macros for every probe
+   defined in pydtrace_provider.d.
+
+   Calling these functions must be guarded by a `PyDTrace_{PROBE}_ENABLED()`
+   check to minimize performance impact when probing is off. For example:
+
+       if (PyDTrace_FUNCTION_ENTRY_ENABLED())
+           PyDTrace_FUNCTION_ENTRY(f);
+*/
+
+#else
+
+/* Without DTrace, compile to nothing. */
+
+#define PyDTrace_LINE(arg0, arg1, arg2, arg3) do ; while (0)
+#define PyDTrace_FUNCTION_ENTRY(arg0, arg1, arg2)  do ; while (0)
+#define PyDTrace_FUNCTION_RETURN(arg0, arg1, arg2) do ; while (0)
+#define PyDTrace_GC_START(arg0)               do ; while (0)
+#define PyDTrace_GC_DONE(arg0)                do ; while (0)
+#define PyDTrace_INSTANCE_NEW_START(arg0)     do ; while (0)
+#define PyDTrace_INSTANCE_NEW_DONE(arg0)      do ; while (0)
+#define PyDTrace_INSTANCE_DELETE_START(arg0)  do ; while (0)
+#define PyDTrace_INSTANCE_DELETE_DONE(arg0)   do ; while (0)
+
+#define PyDTrace_LINE_ENABLED()                  (0)
+#define PyDTrace_FUNCTION_ENTRY_ENABLED()        (0)
+#define PyDTrace_FUNCTION_RETURN_ENABLED()       (0)
+#define PyDTrace_GC_START_ENABLED()              (0)
+#define PyDTrace_GC_DONE_ENABLED()               (0)
+#define PyDTrace_INSTANCE_NEW_START_ENABLED()    (0)
+#define PyDTrace_INSTANCE_NEW_DONE_ENABLED()     (0)
+#define PyDTrace_INSTANCE_DELETE_START_ENABLED() (0)
+#define PyDTrace_INSTANCE_DELETE_DONE_ENABLED()  (0)
+
+#endif /* !WITH_DTRACE */
+
+#endif /* !Py_DTRACE_H */
diff --git a/Lib/test/dtracedata/assert_usable.d b/Lib/test/dtracedata/assert_usable.d
new file mode 100644
index 0000000..0b2d4da
--- /dev/null
+++ b/Lib/test/dtracedata/assert_usable.d
@@ -0,0 +1,5 @@
+BEGIN
+{
+    printf("probe: success\n");
+    exit(0);
+}
diff --git a/Lib/test/dtracedata/assert_usable.stp b/Lib/test/dtracedata/assert_usable.stp
new file mode 100644
index 0000000..88e7e68
--- /dev/null
+++ b/Lib/test/dtracedata/assert_usable.stp
@@ -0,0 +1,5 @@
+probe begin
+{
+  println("probe: success")
+  exit ()
+}
diff --git a/Lib/test/dtracedata/call_stack.d b/Lib/test/dtracedata/call_stack.d
new file mode 100644
index 0000000..450e939
--- /dev/null
+++ b/Lib/test/dtracedata/call_stack.d
@@ -0,0 +1,31 @@
+self int indent;
+
+python$target:::function-entry
+/copyinstr(arg1) == "start"/
+{
+    self->trace = 1;
+}
+
+python$target:::function-entry
+/self->trace/
+{
+    printf("%d\t%*s:", timestamp, 15, probename);
+    printf("%*s", self->indent, ""); 
+    printf("%s:%s:%d\n", basename(copyinstr(arg0)), copyinstr(arg1), arg2);
+    self->indent++;
+}
+
+python$target:::function-return
+/self->trace/
+{
+    self->indent--;
+    printf("%d\t%*s:", timestamp, 15, probename);
+    printf("%*s", self->indent, ""); 
+    printf("%s:%s:%d\n", basename(copyinstr(arg0)), copyinstr(arg1), arg2);
+}
+
+python$target:::function-return
+/copyinstr(arg1) == "start"/
+{
+    self->trace = 0;
+}
diff --git a/Lib/test/dtracedata/call_stack.d.expected b/Lib/test/dtracedata/call_stack.d.expected
new file mode 100644
index 0000000..27849d1
--- /dev/null
+++ b/Lib/test/dtracedata/call_stack.d.expected
@@ -0,0 +1,18 @@
+ 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
diff --git a/Lib/test/dtracedata/call_stack.py b/Lib/test/dtracedata/call_stack.py
new file mode 100644
index 0000000..ee9f3ae
--- /dev/null
+++ b/Lib/test/dtracedata/call_stack.py
@@ -0,0 +1,30 @@
+def function_1():
+    function_3(1, 2)
+
+# Check stacktrace
+def function_2():
+    function_1()
+
+# CALL_FUNCTION_VAR
+def function_3(dummy, dummy2):
+    pass
+
+# CALL_FUNCTION_KW
+def function_4(**dummy):
+    return 1
+    return 2  # unreachable
+
+# CALL_FUNCTION_VAR_KW
+def function_5(dummy, dummy2, **dummy3):
+    if False:
+        return 7
+    return 8
+
+def start():
+    function_1()
+    function_2()
+    function_3(1, 2)
+    function_4(test=42)
+    function_5(*(1, 2), **{"test": 42})
+
+start()
diff --git a/Lib/test/dtracedata/call_stack.stp b/Lib/test/dtracedata/call_stack.stp
new file mode 100644
index 0000000..54082c2
--- /dev/null
+++ b/Lib/test/dtracedata/call_stack.stp
@@ -0,0 +1,41 @@
+global tracing
+
+function basename:string(path:string)
+{
+    last_token = token = tokenize(path, "/");
+    while (token != "") {
+        last_token = token;
+        token = tokenize("", "/");
+    }
+    return last_token;
+}
+
+probe process.mark("function__entry")
+{
+    funcname = user_string($arg2);
+
+    if (funcname == "start") {
+        tracing = 1;
+    }
+}
+
+probe process.mark("function__entry"), process.mark("function__return")
+{
+    filename = user_string($arg1);
+    funcname = user_string($arg2);
+    lineno = $arg3;
+
+    if (tracing) {
+        printf("%d\t%s:%s:%s:%d\n", gettimeofday_us(), $$name,
+               basename(filename), funcname, lineno);
+    }
+}
+
+probe process.mark("function__return")
+{
+    funcname = user_string($arg2);
+
+    if (funcname == "start") {
+        tracing = 0;
+    }
+}
diff --git a/Lib/test/dtracedata/call_stack.stp.expected b/Lib/test/dtracedata/call_stack.stp.expected
new file mode 100644
index 0000000..32cf396
--- /dev/null
+++ b/Lib/test/dtracedata/call_stack.stp.expected
@@ -0,0 +1,14 @@
+function__entry:call_stack.py:start:23
+function__entry:call_stack.py:function_1:1
+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__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
diff --git a/Lib/test/dtracedata/gc.d b/Lib/test/dtracedata/gc.d
new file mode 100644
index 0000000..4d91487
--- /dev/null
+++ b/Lib/test/dtracedata/gc.d
@@ -0,0 +1,18 @@
+python$target:::function-entry
+/copyinstr(arg1) == "start"/
+{
+    self->trace = 1;
+}
+
+python$target:::gc-start,
+python$target:::gc-done
+/self->trace/
+{
+    printf("%d\t%s:%ld\n", timestamp, probename, arg0);
+}
+
+python$target:::function-return
+/copyinstr(arg1) == "start"/
+{
+    self->trace = 0;
+}
diff --git a/Lib/test/dtracedata/gc.d.expected b/Lib/test/dtracedata/gc.d.expected
new file mode 100644
index 0000000..8e5ac2a
--- /dev/null
+++ b/Lib/test/dtracedata/gc.d.expected
@@ -0,0 +1,8 @@
+gc-start:0
+gc-done:0
+gc-start:1
+gc-done:0
+gc-start:2
+gc-done:0
+gc-start:2
+gc-done:1
diff --git a/Lib/test/dtracedata/gc.py b/Lib/test/dtracedata/gc.py
new file mode 100644
index 0000000..144a783
--- /dev/null
+++ b/Lib/test/dtracedata/gc.py
@@ -0,0 +1,13 @@
+import gc
+
+def start():
+    gc.collect(0)
+    gc.collect(1)
+    gc.collect(2)
+    l = []
+    l.append(l)
+    del l
+    gc.collect(2)
+
+gc.collect()
+start()
diff --git a/Lib/test/dtracedata/gc.stp b/Lib/test/dtracedata/gc.stp
new file mode 100644
index 0000000..162c6d3
--- /dev/null
+++ b/Lib/test/dtracedata/gc.stp
@@ -0,0 +1,26 @@
+global tracing
+
+probe process.mark("function__entry")
+{
+    funcname = user_string($arg2);
+
+    if (funcname == "start") {
+        tracing = 1;
+    }
+}
+
+probe process.mark("gc__start"), process.mark("gc__done")
+{
+    if (tracing) {
+        printf("%d\t%s:%ld\n", gettimeofday_us(), $$name, $arg1);
+    }
+}
+
+probe process.mark("function__return")
+{
+    funcname = user_string($arg2);
+
+    if (funcname == "start") {
+        tracing = 0;
+    }
+}
diff --git a/Lib/test/dtracedata/gc.stp.expected b/Lib/test/dtracedata/gc.stp.expected
new file mode 100644
index 0000000..7e6e622
--- /dev/null
+++ b/Lib/test/dtracedata/gc.stp.expected
@@ -0,0 +1,8 @@
+gc__start:0
+gc__done:0
+gc__start:1
+gc__done:0
+gc__start:2
+gc__done:0
+gc__start:2
+gc__done:1
diff --git a/Lib/test/dtracedata/instance.py b/Lib/test/dtracedata/instance.py
new file mode 100644
index 0000000..f142137
--- /dev/null
+++ b/Lib/test/dtracedata/instance.py
@@ -0,0 +1,24 @@
+import gc
+
+class old_style_class():
+    pass
+class new_style_class(object):
+    pass
+
+a = old_style_class()
+del a
+gc.collect()
+b = new_style_class()
+del b
+gc.collect()
+
+a = old_style_class()
+del old_style_class
+gc.collect()
+b = new_style_class()
+del new_style_class
+gc.collect()
+del a
+gc.collect()
+del b
+gc.collect()
diff --git a/Lib/test/dtracedata/line.d b/Lib/test/dtracedata/line.d
new file mode 100644
index 0000000..03f22db
--- /dev/null
+++ b/Lib/test/dtracedata/line.d
@@ -0,0 +1,7 @@
+python$target:::line
+/(copyinstr(arg1)=="test_line")/
+{
+    printf("%d\t%s:%s:%s:%d\n", timestamp,
+        probename, basename(copyinstr(arg0)),
+        copyinstr(arg1), arg2);
+}
diff --git a/Lib/test/dtracedata/line.d.expected b/Lib/test/dtracedata/line.d.expected
new file mode 100644
index 0000000..9b16ce7
--- /dev/null
+++ b/Lib/test/dtracedata/line.d.expected
@@ -0,0 +1,20 @@
+line:line.py:test_line:2
+line:line.py:test_line:3
+line:line.py:test_line:4
+line:line.py:test_line:5
+line:line.py:test_line:6
+line:line.py:test_line:7
+line:line.py:test_line:8
+line:line.py:test_line:9
+line:line.py:test_line:10
+line:line.py:test_line:11
+line:line.py:test_line:4
+line:line.py:test_line:5
+line:line.py:test_line:6
+line:line.py:test_line:7
+line:line.py:test_line:8
+line:line.py:test_line:10
+line:line.py:test_line:11
+line:line.py:test_line:4
+line:line.py:test_line:12
+line:line.py:test_line:13
diff --git a/Lib/test/dtracedata/line.py b/Lib/test/dtracedata/line.py
new file mode 100644
index 0000000..0930ff3
--- /dev/null
+++ b/Lib/test/dtracedata/line.py
@@ -0,0 +1,17 @@
+def test_line():
+    a = 1
+    print('# Preamble', a)
+    for i in range(2):
+        a = i
+        b = i+2
+        c = i+3
+        if c < 4:
+            a = c
+        d = a + b +c
+        print('#', a, b, c, d)
+    a = 1
+    print('# Epilogue', a)
+
+
+if __name__ == '__main__':
+    test_line()
diff --git a/Lib/test/test_dtrace.py b/Lib/test/test_dtrace.py
new file mode 100644
index 0000000..ca239b3
--- /dev/null
+++ b/Lib/test/test_dtrace.py
@@ -0,0 +1,178 @@
+import dis
+import os.path
+import re
+import subprocess
+import sys
+import types
+import unittest
+
+from test.support import findfile, run_unittest
+
+
+def abspath(filename):
+    return os.path.abspath(findfile(filename, subdir="dtracedata"))
+
+
+def normalize_trace_output(output):
+    """Normalize DTrace output for comparison.
+
+    DTrace keeps a per-CPU buffer, and when showing the fired probes, buffers
+    are concatenated. So if the operating system moves our thread around, the
+    straight result can be "non-causal". So we add timestamps to the probe
+    firing, sort by that field, then strip it from the output"""
+
+    # When compiling with '--with-pydebug', strip '[# refs]' debug output.
+    output = re.sub(r"\[[0-9]+ refs\]", "", output)
+    try:
+        result = [
+            row.split("\t")
+            for row in output.splitlines()
+            if row and not row.startswith('#')
+        ]
+        result.sort(key=lambda row: int(row[0]))
+        result = [row[1] for row in result]
+        return "\n".join(result)
+    except (IndexError, ValueError):
+        raise AssertionError(
+            "tracer produced unparseable output:\n{}".format(output)
+        )
+
+
+class TraceBackend:
+    EXTENSION = None
+    COMMAND = None
+    COMMAND_ARGS = []
+
+    def run_case(self, name, optimize_python=None):
+        actual_output = normalize_trace_output(self.trace_python(
+            script_file=abspath(name + self.EXTENSION),
+            python_file=abspath(name + ".py"),
+            optimize_python=optimize_python))
+
+        with open(abspath(name + self.EXTENSION + ".expected")) as f:
+            expected_output = f.read().rstrip()
+
+        return (expected_output, actual_output)
+
+    def generate_trace_command(self, script_file, subcommand=None):
+        command = self.COMMAND + [script_file]
+        if subcommand:
+            command += ["-c", subcommand]
+        return command
+
+    def trace(self, script_file, subcommand=None):
+        command = self.generate_trace_command(script_file, subcommand)
+        stdout, _ = subprocess.Popen(command,
+                                     stdout=subprocess.PIPE,
+                                     stderr=subprocess.STDOUT,
+                                     universal_newlines=True).communicate()
+        return stdout
+
+    def trace_python(self, script_file, python_file, optimize_python=None):
+        python_flags = []
+        if optimize_python:
+            python_flags.extend(["-O"] * optimize_python)
+        subcommand = " ".join([sys.executable] + python_flags + [python_file])
+        return self.trace(script_file, subcommand)
+
+    def assert_usable(self):
+        try:
+            output = self.trace(abspath("assert_usable" + self.EXTENSION))
+            output = output.strip()
+        except FileNotFoundError as fnfe:
+            output = str(fnfe)
+        if output != "probe: success":
+            raise unittest.SkipTest(
+                "{}(1) failed: {}".format(self.COMMAND[0], output)
+            )
+
+
+class DTraceBackend(TraceBackend):
+    EXTENSION = ".d"
+    COMMAND = ["dtrace", "-q", "-s"]
+
+
+class SystemTapBackend(TraceBackend):
+    EXTENSION = ".stp"
+    COMMAND = ["stap", "-g"]
+
+
+class TraceTests(unittest.TestCase):
+    # unittest.TestCase options
+    maxDiff = None
+
+    # TraceTests options
+    backend = None
+    optimize_python = 0
+
+    @classmethod
+    def setUpClass(self):
+        self.backend.assert_usable()
+
+    def run_case(self, name):
+        actual_output, expected_output = self.backend.run_case(
+            name, optimize_python=self.optimize_python)
+        self.assertEqual(actual_output, expected_output)
+
+    def test_function_entry_return(self):
+        self.run_case("call_stack")
+
+    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"])
+
+        with open(abspath("call_stack.py")) as f:
+            code_string = f.read()
+
+        def get_function_instructions(funcname):
+            # Recompile with appropriate optimization setting
+            code = compile(source=code_string,
+                           filename="<string>",
+                           mode="exec",
+                           optimize=self.optimize_python)
+
+            for c in code.co_consts:
+                if isinstance(c, types.CodeType) and c.co_name == funcname:
+                    return dis.get_instructions(c)
+            return []
+
+        for instruction in get_function_instructions('start'):
+            opcodes.discard(instruction.opname)
+
+        self.assertEqual(set(), opcodes)
+
+    def test_gc(self):
+        self.run_case("gc")
+
+    def test_line(self):
+        self.run_case("line")
+
+
+class DTraceNormalTests(TraceTests):
+    backend = DTraceBackend()
+    optimize_python = 0
+
+
+class DTraceOptimizedTests(TraceTests):
+    backend = DTraceBackend()
+    optimize_python = 2
+
+
+class SystemTapNormalTests(TraceTests):
+    backend = SystemTapBackend()
+    optimize_python = 0
+
+
+class SystemTapOptimizedTests(TraceTests):
+    backend = SystemTapBackend()
+    optimize_python = 2
+
+
+def test_main():
+    run_unittest(DTraceNormalTests, DTraceOptimizedTests, SystemTapNormalTests,
+                 SystemTapOptimizedTests)
+
+
+if __name__ == '__main__':
+    test_main()
diff --git a/Makefile.pre.in b/Makefile.pre.in
index 4445b24..908ca52 100644
--- a/Makefile.pre.in
+++ b/Makefile.pre.in
@@ -48,6 +48,10 @@ PGO_PROF_USE_FLAG=@PGO_PROF_USE_FLAG@
 LLVM_PROF_MERGER=@LLVM_PROF_MERGER@
 LLVM_PROF_FILE=@LLVM_PROF_FILE@
 LLVM_PROF_ERR=@LLVM_PROF_ERR@
+DTRACE=         @DTRACE@
+DFLAGS=         @DFLAGS@
+DTRACE_HEADERS= @DTRACE_HEADERS@
+DTRACE_OBJS=    @DTRACE_OBJS@
 
 GNULD=		@GNULD@
 
@@ -315,7 +319,7 @@ OPCODE_H_DIR= 	$(srcdir)/Include
 OPCODE_H_SCRIPT= $(srcdir)/Tools/scripts/generate_opcode_h.py
 OPCODE_H=	$(OPCODE_H_DIR)/opcode.h
 OPCODE_H_GEN=	$(PYTHON_FOR_GEN) $(OPCODE_H_SCRIPT) $(srcdir)/Lib/opcode.py $(OPCODE_H)
-#
+
 ##########################################################################
 # AST
 AST_H_DIR=	Include
@@ -391,7 +395,8 @@ PYTHON_OBJS=	\
 		Python/$(DYNLOADFILE) \
 		$(LIBOBJS) \
 		$(MACHDEP_OBJS) \
-		$(THREADOBJ)
+		$(THREADOBJ) \
+		$(DTRACE_OBJS)
 
 
 ##########################################################################
@@ -451,6 +456,15 @@ LIBRARY_OBJS=	\
 		$(LIBRARY_OBJS_OMIT_FROZEN) \
 		Python/frozen.o
 
+##########################################################################
+# DTrace
+
+# On some systems, object files that reference DTrace probes need to be modified
+# in-place by dtrace(1).
+DTRACE_DEPS = \
+	Python/ceval.o
+# XXX: should gcmodule, etc. be here, too?
+
 #########################################################################
 # Rules
 
@@ -852,6 +866,18 @@ Python/ceval.o: $(OPCODETARGETS_H) $(srcdir)/Python/ceval_gil.h
 
 Python/frozen.o: Python/importlib.h Python/importlib_external.h
 
+# Generate DTrace probe macros, then rename them (PYTHON_ -> PyDTrace_) to
+# follow our naming conventions. dtrace(1) uses the output filename to generate
+# an include guard, so we can't use a pipeline to transform its output.
+Include/pydtrace_probes.h: $(srcdir)/Include/pydtrace.d
+	$(DTRACE) $(DFLAGS) -o $@ -h -s $<
+	: sed in-place edit with POSIX-only tools
+	sed 's/PYTHON_/PyDTrace_/' $@ > $@.tmp
+	mv $@.tmp $@
+
+Python/pydtrace.o: $(srcdir)/Include/pydtrace.d $(DTRACE_DEPS)
+	$(DTRACE) $(DFLAGS) -o $@ -G -s $< $(DTRACE_DEPS)
+
 Objects/typeobject.o: Objects/typeslots.inc
 Objects/typeslots.inc: $(srcdir)/Include/typeslots.h $(srcdir)/Objects/typeslots.py
 	$(PYTHON_FOR_GEN) $(srcdir)/Objects/typeslots.py < $(srcdir)/Include/typeslots.h Objects/typeslots.inc
@@ -918,6 +944,7 @@ PYTHON_HEADERS= \
 		$(srcdir)/Include/pycapsule.h \
 		$(srcdir)/Include/pyctype.h \
 		$(srcdir)/Include/pydebug.h \
+		$(srcdir)/Include/pydtrace.h \
 		$(srcdir)/Include/pyerrors.h \
 		$(srcdir)/Include/pyfpe.h \
 		$(srcdir)/Include/pyhash.h \
@@ -949,7 +976,8 @@ PYTHON_HEADERS= \
 		$(srcdir)/Include/weakrefobject.h \
 		pyconfig.h \
 		$(PARSER_HEADERS) \
-		$(AST_H)
+		$(AST_H) \
+		$(DTRACE_HEADERS)
 
 $(LIBRARY_OBJS) $(MODOBJS) Programs/python.o: $(PYTHON_HEADERS)
 
@@ -1158,6 +1186,7 @@ LIBSUBDIRS=	tkinter tkinter/test tkinter/test/test_tkinter \
 		test/audiodata \
 		test/capath test/data \
 		test/cjkencodings test/decimaltestdata test/xmltestdata \
+		test/dtracedata \
 		test/eintrdata \
 		test/imghdrdata \
 		test/libregrtest \
@@ -1569,6 +1598,7 @@ clean: pycremoval
 	-rm -f Lib/lib2to3/*Grammar*.pickle
 	-rm -f Programs/_testembed Programs/_freeze_importlib
 	-find build -type f -a ! -name '*.gc??' -exec rm -f {} ';'
+	-rm -f Include/pydtrace_probes.h
 
 profile-removal:
 	find . -name '*.gc??' -exec rm -f {} ';'
diff --git a/Misc/ACKS b/Misc/ACKS
index 414cda9..0aeb1b1 100644
--- a/Misc/ACKS
+++ b/Misc/ACKS
@@ -120,6 +120,7 @@ Ben Bell
 Thomas Bellman
 Alexander “Саша” Belopolsky
 Eli Bendersky
+Nikhil Benesch
 David Benjamin
 Oscar Benjamin
 Andrew Bennetts
diff --git a/Misc/NEWS b/Misc/NEWS
index 28d9bb1b..a6979dc 100644
--- a/Misc/NEWS
+++ b/Misc/NEWS
@@ -329,6 +329,8 @@ Build
   make time when --with-optimizations is enabled.  Also improve our
   ability to find the llvm-profdata tool on MacOS and some Linuxes.
 
+- Issue #21590: Support for DTrace and SystemTap probes.
+
 - Issue #26307: The profile-opt build now applys PGO to the built-in modules.
 
 - Issue #26539: Add the --with-optimizations flag to turn on LTO and PGO build
diff --git a/Modules/gcmodule.c b/Modules/gcmodule.c
index 07950a6..2575d96 100644
--- a/Modules/gcmodule.c
+++ b/Modules/gcmodule.c
@@ -25,6 +25,7 @@
 
 #include "Python.h"
 #include "frameobject.h"        /* for PyFrame_ClearFreeList */
+#include "pydtrace.h"
 #include "pytime.h"             /* for _PyTime_GetMonotonicClock() */
 
 /* Get an object's GC head */
@@ -925,6 +926,9 @@ collect(int generation, Py_ssize_t *n_collected, Py_ssize_t *n_uncollectable,
         PySys_WriteStderr("\n");
     }
 
+    if (PyDTrace_GC_START_ENABLED())
+        PyDTrace_GC_START(generation);
+
     /* update collection and allocation counters */
     if (generation+1 < NUM_GENERATIONS)
         generations[generation+1].count += 1;
@@ -1069,6 +1073,10 @@ collect(int generation, Py_ssize_t *n_collected, Py_ssize_t *n_uncollectable,
     stats->collections++;
     stats->collected += m;
     stats->uncollectable += n;
+
+    if (PyDTrace_GC_DONE_ENABLED())
+        PyDTrace_GC_DONE(n+m);
+
     return n+m;
 }
 
diff --git a/Python/ceval.c b/Python/ceval.c
index d3bd8b5..a396e81 100644
--- a/Python/ceval.c
+++ b/Python/ceval.c
@@ -15,6 +15,7 @@
 #include "dictobject.h"
 #include "frameobject.h"
 #include "opcode.h"
+#include "pydtrace.h"
 #include "setobject.h"
 #include "structmember.h"
 
@@ -50,6 +51,9 @@ static void call_exc_trace(Py_tracefunc, PyObject *,
                            PyThreadState *, PyFrameObject *);
 static int maybe_call_line_trace(Py_tracefunc, PyObject *,
                                  PyThreadState *, PyFrameObject *, int *, int *, int *);
+static void maybe_dtrace_line(PyFrameObject *, int *, int *, int *);
+static void dtrace_function_entry(PyFrameObject *);
+static void dtrace_function_return(PyFrameObject *);
 
 static PyObject * cmp_outcome(int, PyObject *, PyObject *);
 static PyObject * import_name(PyFrameObject *, PyObject *, PyObject *, PyObject *);
@@ -822,7 +826,7 @@ _PyEval_EvalFrameDefault(PyFrameObject *f, int throwflag)
 #ifdef LLTRACE
 #define FAST_DISPATCH() \
     { \
-        if (!lltrace && !_Py_TracingPossible) { \
+        if (!lltrace && !_Py_TracingPossible && !PyDTrace_LINE_ENABLED()) { \
             f->f_lasti = INSTR_OFFSET(); \
             NEXTOPARG(); \
             goto *opcode_targets[opcode]; \
@@ -832,7 +836,7 @@ _PyEval_EvalFrameDefault(PyFrameObject *f, int throwflag)
 #else
 #define FAST_DISPATCH() \
     { \
-        if (!_Py_TracingPossible) { \
+        if (!_Py_TracingPossible && !PyDTrace_LINE_ENABLED()) { \
             f->f_lasti = INSTR_OFFSET(); \
             NEXTOPARG(); \
             goto *opcode_targets[opcode]; \
@@ -1042,6 +1046,9 @@ _PyEval_EvalFrameDefault(PyFrameObject *f, int throwflag)
         }
     }
 
+    if (PyDTrace_FUNCTION_ENTRY_ENABLED())
+        dtrace_function_entry(f);
+
     co = f->f_code;
     names = co->co_names;
     consts = co->co_consts;
@@ -1162,6 +1169,9 @@ _PyEval_EvalFrameDefault(PyFrameObject *f, int throwflag)
     fast_next_opcode:
         f->f_lasti = INSTR_OFFSET();
 
+        if (PyDTrace_LINE_ENABLED())
+            maybe_dtrace_line(f, &instr_lb, &instr_ub, &instr_prev);
+
         /* line-by-line tracing support */
 
         if (_Py_TracingPossible &&
@@ -3620,6 +3630,8 @@ fast_yield:
 
     /* pop frame */
 exit_eval_frame:
+    if (PyDTrace_FUNCTION_RETURN_ENABLED())
+        dtrace_function_return(f);
     Py_LeaveRecursiveCall();
     f->f_executing = 0;
     tstate->frame = f->f_back;
@@ -5415,3 +5427,65 @@ _PyEval_RequestCodeExtraIndex(freefunc free)
     tstate->co_extra_freefuncs[new_index] = free;
     return new_index;
 }
+
+static void
+dtrace_function_entry(PyFrameObject *f)
+{
+    char* filename;
+    char* funcname;
+    int lineno;
+
+    filename = PyUnicode_AsUTF8(f->f_code->co_filename);
+    funcname = PyUnicode_AsUTF8(f->f_code->co_name);
+    lineno = PyCode_Addr2Line(f->f_code, f->f_lasti);
+
+    PyDTrace_FUNCTION_ENTRY(filename, funcname, lineno);
+}
+
+static void
+dtrace_function_return(PyFrameObject *f)
+{
+    char* filename;
+    char* funcname;
+    int lineno;
+
+    filename = PyUnicode_AsUTF8(f->f_code->co_filename);
+    funcname = PyUnicode_AsUTF8(f->f_code->co_name);
+    lineno = PyCode_Addr2Line(f->f_code, f->f_lasti);
+
+    PyDTrace_FUNCTION_RETURN(filename, funcname, lineno);
+}
+
+/* DTrace equivalent of maybe_call_line_trace. */
+static void
+maybe_dtrace_line(PyFrameObject *frame,
+                  int *instr_lb, int *instr_ub, int *instr_prev)
+{
+    int line = frame->f_lineno;
+    char *co_filename, *co_name;
+
+    /* If the last instruction executed isn't in the current
+       instruction window, reset the window.
+    */
+    if (frame->f_lasti < *instr_lb || frame->f_lasti >= *instr_ub) {
+        PyAddrPair bounds;
+        line = _PyCode_CheckLineNumber(frame->f_code, frame->f_lasti,
+                                       &bounds);
+        *instr_lb = bounds.ap_lower;
+        *instr_ub = bounds.ap_upper;
+    }
+    /* If the last instruction falls at the start of a line or if
+       it represents a jump backwards, update the frame's line
+       number and call the trace function. */
+    if (frame->f_lasti == *instr_lb || frame->f_lasti < *instr_prev) {
+        frame->f_lineno = line;
+        co_filename = PyUnicode_AsUTF8(frame->f_code->co_filename);
+        if (!co_filename)
+            co_filename = "?";
+        co_name = PyUnicode_AsUTF8(frame->f_code->co_name);
+        if (!co_name)
+            co_name = "?";
+        PyDTrace_LINE(co_filename, co_name, line);
+    }
+    *instr_prev = frame->f_lasti;
+}
diff --git a/configure b/configure
index 3998a32..fae0791 100755
--- a/configure
+++ b/configure
@@ -642,6 +642,10 @@ TRUE
 MACHDEP_OBJS
 DYNLOADFILE
 DLINCLDIR
+DTRACE_OBJS
+DTRACE_HEADERS
+DFLAGS
+DTRACE
 THREADOBJ
 LDLAST
 USE_THREAD_MODULE
@@ -713,6 +717,7 @@ MULTIARCH
 ac_ct_CXX
 MAINCC
 CXX
+SED
 GREP
 CPP
 OBJEXT
@@ -780,7 +785,6 @@ infodir
 docdir
 oldincludedir
 includedir
-runstatedir
 localstatedir
 sharedstatedir
 sysconfdir
@@ -832,6 +836,7 @@ enable_ipv6
 with_doc_strings
 with_pymalloc
 with_valgrind
+with_dtrace
 with_fpectl
 with_libm
 with_libc
@@ -890,7 +895,6 @@ datadir='${datarootdir}'
 sysconfdir='${prefix}/etc'
 sharedstatedir='${prefix}/com'
 localstatedir='${prefix}/var'
-runstatedir='${localstatedir}/run'
 includedir='${prefix}/include'
 oldincludedir='/usr/include'
 docdir='${datarootdir}/doc/${PACKAGE_TARNAME}'
@@ -1143,15 +1147,6 @@ do
   | -silent | --silent | --silen | --sile | --sil)
     silent=yes ;;
 
-  -runstatedir | --runstatedir | --runstatedi | --runstated \
-  | --runstate | --runstat | --runsta | --runst | --runs \
-  | --run | --ru | --r)
-    ac_prev=runstatedir ;;
-  -runstatedir=* | --runstatedir=* | --runstatedi=* | --runstated=* \
-  | --runstate=* | --runstat=* | --runsta=* | --runst=* | --runs=* \
-  | --run=* | --ru=* | --r=*)
-    runstatedir=$ac_optarg ;;
-
   -sbindir | --sbindir | --sbindi | --sbind | --sbin | --sbi | --sb)
     ac_prev=sbindir ;;
   -sbindir=* | --sbindir=* | --sbindi=* | --sbind=* | --sbin=* \
@@ -1289,7 +1284,7 @@ fi
 for ac_var in	exec_prefix prefix bindir sbindir libexecdir datarootdir \
 		datadir sysconfdir sharedstatedir localstatedir includedir \
 		oldincludedir docdir infodir htmldir dvidir pdfdir psdir \
-		libdir localedir mandir runstatedir
+		libdir localedir mandir
 do
   eval ac_val=\$$ac_var
   # Remove trailing slashes.
@@ -1442,7 +1437,6 @@ Fine tuning of the installation directories:
   --sysconfdir=DIR        read-only single-machine data [PREFIX/etc]
   --sharedstatedir=DIR    modifiable architecture-independent data [PREFIX/com]
   --localstatedir=DIR     modifiable single-machine data [PREFIX/var]
-  --runstatedir=DIR       modifiable per-process data [LOCALSTATEDIR/run]
   --libdir=DIR            object code libraries [EPREFIX/lib]
   --includedir=DIR        C header files [PREFIX/include]
   --oldincludedir=DIR     C header files for non-gcc [/usr/include]
@@ -1535,6 +1529,7 @@ Optional Packages:
   --with(out)-doc-strings disable/enable documentation strings
   --with(out)-pymalloc    disable/enable specialized mallocs
   --with-valgrind         Enable Valgrind support
+  --with(out)-dtrace      disable/enable DTrace support
   --with-fpectl           enable SIGFPE catching
   --with-libm=STRING      math library
   --with-libc=STRING      C library
@@ -4581,6 +4576,75 @@ $as_echo "$ac_cv_path_GREP" >&6; }
  GREP="$ac_cv_path_GREP"
 
 
+{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for a sed that does not truncate output" >&5
+$as_echo_n "checking for a sed that does not truncate output... " >&6; }
+if ${ac_cv_path_SED+:} false; then :
+  $as_echo_n "(cached) " >&6
+else
+            ac_script=s/aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa/bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb/
+     for ac_i in 1 2 3 4 5 6 7; do
+       ac_script="$ac_script$as_nl$ac_script"
+     done
+     echo "$ac_script" 2>/dev/null | sed 99q >conftest.sed
+     { ac_script=; unset ac_script;}
+     if test -z "$SED"; then
+  ac_path_SED_found=false
+  # Loop through the user's path and test for each of PROGNAME-LIST
+  as_save_IFS=$IFS; IFS=$PATH_SEPARATOR
+for as_dir in $PATH
+do
+  IFS=$as_save_IFS
+  test -z "$as_dir" && as_dir=.
+    for ac_prog in sed gsed; do
+    for ac_exec_ext in '' $ac_executable_extensions; do
+      ac_path_SED="$as_dir/$ac_prog$ac_exec_ext"
+      as_fn_executable_p "$ac_path_SED" || continue
+# Check for GNU ac_path_SED and select it if it is found.
+  # Check for GNU $ac_path_SED
+case `"$ac_path_SED" --version 2>&1` in
+*GNU*)
+  ac_cv_path_SED="$ac_path_SED" ac_path_SED_found=:;;
+*)
+  ac_count=0
+  $as_echo_n 0123456789 >"conftest.in"
+  while :
+  do
+    cat "conftest.in" "conftest.in" >"conftest.tmp"
+    mv "conftest.tmp" "conftest.in"
+    cp "conftest.in" "conftest.nl"
+    $as_echo '' >> "conftest.nl"
+    "$ac_path_SED" -f conftest.sed < "conftest.nl" >"conftest.out" 2>/dev/null || break
+    diff "conftest.out" "conftest.nl" >/dev/null 2>&1 || break
+    as_fn_arith $ac_count + 1 && ac_count=$as_val
+    if test $ac_count -gt ${ac_path_SED_max-0}; then
+      # Best one so far, save it but keep looking for a better one
+      ac_cv_path_SED="$ac_path_SED"
+      ac_path_SED_max=$ac_count
+    fi
+    # 10*(2^10) chars as input seems more than enough
+    test $ac_count -gt 10 && break
+  done
+  rm -f conftest.in conftest.tmp conftest.nl conftest.out;;
+esac
+
+      $ac_path_SED_found && break 3
+    done
+  done
+  done
+IFS=$as_save_IFS
+  if test -z "$ac_cv_path_SED"; then
+    as_fn_error $? "no acceptable sed could be found in \$PATH" "$LINENO" 5
+  fi
+else
+  ac_cv_path_SED=$SED
+fi
+
+fi
+{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $ac_cv_path_SED" >&5
+$as_echo "$ac_cv_path_SED" >&6; }
+ SED="$ac_cv_path_SED"
+  rm -f conftest.sed
+
 
 
 
@@ -10864,6 +10928,102 @@ fi
     OPT="-DDYNAMIC_ANNOTATIONS_ENABLED=1 $OPT"
 fi
 
+# Check for DTrace support
+{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for --with-dtrace" >&5
+$as_echo_n "checking for --with-dtrace... " >&6; }
+
+# Check whether --with-dtrace was given.
+if test "${with_dtrace+set}" = set; then :
+  withval=$with_dtrace;
+else
+  with_dtrace=no
+fi
+
+{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $with_dtrace" >&5
+$as_echo "$with_dtrace" >&6; }
+
+
+
+
+
+DTRACE=
+DFLAGS=
+DTRACE_HEADERS=
+DTRACE_OBJS=
+
+if test "$with_dtrace" = "yes"
+then
+    # Extract the first word of "dtrace", so it can be a program name with args.
+set dummy dtrace; ac_word=$2
+{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for $ac_word" >&5
+$as_echo_n "checking for $ac_word... " >&6; }
+if ${ac_cv_path_DTRACE+:} false; then :
+  $as_echo_n "(cached) " >&6
+else
+  case $DTRACE in
+  [\\/]* | ?:[\\/]*)
+  ac_cv_path_DTRACE="$DTRACE" # Let the user override the test with a path.
+  ;;
+  *)
+  as_save_IFS=$IFS; IFS=$PATH_SEPARATOR
+for as_dir in $PATH
+do
+  IFS=$as_save_IFS
+  test -z "$as_dir" && as_dir=.
+    for ac_exec_ext in '' $ac_executable_extensions; do
+  if as_fn_executable_p "$as_dir/$ac_word$ac_exec_ext"; then
+    ac_cv_path_DTRACE="$as_dir/$ac_word$ac_exec_ext"
+    $as_echo "$as_me:${as_lineno-$LINENO}: found $as_dir/$ac_word$ac_exec_ext" >&5
+    break 2
+  fi
+done
+  done
+IFS=$as_save_IFS
+
+  test -z "$ac_cv_path_DTRACE" && ac_cv_path_DTRACE="not found"
+  ;;
+esac
+fi
+DTRACE=$ac_cv_path_DTRACE
+if test -n "$DTRACE"; then
+  { $as_echo "$as_me:${as_lineno-$LINENO}: result: $DTRACE" >&5
+$as_echo "$DTRACE" >&6; }
+else
+  { $as_echo "$as_me:${as_lineno-$LINENO}: result: no" >&5
+$as_echo "no" >&6; }
+fi
+
+
+    if test "$DTRACE" = "not found"; then
+        as_fn_error $? "dtrace command not found on \$PATH" "$LINENO" 5
+    fi
+
+$as_echo "#define WITH_DTRACE 1" >>confdefs.h
+
+    DTRACE_HEADERS="Include/pydtrace_probes.h"
+
+    # On OS X, DTrace providers do not need to be explicitly compiled and
+    # linked into the binary. Correspondingly, dtrace(1) is missing the ELF
+    # generation flag '-G'. We check for presence of this flag, rather than
+    # hardcoding support by OS, in the interest of robustness.
+    { $as_echo "$as_me:${as_lineno-$LINENO}: checking whether DTrace probes require linking" >&5
+$as_echo_n "checking whether DTrace probes require linking... " >&6; }
+if ${ac_cv_dtrace_link+:} false; then :
+  $as_echo_n "(cached) " >&6
+else
+              ac_cv_dtrace_link=no
+            echo 'BEGIN' > conftest.d
+            "$DTRACE" -G -s conftest.d -o conftest.o > /dev/null 2>&1 && \
+                ac_cv_dtrace_link=yes
+
+fi
+{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $ac_cv_dtrace_link" >&5
+$as_echo "$ac_cv_dtrace_link" >&6; }
+    if test "$ac_cv_dtrace_link" = "yes"; then
+        DTRACE_OBJS="Python/pydtrace.o"
+    fi
+fi
+
 # -I${DLINCLDIR} is added to the compile rule for importdl.o
 
 DLINCLDIR=.
diff --git a/configure.ac b/configure.ac
index 269c41e..b1af05e 100644
--- a/configure.ac
+++ b/configure.ac
@@ -694,6 +694,7 @@ fi
 AC_PROG_CC
 AC_PROG_CPP
 AC_PROG_GREP
+AC_PROG_SED
 
 AC_SUBST(CXX)
 AC_SUBST(MAINCC)
@@ -3245,6 +3246,47 @@ if test "$with_valgrind" != no; then
     OPT="-DDYNAMIC_ANNOTATIONS_ENABLED=1 $OPT"
 fi
 
+# Check for DTrace support
+AC_MSG_CHECKING(for --with-dtrace)
+AC_ARG_WITH(dtrace,
+  AC_HELP_STRING(--with(out)-dtrace, [disable/enable DTrace support]),,
+  with_dtrace=no)
+AC_MSG_RESULT($with_dtrace)
+
+AC_SUBST(DTRACE)
+AC_SUBST(DFLAGS)
+AC_SUBST(DTRACE_HEADERS)
+AC_SUBST(DTRACE_OBJS)
+DTRACE=
+DFLAGS=
+DTRACE_HEADERS=
+DTRACE_OBJS=
+
+if test "$with_dtrace" = "yes"
+then
+    AC_PATH_PROG(DTRACE, [dtrace], [not found])
+    if test "$DTRACE" = "not found"; then
+        AC_MSG_ERROR([dtrace command not found on \$PATH])
+    fi
+    AC_DEFINE(WITH_DTRACE, 1, [Define if you want to compile in DTrace support])
+    DTRACE_HEADERS="Include/pydtrace_probes.h"
+
+    # On OS X, DTrace providers do not need to be explicitly compiled and
+    # linked into the binary. Correspondingly, dtrace(1) is missing the ELF
+    # generation flag '-G'. We check for presence of this flag, rather than
+    # hardcoding support by OS, in the interest of robustness.
+    AC_CACHE_CHECK([whether DTrace probes require linking],
+        [ac_cv_dtrace_link], [dnl
+            ac_cv_dtrace_link=no
+            echo 'BEGIN' > conftest.d
+            "$DTRACE" -G -s conftest.d -o conftest.o > /dev/null 2>&1 && \
+                ac_cv_dtrace_link=yes
+      ])
+    if test "$ac_cv_dtrace_link" = "yes"; then
+        DTRACE_OBJS="Python/pydtrace.o"
+    fi
+fi
+
 # -I${DLINCLDIR} is added to the compile rule for importdl.o
 AC_SUBST(DLINCLDIR)
 DLINCLDIR=.
diff --git a/pyconfig.h.in b/pyconfig.h.in
index 7682c48..d3f61f7 100644
--- a/pyconfig.h.in
+++ b/pyconfig.h.in
@@ -1370,6 +1370,9 @@
 /* Define if you want documentation strings in extension modules */
 #undef WITH_DOC_STRINGS
 
+/* Define if you want to compile in DTrace support */
+#undef WITH_DTRACE
+
 /* Define if you want to use the new-style (Openstep, Rhapsody, MacOS) dynamic
    linker (dyld) instead of the old-style (NextStep) dynamic linker (rld).
    Dyld is necessary to support frameworks. */