[PATCH 09/10] btrace, python: Enable calling the ptwrite listener.

Willgerodt, Felix felix.willgerodt@intel.com
Mon Jun 14 14:53:13 GMT 2021


On 6/4/19 2:37 PM, Metzger, Markus T wrote:
> Hello Felix,
>
>> diff --git a/gdb/btrace.h b/gdb/btrace.h
>> index 7a80f5a44e1..7ad3cb65a28 100644
>> --- a/gdb/btrace.h
>> +++ b/gdb/btrace.h
>> @@ -354,6 +354,14 @@ struct btrace_thread_info
>>        stepping through the execution history.  */
>>     std::vector<std::string> aux_data;
>>
>> +  /* Function pointer to the ptwrite callback.  Returns the string returned
>> +     by the ptwrite listener function or nullptr if no string is supposed to
>> +     be printed.  */
>> +  gdb::unique_xmalloc_ptr<char> (*ptw_callback_fun) (
>> +const uint64_t *payload,
>> +const uint64_t *ip,
>> +const void *ptw_listener);
>> +
>>     /* PyObject pointer to the ptwrite listener function.  */
>>     void *ptw_listener = nullptr;
>>
>> diff --git a/gdb/python/py-record-btrace.c b/gdb/python/py-record-btrace.c
>> index c7ad47d64a0..42a0c29a348 100644
>> --- a/gdb/python/py-record-btrace.c
>> +++ b/gdb/python/py-record-btrace.c
>> @@ -772,6 +772,68 @@ recpy_bt_function_call_history (PyObject *self, void
>> *closure)
>>     return btpy_list_new (tinfo, first, last, 1, &recpy_func_type);
>>   }
>>
>> +/* Calling the ptwrite listener.  Returns a pointer to the string that will be
>> +   printed or nullptr if nothing should be printed.  */
>> +gdb::unique_xmalloc_ptr<char>
>> +recpy_call_listener (const uint64_t *payload, const uint64_t *ip,
>> +     const void *ptw_listener)
>> +{
>> +  if ((PyObject *) ptw_listener == Py_None)
>> +    return nullptr;
>> +  else if ((PyObject *) ptw_listener == nullptr)
>> +    error (_("No valid ptwrite listener."));

Markus> PTW_LISTENER is declared void * in struct btrace_info.  Couldn't we treat nullptr
Markus> as not-present?

If ptw_listener is a nullptr, something went rather wrong. We have a listener active by
default and can only "deactivate" the listener by registering Py_None. I don't think this
extra check hurts. But I don't have a strong opinion if you want it removed.


>> +
>> +  /* As Python is started as a seperate thread, we need to
>> +     acquire the GIL to safely call the listener function.  */
>> +  PyGILState_STATE gstate = PyGILState_Ensure ();
>> +
>> +  PyObject *py_ip = Py_None;
>> +  PyObject *py_payload = PyLong_FromUnsignedLongLong (*payload);
>> +  Py_INCREF (Py_None);

Markus> Are we dropping that reference again in the case that PY_IP gets overwritten?

We are now, thanks.

>> +
>> +  if (ip != nullptr)
>> +    py_ip = PyLong_FromUnsignedLongLong (*ip);

Markus> It wouldn't hurt to document that IP can be NULL whereas PAYLOAD cannot
Markus> in the function's comment.

Done

>> +
>> +  PyObject *py_result = PyObject_CallFunctionObjArgs ((PyObject *)
>> ptw_listener,
>> +      py_payload, py_ip, NULL);
>> +
>> +  if (PyErr_Occurred ())
>> +    {
>> +      gdbpy_print_stack ();
>> +      gdb_Py_DECREF (py_ip);
>> +      gdb_Py_DECREF (py_payload);
>> +      gdb_Py_DECREF (py_result);
>> +      PyGILState_Release (gstate);
>> +      error (_("Error while executing Python code."));
>> +    }
>> +
>> +  gdb_Py_DECREF (py_ip);
>> +  gdb_Py_DECREF (py_payload);
>> +
>> +  if (py_result == Py_None)
>> +    {
>> +      gdb_Py_DECREF (py_result);
>> +      PyGILState_Release (gstate);
>> +      return nullptr;
>> +    }
>> +
>> +  gdb::unique_xmalloc_ptr<char> resultstring = gdbpy_obj_to_string
>> (py_result);
>> +
>> +  if (PyErr_Occurred ())
>> +    {
>> +      gdbpy_print_stack ();
>> +      gdb_Py_DECREF (py_result);
>> +      PyGILState_Release (gstate);
>> +      error (_("Error while executing Python code."));
>> +    }
>> +
>> +  if (py_result != nullptr)
>> +    gdb_Py_DECREF (py_result);

Markus> Shouldn't we check that before using PY_RESULT above?

I removed the check as it shouldn't be necessary at all.
PyObject_CallFunctionObjArgs only returns nullptr if an error occured,
but that is already checked via PyErr_Occurred ().

Thanks,
Felix
Intel Deutschland GmbH
Registered Address: Am Campeon 10, 85579 Neubiberg, Germany
Tel: +49 89 99 8853-0, www.intel.de <http://www.intel.de>
Managing Directors: Christin Eisenschmid, Sharon Heck, Tiffany Doon Silva  
Chairperson of the Supervisory Board: Nicole Lau
Registered Office: Munich
Commercial Register: Amtsgericht Muenchen HRB 186928


More information about the Gdb-patches mailing list