Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
// Copyright 2018-2019 - Omar Sandoval
|
|
|
|
// SPDX-License-Identifier: GPL-3.0+
|
|
|
|
|
|
|
|
#include <limits.h>
|
|
|
|
|
|
|
|
#include "drgnpy.h"
|
2019-04-03 19:33:37 +01:00
|
|
|
#include "../error.h"
|
|
|
|
#include "../object.h"
|
|
|
|
#include "../serialize.h"
|
|
|
|
#include "../type.h"
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
|
|
|
|
static DrgnObject *DrgnObject_new(PyTypeObject *subtype, PyObject *args,
|
|
|
|
PyObject *kwds)
|
|
|
|
{
|
|
|
|
PyObject *arg;
|
|
|
|
|
|
|
|
if (PyTuple_GET_SIZE(args) < 1) {
|
|
|
|
PyErr_SetString(PyExc_TypeError,
|
|
|
|
"Object() missing required argument 'prog' (pos 1)");
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
arg = PyTuple_GET_ITEM(args, 0);
|
|
|
|
if (!PyObject_TypeCheck(arg, &Program_type)) {
|
|
|
|
PyErr_SetString(PyExc_TypeError,
|
|
|
|
"Object() argument 1 must be Program");
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
return DrgnObject_alloc((Program *)arg);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void DrgnObject_dealloc(DrgnObject *self)
|
|
|
|
{
|
|
|
|
Py_DECREF(DrgnObject_prog(self));
|
|
|
|
drgn_object_deinit(&self->obj);
|
|
|
|
Py_TYPE(self)->tp_free((PyObject *)self);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int DrgnObject_literal(struct drgn_object *res, PyObject *literal)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
|
|
|
|
if (PyBool_Check(literal)) {
|
|
|
|
err = drgn_object_bool_literal(res, literal == Py_True);
|
|
|
|
} else if (PyLong_Check(literal)) {
|
|
|
|
unsigned long long uvalue;
|
|
|
|
bool is_negative;
|
|
|
|
|
|
|
|
is_negative = Py_SIZE(literal) < 0;
|
|
|
|
if (is_negative) {
|
|
|
|
literal = PyNumber_Negative(literal);
|
|
|
|
if (!literal)
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
uvalue = PyLong_AsUnsignedLongLong(literal);
|
|
|
|
if (is_negative)
|
|
|
|
Py_DECREF(literal);
|
|
|
|
if (uvalue == (unsigned long long)-1 && PyErr_Occurred())
|
|
|
|
return -1;
|
|
|
|
err = drgn_object_integer_literal(res, uvalue);
|
|
|
|
if (!err && is_negative)
|
|
|
|
err = drgn_object_neg(res, res);
|
|
|
|
} else if (PyFloat_Check(literal)) {
|
|
|
|
err = drgn_object_float_literal(res,
|
|
|
|
PyFloat_AS_DOUBLE(literal));
|
|
|
|
} else {
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
if (err) {
|
|
|
|
set_drgn_error(err);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int serialize_py_object(struct drgn_program *prog, char *buf,
|
|
|
|
uint64_t buf_bit_size, uint64_t bit_offset,
|
|
|
|
PyObject *value_obj,
|
|
|
|
struct drgn_object_type *type,
|
|
|
|
bool little_endian);
|
|
|
|
|
|
|
|
static int serialize_compound_value(struct drgn_program *prog, char *buf,
|
|
|
|
uint64_t buf_bit_size, uint64_t bit_offset,
|
|
|
|
PyObject *value_obj,
|
|
|
|
struct drgn_object_type *type,
|
|
|
|
bool little_endian)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
PyObject *tmp, *items;
|
|
|
|
Py_ssize_t num_items, i;
|
|
|
|
int ret = -1;
|
|
|
|
|
|
|
|
if (!PyMapping_Check(value_obj)) {
|
|
|
|
set_error_type_name("'%s' value must be dictionary or mapping",
|
|
|
|
drgn_object_type_qualified(type));
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
tmp = PyMapping_Items(value_obj);
|
|
|
|
if (!tmp)
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Since Python 3.7, PyMapping_Items() always returns a list. However,
|
|
|
|
* before that, it could also return a tuple.
|
|
|
|
*/
|
|
|
|
items = PySequence_Fast(tmp, "items must be sequence");
|
|
|
|
Py_DECREF(tmp);
|
|
|
|
if (!items)
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
num_items = PySequence_Fast_GET_SIZE(items);
|
|
|
|
for (i = 0; i < num_items; i++) {
|
|
|
|
PyObject *item, *key;
|
|
|
|
const char *member_name;
|
2019-04-22 22:21:25 +01:00
|
|
|
struct drgn_member_info member;
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
struct drgn_object_type member_type;
|
|
|
|
|
|
|
|
item = PySequence_Fast_GET_ITEM(items, i);
|
|
|
|
if (!PyTuple_Check(item) || PyTuple_GET_SIZE(item) != 2) {
|
|
|
|
PyErr_SetString(PyExc_TypeError, "invalid item");
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
key = PyTuple_GET_ITEM(item, 0);
|
|
|
|
if (!PyUnicode_Check(key)) {
|
|
|
|
PyErr_SetString(PyExc_TypeError,
|
|
|
|
"member key must be string");
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
member_name = PyUnicode_AsUTF8(key);
|
|
|
|
if (!member_name)
|
|
|
|
goto out;
|
|
|
|
|
2019-04-22 22:21:25 +01:00
|
|
|
err = drgn_program_member_info(prog, type->underlying_type,
|
|
|
|
member_name, &member);
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
if (err) {
|
|
|
|
set_drgn_error(err);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2019-04-22 22:21:25 +01:00
|
|
|
member_type.type = member.qualified_type.type;
|
|
|
|
member_type.qualifiers = member.qualified_type.qualifiers;
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
member_type.underlying_type =
|
|
|
|
drgn_underlying_type(member_type.type);
|
2019-04-22 22:21:25 +01:00
|
|
|
member_type.bit_field_size = member.bit_field_size;
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
|
|
|
|
if (serialize_py_object(prog, buf, buf_bit_size,
|
2019-04-22 22:21:25 +01:00
|
|
|
bit_offset + member.bit_offset,
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
PyTuple_GET_ITEM(item, 1),
|
|
|
|
&member_type, little_endian) == -1)
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = 0;
|
|
|
|
out:
|
|
|
|
Py_DECREF(items);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int serialize_array_value(struct drgn_program *prog, char *buf,
|
|
|
|
uint64_t buf_bit_size, uint64_t bit_offset,
|
|
|
|
PyObject *value_obj,
|
|
|
|
struct drgn_object_type *type,
|
|
|
|
bool little_endian)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
struct drgn_qualified_type qualified_type;
|
|
|
|
struct drgn_object_type element_type;
|
|
|
|
uint64_t element_bit_size, length;
|
|
|
|
PyObject *seq;
|
|
|
|
size_t seq_length, i;
|
|
|
|
|
|
|
|
qualified_type = drgn_type_type(type->underlying_type);
|
|
|
|
element_type.type = qualified_type.type;
|
|
|
|
element_type.qualifiers = qualified_type.qualifiers;
|
|
|
|
element_type.underlying_type = drgn_underlying_type(element_type.type);
|
|
|
|
element_type.bit_field_size = 0;
|
|
|
|
err = drgn_type_bit_size(element_type.underlying_type,
|
|
|
|
&element_bit_size);
|
|
|
|
if (err) {
|
|
|
|
set_drgn_error(err);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
length = drgn_type_length(type->underlying_type);
|
|
|
|
if (length > PY_SSIZE_T_MAX) {
|
|
|
|
PyErr_NoMemory();
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
seq = PySequence_Fast(value_obj, "");
|
|
|
|
if (!seq) {
|
|
|
|
if (PyErr_ExceptionMatches(PyExc_TypeError)) {
|
|
|
|
set_error_type_name("'%s' value must be iterable",
|
|
|
|
drgn_object_type_qualified(type));
|
|
|
|
}
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
seq_length = PySequence_Fast_GET_SIZE(seq);
|
|
|
|
if (seq_length > length) {
|
|
|
|
Py_DECREF(seq);
|
|
|
|
PyErr_SetString(PyExc_ValueError,
|
|
|
|
"too many items in array value");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (i = 0; i < seq_length; i++) {
|
|
|
|
if (serialize_py_object(prog, buf, buf_bit_size,
|
|
|
|
bit_offset + i * element_bit_size,
|
|
|
|
PySequence_Fast_GET_ITEM(seq, i),
|
|
|
|
&element_type, little_endian) == -1) {
|
|
|
|
Py_DECREF(seq);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
Py_DECREF(seq);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int serialize_py_object(struct drgn_program *prog, char *buf,
|
|
|
|
uint64_t buf_bit_size, uint64_t bit_offset,
|
|
|
|
PyObject *value_obj,
|
|
|
|
struct drgn_object_type *type,
|
|
|
|
bool little_endian)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
enum drgn_object_kind kind;
|
|
|
|
uint64_t bit_size, bit_end;
|
|
|
|
|
|
|
|
err = drgn_object_type_kind_and_size(type, &kind, &bit_size);
|
|
|
|
if (err) {
|
|
|
|
set_drgn_error(err);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
err = sanity_check_object(kind, type->bit_field_size, bit_size);
|
|
|
|
if (err) {
|
|
|
|
set_drgn_error(err);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (__builtin_add_overflow(bit_offset, bit_size, &bit_end) ||
|
|
|
|
bit_end > buf_bit_size) {
|
|
|
|
err = drgn_error_create(DRGN_ERROR_FAULT,
|
|
|
|
"out of bounds of value");
|
|
|
|
set_drgn_error(err);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
switch (kind) {
|
|
|
|
case DRGN_OBJECT_SIGNED:
|
|
|
|
case DRGN_OBJECT_UNSIGNED: {
|
2019-11-23 01:18:55 +00:00
|
|
|
PyObject *long_obj;
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
union {
|
|
|
|
int64_t svalue;
|
|
|
|
uint64_t uvalue;
|
|
|
|
} tmp;
|
|
|
|
|
|
|
|
if (!PyNumber_Check(value_obj)) {
|
|
|
|
set_error_type_name("'%s' value must be number",
|
|
|
|
drgn_object_type_qualified(type));
|
|
|
|
return -1;
|
|
|
|
}
|
2019-11-23 01:18:55 +00:00
|
|
|
long_obj = PyNumber_Long(value_obj);
|
|
|
|
if (!long_obj)
|
|
|
|
return -1;
|
|
|
|
tmp.uvalue = PyLong_AsUnsignedLongLongMask(long_obj);
|
|
|
|
Py_DECREF(long_obj);
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
if (tmp.uvalue == (unsigned long long)-1 && PyErr_Occurred())
|
|
|
|
return -1;
|
|
|
|
if (kind == DRGN_OBJECT_SIGNED)
|
|
|
|
tmp.svalue = truncate_signed(tmp.svalue, bit_size);
|
|
|
|
else
|
|
|
|
tmp.uvalue = truncate_unsigned(tmp.uvalue, bit_size);
|
|
|
|
serialize_bits(buf, bit_offset, tmp.uvalue, bit_size,
|
|
|
|
little_endian);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
case DRGN_OBJECT_FLOAT: {
|
|
|
|
union {
|
|
|
|
uint64_t uvalue;
|
|
|
|
double fvalue64;
|
|
|
|
float fvalue32;
|
|
|
|
} tmp;
|
|
|
|
double fvalue;
|
|
|
|
|
|
|
|
if (!PyNumber_Check(value_obj)) {
|
|
|
|
set_error_type_name("'%s' value must be number",
|
|
|
|
drgn_object_type_qualified(type));
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
fvalue = PyFloat_AsDouble(value_obj);
|
|
|
|
if (fvalue == -1.0 && PyErr_Occurred())
|
|
|
|
return -1;
|
|
|
|
if (bit_size == 64)
|
|
|
|
tmp.fvalue64 = fvalue;
|
|
|
|
else
|
|
|
|
tmp.fvalue32 = fvalue;
|
|
|
|
serialize_bits(buf, bit_offset, tmp.uvalue, bit_size,
|
|
|
|
little_endian);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
case DRGN_OBJECT_BUFFER:
|
|
|
|
switch (drgn_type_kind(type->underlying_type)) {
|
|
|
|
case DRGN_TYPE_COMPLEX:
|
|
|
|
PyErr_SetString(PyExc_NotImplementedError,
|
|
|
|
"creating complex object is not implemented");
|
|
|
|
return -1;
|
|
|
|
case DRGN_TYPE_STRUCT:
|
|
|
|
case DRGN_TYPE_UNION:
|
2019-11-15 01:12:47 +00:00
|
|
|
case DRGN_TYPE_CLASS:
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
return serialize_compound_value(prog, buf, buf_bit_size,
|
|
|
|
bit_offset, value_obj,
|
|
|
|
type, little_endian);
|
|
|
|
case DRGN_TYPE_ARRAY:
|
|
|
|
return serialize_array_value(prog, buf, buf_bit_size,
|
|
|
|
bit_offset, value_obj,
|
|
|
|
type, little_endian);
|
|
|
|
default:
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
DRGN_UNREACHABLE();
|
|
|
|
}
|
|
|
|
|
|
|
|
static int buffer_object_from_value(struct drgn_object *res,
|
|
|
|
struct drgn_qualified_type qualified_type,
|
|
|
|
PyObject *value_obj, uint64_t bit_offset,
|
|
|
|
enum drgn_byte_order byte_order)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
union drgn_value value;
|
|
|
|
struct drgn_object_type type;
|
|
|
|
enum drgn_object_kind kind;
|
|
|
|
uint64_t bit_size, size;
|
|
|
|
char *buf;
|
|
|
|
|
|
|
|
err = drgn_byte_order_to_little_endian(res->prog, byte_order,
|
|
|
|
&value.little_endian);
|
|
|
|
if (err) {
|
|
|
|
set_drgn_error(err);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
err = drgn_object_set_common(qualified_type, 0, &type, &kind,
|
|
|
|
&bit_size);
|
|
|
|
if (err) {
|
|
|
|
set_drgn_error(err);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (bit_offset >= 8) {
|
|
|
|
PyErr_SetString(PyExc_ValueError,
|
|
|
|
"bit offset must be less than 8");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
if (bit_size > UINT64_MAX - bit_offset) {
|
|
|
|
PyErr_SetString(PyExc_OverflowError, "object is too large");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
size = drgn_value_size(bit_size, bit_offset);
|
|
|
|
if (size > SIZE_MAX) {
|
|
|
|
PyErr_NoMemory();
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
if (size <= sizeof(value.ibuf)) {
|
|
|
|
buf = value.ibuf;
|
|
|
|
} else {
|
|
|
|
buf = malloc(size);
|
|
|
|
if (!buf) {
|
|
|
|
PyErr_NoMemory();
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
value.bufp = buf;
|
|
|
|
}
|
|
|
|
memset(buf, 0, size);
|
|
|
|
value.bit_offset = bit_offset;
|
|
|
|
|
|
|
|
if (serialize_py_object(res->prog, buf, bit_offset + bit_size,
|
|
|
|
bit_offset, value_obj, &type,
|
|
|
|
value.little_endian) == -1) {
|
|
|
|
if (buf != value.ibuf)
|
|
|
|
free(buf);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
drgn_object_reinit(res, &type, kind, bit_size, false);
|
|
|
|
res->value = value;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int DrgnObject_init(DrgnObject *self, PyObject *args, PyObject *kwds)
|
|
|
|
{
|
|
|
|
static char *keywords[] = {
|
|
|
|
"prog", "type", "value", "address", "byteorder",
|
2019-04-04 09:27:23 +01:00
|
|
|
"bit_offset", "bit_field_size", NULL,
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
};
|
|
|
|
struct drgn_error *err;
|
|
|
|
Program *prog;
|
|
|
|
PyObject *type_obj = Py_None, *value_obj = Py_None;
|
2019-10-16 02:03:21 +01:00
|
|
|
struct index_arg address = { .allow_none = true, .is_none = true };
|
2019-10-16 05:21:21 +01:00
|
|
|
struct byteorder_arg byteorder = {
|
|
|
|
.allow_none = true,
|
|
|
|
.is_none = true,
|
|
|
|
.value = DRGN_PROGRAM_ENDIAN,
|
|
|
|
};
|
2019-10-16 02:03:21 +01:00
|
|
|
struct index_arg bit_offset = { .allow_none = true, .is_none = true };
|
|
|
|
struct index_arg bit_field_size = { .allow_none = true, .is_none = true };
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
struct drgn_qualified_type qualified_type;
|
|
|
|
|
2019-10-16 05:21:21 +01:00
|
|
|
if (!PyArg_ParseTupleAndKeywords(args, kwds, "O!|O$OO&O&O&O&:Object",
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
keywords, &Program_type, &prog,
|
2019-10-16 02:03:21 +01:00
|
|
|
&type_obj, &value_obj, index_converter,
|
2019-10-16 05:21:21 +01:00
|
|
|
&address, byteorder_converter,
|
|
|
|
&byteorder, index_converter,
|
|
|
|
&bit_offset, index_converter,
|
|
|
|
&bit_field_size))
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
return -1;
|
|
|
|
|
|
|
|
if (&prog->prog != self->obj.prog) {
|
|
|
|
PyErr_SetString(PyExc_ValueError,
|
|
|
|
"cannot change object program");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
2019-04-12 07:26:39 +01:00
|
|
|
if (Program_type_arg(DrgnObject_prog(self), type_obj, true,
|
|
|
|
&qualified_type) == -1)
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
return -1;
|
|
|
|
|
2019-11-30 04:40:40 +00:00
|
|
|
if (!bit_field_size.is_none && bit_field_size.uvalue == 0) {
|
2019-10-16 02:03:21 +01:00
|
|
|
PyErr_SetString(PyExc_ValueError,
|
|
|
|
"bit field size cannot be zero");
|
|
|
|
return -1;
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
}
|
|
|
|
|
2019-10-16 02:03:21 +01:00
|
|
|
if (!address.is_none && value_obj != Py_None) {
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
PyErr_SetString(PyExc_ValueError,
|
|
|
|
"object cannot have address and value");
|
|
|
|
return -1;
|
2019-10-16 02:03:21 +01:00
|
|
|
} else if (!address.is_none) {
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
if (!qualified_type.type) {
|
|
|
|
PyErr_SetString(PyExc_ValueError,
|
|
|
|
"reference must have type");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
err = drgn_object_set_reference(&self->obj, qualified_type,
|
2019-11-30 04:40:40 +00:00
|
|
|
address.uvalue,
|
|
|
|
bit_offset.uvalue,
|
|
|
|
bit_field_size.uvalue,
|
2019-10-16 05:21:21 +01:00
|
|
|
byteorder.value);
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
} else if (value_obj != Py_None && !qualified_type.type) {
|
|
|
|
int ret;
|
|
|
|
|
2019-10-16 05:21:21 +01:00
|
|
|
if (!byteorder.is_none) {
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
PyErr_SetString(PyExc_ValueError,
|
|
|
|
"literal cannot have byteorder");
|
|
|
|
return -1;
|
|
|
|
}
|
2019-10-16 02:03:21 +01:00
|
|
|
if (!bit_offset.is_none) {
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
PyErr_SetString(PyExc_ValueError,
|
|
|
|
"literal cannot have bit offset");
|
|
|
|
return -1;
|
|
|
|
}
|
2019-10-16 02:03:21 +01:00
|
|
|
if (!bit_field_size.is_none) {
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
PyErr_SetString(PyExc_ValueError,
|
|
|
|
"literal cannot be bit field");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = DrgnObject_literal(&self->obj, value_obj);
|
|
|
|
if (ret == -1) {
|
|
|
|
return -1;
|
|
|
|
} else if (ret) {
|
|
|
|
PyErr_Format(PyExc_TypeError,
|
|
|
|
"cannot create %s literal",
|
|
|
|
Py_TYPE(value_obj)->tp_name);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
err = NULL;
|
|
|
|
} else if (value_obj != Py_None) {
|
|
|
|
enum drgn_object_kind kind;
|
|
|
|
|
|
|
|
kind = drgn_type_object_kind(qualified_type.type);
|
|
|
|
if (!drgn_object_kind_is_complete(kind)) {
|
|
|
|
err = drgn_error_incomplete_type("cannot create object with %s type",
|
|
|
|
qualified_type.type);
|
|
|
|
set_drgn_error(err);
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
}
|
2019-10-16 02:03:21 +01:00
|
|
|
if (!bit_field_size.is_none && kind != DRGN_OBJECT_SIGNED &&
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
kind != DRGN_OBJECT_UNSIGNED) {
|
|
|
|
PyErr_SetString(PyExc_ValueError,
|
|
|
|
"bit field must be integer");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
if (kind != DRGN_OBJECT_BUFFER) {
|
2019-10-16 05:21:21 +01:00
|
|
|
if (!byteorder.is_none) {
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
PyErr_SetString(PyExc_ValueError,
|
|
|
|
"primitive value cannot have byteorder");
|
|
|
|
return -1;
|
|
|
|
}
|
2019-10-16 02:03:21 +01:00
|
|
|
if (!bit_offset.is_none) {
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
PyErr_SetString(PyExc_ValueError,
|
|
|
|
"primitive value cannot have bit offset");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
switch (kind) {
|
|
|
|
case DRGN_OBJECT_BUFFER:
|
|
|
|
if (buffer_object_from_value(&self->obj, qualified_type,
|
2019-10-16 02:03:21 +01:00
|
|
|
value_obj,
|
2019-11-30 04:40:40 +00:00
|
|
|
bit_offset.uvalue,
|
2019-10-16 05:21:21 +01:00
|
|
|
byteorder.value) == -1)
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
return -1;
|
|
|
|
err = NULL;
|
|
|
|
break;
|
|
|
|
case DRGN_OBJECT_SIGNED:
|
|
|
|
case DRGN_OBJECT_UNSIGNED: {
|
2019-11-23 01:18:55 +00:00
|
|
|
PyObject *long_obj;
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
union {
|
|
|
|
int64_t svalue;
|
|
|
|
uint64_t uvalue;
|
|
|
|
} tmp;
|
|
|
|
|
|
|
|
if (!PyNumber_Check(value_obj)) {
|
|
|
|
set_error_type_name("'%s' value must be number",
|
|
|
|
qualified_type);
|
|
|
|
return -1;
|
|
|
|
}
|
2019-11-23 01:18:55 +00:00
|
|
|
long_obj = PyNumber_Long(value_obj);
|
|
|
|
if (!long_obj)
|
|
|
|
return -1;
|
|
|
|
tmp.uvalue = PyLong_AsUnsignedLongLongMask(long_obj);
|
|
|
|
Py_DECREF(long_obj);
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
if (tmp.uvalue == (unsigned long long)-1 &&
|
|
|
|
PyErr_Occurred())
|
|
|
|
return -1;
|
|
|
|
if (kind == DRGN_OBJECT_SIGNED) {
|
|
|
|
err = drgn_object_set_signed(&self->obj,
|
|
|
|
qualified_type,
|
|
|
|
tmp.svalue,
|
2019-11-30 04:40:40 +00:00
|
|
|
bit_field_size.uvalue);
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
} else {
|
|
|
|
err = drgn_object_set_unsigned(&self->obj,
|
|
|
|
qualified_type,
|
|
|
|
tmp.uvalue,
|
2019-11-30 04:40:40 +00:00
|
|
|
bit_field_size.uvalue);
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
}
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
case DRGN_OBJECT_FLOAT: {
|
|
|
|
double fvalue;
|
|
|
|
|
|
|
|
if (!PyNumber_Check(value_obj)) {
|
|
|
|
set_error_type_name("'%s' value must be number",
|
|
|
|
qualified_type);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
fvalue = PyFloat_AsDouble(value_obj);
|
|
|
|
if (fvalue == -1.0 && PyErr_Occurred())
|
|
|
|
return -1;
|
|
|
|
err = drgn_object_set_float(&self->obj, qualified_type,
|
|
|
|
fvalue);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
default:
|
|
|
|
DRGN_UNREACHABLE();
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
PyErr_SetString(PyExc_ValueError,
|
|
|
|
"object must have either address or value");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
if (err) {
|
|
|
|
set_drgn_error(err);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static PyObject *DrgnObject_value_impl(struct drgn_object *obj);
|
|
|
|
|
|
|
|
static PyObject *DrgnObject_compound_value(struct drgn_object *obj,
|
|
|
|
struct drgn_type *underlying_type)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
PyObject *dict;
|
|
|
|
struct drgn_object member;
|
|
|
|
struct drgn_type_member *members;
|
|
|
|
size_t num_members, i;
|
|
|
|
|
|
|
|
if (!drgn_type_is_complete(underlying_type)) {
|
|
|
|
PyErr_Format(PyExc_TypeError,
|
|
|
|
"cannot get value of incomplete %s",
|
2019-11-15 01:12:47 +00:00
|
|
|
drgn_type_kind_spelling[drgn_type_kind(underlying_type)]);
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
dict = PyDict_New();
|
|
|
|
if (!dict)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
drgn_object_init(&member, obj->prog);
|
|
|
|
members = drgn_type_members(underlying_type);
|
|
|
|
num_members = drgn_type_num_members(underlying_type);
|
|
|
|
for (i = 0; i < num_members; i++) {
|
|
|
|
struct drgn_qualified_type member_type;
|
|
|
|
PyObject *member_value;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
err = drgn_member_type(&members[i], &member_type);
|
|
|
|
if (err) {
|
|
|
|
set_drgn_error(err);
|
|
|
|
Py_CLEAR(dict);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
err = drgn_object_slice(&member, obj, member_type,
|
|
|
|
members[i].bit_offset,
|
|
|
|
members[i].bit_field_size);
|
|
|
|
if (err) {
|
|
|
|
set_drgn_error(err);
|
|
|
|
Py_CLEAR(dict);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
member_value = DrgnObject_value_impl(&member);
|
|
|
|
if (!member_value) {
|
|
|
|
Py_CLEAR(dict);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (members[i].name) {
|
|
|
|
ret = PyDict_SetItemString(dict, members[i].name,
|
|
|
|
member_value);
|
|
|
|
} else {
|
|
|
|
ret = PyDict_Update(dict, member_value);
|
|
|
|
}
|
|
|
|
Py_DECREF(member_value);
|
|
|
|
if (ret) {
|
|
|
|
Py_CLEAR(dict);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
out:
|
|
|
|
drgn_object_deinit(&member);
|
|
|
|
return dict;
|
|
|
|
}
|
|
|
|
|
|
|
|
static PyObject *DrgnObject_array_value(struct drgn_object *obj,
|
|
|
|
struct drgn_type *underlying_type)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
struct drgn_qualified_type element_type;
|
|
|
|
uint64_t element_bit_size, length, i;
|
|
|
|
PyObject *list;
|
|
|
|
struct drgn_object element;
|
|
|
|
|
|
|
|
element_type = drgn_type_type(underlying_type);
|
|
|
|
err = drgn_type_bit_size(element_type.type, &element_bit_size);
|
|
|
|
if (err)
|
|
|
|
return set_drgn_error(err);
|
|
|
|
|
|
|
|
length = drgn_type_length(underlying_type);
|
|
|
|
if (length > PY_SSIZE_T_MAX) {
|
|
|
|
PyErr_NoMemory();
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
list = PyList_New(length);
|
|
|
|
if (!list)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
drgn_object_init(&element, obj->prog);
|
|
|
|
for (i = 0; i < length; i++) {
|
|
|
|
PyObject *element_value;
|
|
|
|
|
|
|
|
err = drgn_object_slice(&element, obj, element_type,
|
|
|
|
i * element_bit_size, 0);
|
|
|
|
if (err) {
|
|
|
|
set_drgn_error(err);
|
|
|
|
Py_CLEAR(list);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
element_value = DrgnObject_value_impl(&element);
|
|
|
|
if (!element_value) {
|
|
|
|
Py_CLEAR(list);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
PyList_SET_ITEM(list, i, element_value);
|
|
|
|
}
|
|
|
|
|
|
|
|
out:
|
|
|
|
drgn_object_deinit(&element);
|
|
|
|
return list;
|
|
|
|
}
|
|
|
|
|
|
|
|
static PyObject *DrgnObject_value_impl(struct drgn_object *obj)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
struct drgn_type *underlying_type;
|
|
|
|
|
|
|
|
if (!drgn_object_kind_is_complete(obj->kind)) {
|
|
|
|
err = drgn_error_incomplete_type("cannot read object with %s type",
|
|
|
|
obj->type);
|
2019-07-28 08:54:45 +01:00
|
|
|
return set_drgn_error(err);
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
underlying_type = drgn_underlying_type(obj->type);
|
|
|
|
switch (obj->kind) {
|
|
|
|
case DRGN_OBJECT_SIGNED: {
|
|
|
|
int64_t svalue;
|
|
|
|
|
|
|
|
err = drgn_object_read_signed(obj, &svalue);
|
|
|
|
if (err)
|
|
|
|
return set_drgn_error(err);
|
|
|
|
return PyLong_FromLongLong(svalue);
|
|
|
|
}
|
|
|
|
case DRGN_OBJECT_UNSIGNED: {
|
|
|
|
uint64_t uvalue;
|
|
|
|
|
|
|
|
err = drgn_object_read_unsigned(obj, &uvalue);
|
|
|
|
if (err)
|
|
|
|
return set_drgn_error(err);
|
|
|
|
if (drgn_type_kind(underlying_type) == DRGN_TYPE_BOOL)
|
|
|
|
return PyBool_FromLong(!!uvalue);
|
|
|
|
else
|
|
|
|
return PyLong_FromUnsignedLongLong(uvalue);
|
|
|
|
}
|
|
|
|
case DRGN_OBJECT_FLOAT: {
|
|
|
|
double fvalue;
|
|
|
|
|
|
|
|
err = drgn_object_read_float(obj, &fvalue);
|
|
|
|
if (err)
|
|
|
|
return set_drgn_error(err);
|
|
|
|
return PyFloat_FromDouble(fvalue);
|
|
|
|
}
|
|
|
|
case DRGN_OBJECT_BUFFER:
|
|
|
|
switch (drgn_type_kind(underlying_type)) {
|
|
|
|
case DRGN_TYPE_COMPLEX:
|
|
|
|
PyErr_SetString(PyExc_NotImplementedError,
|
|
|
|
"reading complex type is not implemented");
|
|
|
|
return NULL;
|
|
|
|
case DRGN_TYPE_STRUCT:
|
|
|
|
case DRGN_TYPE_UNION:
|
2019-11-15 01:12:47 +00:00
|
|
|
case DRGN_TYPE_CLASS:
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
return DrgnObject_compound_value(obj, underlying_type);
|
|
|
|
case DRGN_TYPE_ARRAY:
|
|
|
|
return DrgnObject_array_value(obj, underlying_type);
|
|
|
|
default:
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
DRGN_UNREACHABLE();
|
|
|
|
}
|
|
|
|
|
|
|
|
static PyObject *DrgnObject_value(DrgnObject *self)
|
|
|
|
{
|
|
|
|
return DrgnObject_value_impl(&self->obj);
|
|
|
|
}
|
|
|
|
|
|
|
|
static PyObject *DrgnObject_string(DrgnObject *self)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
char *str;
|
|
|
|
PyObject *ret;
|
|
|
|
|
|
|
|
err = drgn_object_read_c_string(&self->obj, &str);
|
|
|
|
if (err)
|
|
|
|
return set_drgn_error(err);
|
|
|
|
|
|
|
|
ret = PyBytes_FromString(str);
|
|
|
|
free(str);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static DrgnObject *DrgnObject_address_of(DrgnObject *self)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
DrgnObject *res;
|
|
|
|
|
|
|
|
res = DrgnObject_alloc(DrgnObject_prog(self));
|
|
|
|
if (!res)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
err = drgn_object_address_of(&res->obj, &self->obj);
|
|
|
|
if (err) {
|
|
|
|
Py_DECREF(res);
|
2019-07-28 08:54:45 +01:00
|
|
|
return set_drgn_error(err);
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
}
|
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
|
|
|
static DrgnObject *DrgnObject_read(DrgnObject *self)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
DrgnObject *res;
|
|
|
|
|
|
|
|
if (!self->obj.is_reference) {
|
|
|
|
Py_INCREF(self);
|
|
|
|
return self;
|
|
|
|
}
|
|
|
|
|
|
|
|
res = DrgnObject_alloc(DrgnObject_prog(self));
|
|
|
|
if (!res)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
err = drgn_object_read(&res->obj, &self->obj);
|
|
|
|
if (err) {
|
|
|
|
Py_DECREF(res);
|
2019-07-28 08:54:45 +01:00
|
|
|
return set_drgn_error(err);
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
}
|
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
|
|
|
static PyObject *DrgnObject_repr(DrgnObject *self)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
PyObject *parts, *tmp, *sep, *ret = NULL;
|
|
|
|
char *type_name;
|
|
|
|
|
|
|
|
parts = PyList_New(0);
|
|
|
|
if (!parts)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
err = drgn_pretty_print_type_name(drgn_object_qualified_type(&self->obj),
|
|
|
|
&type_name);
|
|
|
|
if (err) {
|
|
|
|
set_drgn_error(err);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
tmp = PyUnicode_FromString(type_name);
|
|
|
|
free(type_name);
|
|
|
|
if (!tmp)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
if (append_format(parts, "Object(prog, %R, ", tmp) == -1) {
|
|
|
|
Py_DECREF(tmp);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
Py_DECREF(tmp);
|
|
|
|
|
|
|
|
if (self->obj.is_reference) {
|
|
|
|
char buf[17];
|
|
|
|
|
|
|
|
snprintf(buf, sizeof(buf), "%" PRIx64,
|
|
|
|
self->obj.reference.address);
|
|
|
|
if (append_format(parts, "address=0x%s", buf) == -1)
|
|
|
|
goto out;
|
|
|
|
} else {
|
|
|
|
PyObject *value_obj;
|
|
|
|
|
2019-04-05 18:36:03 +01:00
|
|
|
if (append_string(parts, "value=") == -1)
|
|
|
|
goto out;
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
value_obj = DrgnObject_value(self);
|
|
|
|
if (!value_obj)
|
|
|
|
goto out;
|
2019-04-05 18:36:03 +01:00
|
|
|
if (drgn_type_kind(drgn_underlying_type(self->obj.type)) ==
|
|
|
|
DRGN_TYPE_POINTER)
|
|
|
|
tmp = PyNumber_ToBase(value_obj, 16);
|
|
|
|
else
|
|
|
|
tmp = PyObject_Repr(value_obj);
|
|
|
|
Py_DECREF(value_obj);
|
|
|
|
if (!tmp)
|
|
|
|
goto out;
|
|
|
|
if (PyList_Append(parts, tmp) == -1) {
|
|
|
|
Py_DECREF(tmp);
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
goto out;
|
|
|
|
}
|
2019-04-05 18:36:03 +01:00
|
|
|
Py_DECREF(tmp);
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
}
|
|
|
|
|
2019-07-29 08:57:28 +01:00
|
|
|
if ((self->obj.is_reference || self->obj.kind == DRGN_OBJECT_BUFFER) &&
|
|
|
|
self->obj.prog->has_platform) {
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
bool little_endian;
|
|
|
|
|
|
|
|
if (self->obj.is_reference)
|
|
|
|
little_endian = self->obj.reference.little_endian;
|
|
|
|
else
|
|
|
|
little_endian = self->obj.value.little_endian;
|
|
|
|
if (little_endian !=
|
2019-07-29 08:57:28 +01:00
|
|
|
drgn_program_is_little_endian(self->obj.prog) &&
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
append_format(parts, ", byteorder='%s'",
|
|
|
|
little_endian ? "little" : "big") == -1)
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (self->obj.is_reference || self->obj.kind == DRGN_OBJECT_BUFFER) {
|
|
|
|
uint8_t bit_offset;
|
|
|
|
|
|
|
|
if (self->obj.is_reference)
|
|
|
|
bit_offset = self->obj.reference.bit_offset;
|
|
|
|
else
|
|
|
|
bit_offset = self->obj.value.bit_offset;
|
|
|
|
if (bit_offset &&
|
|
|
|
append_format(parts, ", bit_offset=%d", bit_offset) == -1)
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (self->obj.is_bit_field &&
|
|
|
|
append_format(parts, ", bit_field_size=%llu",
|
|
|
|
(unsigned long long)self->obj.bit_size) == -1)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
if (append_string(parts, ")") == -1)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
sep = PyUnicode_New(0, 0);
|
|
|
|
if (!sep)
|
|
|
|
goto out;
|
|
|
|
ret = PyUnicode_Join(sep, parts);
|
|
|
|
Py_DECREF(sep);
|
|
|
|
out:
|
|
|
|
Py_DECREF(parts);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static PyObject *DrgnObject_str(DrgnObject *self)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
char *str;
|
|
|
|
PyObject *ret;
|
|
|
|
|
|
|
|
err = drgn_pretty_print_object(&self->obj, SIZE_MAX, &str);
|
|
|
|
if (err)
|
|
|
|
return set_drgn_error(err);
|
|
|
|
|
|
|
|
ret = PyUnicode_FromString(str);
|
|
|
|
free(str);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static PyObject *DrgnObject_format(DrgnObject *self, PyObject *arg)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
const char *format_spec;
|
|
|
|
size_t columns = SIZE_MAX;
|
|
|
|
char *end, *str;
|
|
|
|
PyObject *ret;
|
|
|
|
|
|
|
|
format_spec = PyUnicode_AsUTF8(arg);
|
|
|
|
if (!format_spec)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
if (format_spec[0] == '.') {
|
|
|
|
unsigned long tmp;
|
|
|
|
|
|
|
|
if (format_spec[1] == '-') {
|
|
|
|
PyErr_SetString(PyExc_ValueError,
|
|
|
|
"Format specifier precision must be non-negative");
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
errno = 0;
|
|
|
|
tmp = strtoul(format_spec + 1, &end, 10);
|
|
|
|
if (tmp == ULONG_MAX && errno == ERANGE) {
|
|
|
|
PyErr_SetString(PyExc_OverflowError,
|
|
|
|
"Format specifier precision is too large");
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
columns = tmp;
|
|
|
|
} else {
|
|
|
|
end = (char *)format_spec;
|
|
|
|
}
|
|
|
|
if (*end) {
|
|
|
|
PyErr_SetString(PyExc_ValueError,
|
|
|
|
"Format specifier can only include precision");
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
err = drgn_pretty_print_object(&self->obj, columns, &str);
|
|
|
|
if (err)
|
|
|
|
return set_drgn_error(err);
|
|
|
|
|
|
|
|
ret = PyUnicode_FromString(str);
|
|
|
|
free(str);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static Program *DrgnObject_get_prog(DrgnObject *self, void *arg)
|
|
|
|
{
|
|
|
|
Py_INCREF(DrgnObject_prog(self));
|
|
|
|
return DrgnObject_prog(self);
|
|
|
|
}
|
|
|
|
|
|
|
|
static PyObject *DrgnObject_get_type(DrgnObject *self, void *arg)
|
|
|
|
{
|
|
|
|
return DrgnType_wrap(drgn_object_qualified_type(&self->obj),
|
2019-04-30 20:38:41 +01:00
|
|
|
DrgnObject_prog(self)->objects);
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static PyObject *DrgnObject_get_address(DrgnObject *self, void *arg)
|
|
|
|
{
|
|
|
|
if (self->obj.is_reference)
|
|
|
|
return PyLong_FromUnsignedLongLong(self->obj.reference.address);
|
|
|
|
else
|
|
|
|
Py_RETURN_NONE;
|
|
|
|
}
|
|
|
|
|
|
|
|
static PyObject *DrgnObject_get_byteorder(DrgnObject *self, void *arg)
|
|
|
|
{
|
|
|
|
if (self->obj.is_reference)
|
|
|
|
return byteorder_string(self->obj.reference.little_endian);
|
|
|
|
else if (self->obj.kind == DRGN_OBJECT_BUFFER)
|
|
|
|
return byteorder_string(self->obj.value.little_endian);
|
|
|
|
else
|
|
|
|
Py_RETURN_NONE;
|
|
|
|
}
|
|
|
|
|
|
|
|
static PyObject *DrgnObject_get_bit_offset(DrgnObject *self, void *arg)
|
|
|
|
{
|
|
|
|
if (self->obj.is_reference)
|
|
|
|
return PyLong_FromLong(self->obj.reference.bit_offset);
|
|
|
|
else if (self->obj.kind == DRGN_OBJECT_BUFFER)
|
|
|
|
return PyLong_FromLong(self->obj.value.bit_offset);
|
|
|
|
else
|
|
|
|
Py_RETURN_NONE;
|
|
|
|
}
|
|
|
|
|
|
|
|
static PyObject *DrgnObject_get_bit_field_size(DrgnObject *self, void *arg)
|
|
|
|
{
|
|
|
|
if (self->obj.is_bit_field)
|
|
|
|
return PyLong_FromUnsignedLongLong(self->obj.bit_size);
|
|
|
|
else
|
|
|
|
Py_RETURN_NONE;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int DrgnObject_binary_operand(PyObject *self, PyObject *other,
|
|
|
|
struct drgn_object **obj,
|
|
|
|
struct drgn_object *tmp)
|
|
|
|
{
|
|
|
|
if (PyObject_TypeCheck(self, &DrgnObject_type)) {
|
|
|
|
*obj = &((DrgnObject *)self)->obj;
|
|
|
|
return 0;
|
|
|
|
} else {
|
|
|
|
*obj = tmp;
|
|
|
|
/* If self isn't a DrgnObject, then other must be. */
|
|
|
|
drgn_object_init(tmp, ((DrgnObject *)other)->obj.prog);
|
|
|
|
return DrgnObject_literal(tmp, self);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
#define DrgnObject_BINARY_OP(op) \
|
|
|
|
static PyObject *DrgnObject_##op(PyObject *left, PyObject *right) \
|
|
|
|
{ \
|
|
|
|
struct drgn_error *err; \
|
|
|
|
struct drgn_object *lhs, lhs_tmp, *rhs, rhs_tmp; \
|
|
|
|
DrgnObject *res = NULL; \
|
|
|
|
int ret; \
|
|
|
|
\
|
|
|
|
ret = DrgnObject_binary_operand(left, right, &lhs, &lhs_tmp); \
|
|
|
|
if (ret) \
|
|
|
|
goto out; \
|
|
|
|
ret = DrgnObject_binary_operand(right, left, &rhs, &rhs_tmp); \
|
|
|
|
if (ret) \
|
|
|
|
goto out_lhs; \
|
|
|
|
\
|
|
|
|
res = DrgnObject_alloc(container_of(lhs->prog, Program, prog)); \
|
|
|
|
if (!res) { \
|
|
|
|
ret = -1; \
|
|
|
|
goto out_rhs; \
|
|
|
|
} \
|
|
|
|
\
|
|
|
|
err = drgn_object_##op(&res->obj, lhs, rhs); \
|
|
|
|
if (err) { \
|
|
|
|
set_drgn_error(err); \
|
|
|
|
Py_DECREF(res); \
|
|
|
|
ret = -1; \
|
|
|
|
goto out_rhs; \
|
|
|
|
} \
|
|
|
|
\
|
|
|
|
out_rhs: \
|
|
|
|
if (rhs == &rhs_tmp) \
|
|
|
|
drgn_object_deinit(&rhs_tmp); \
|
|
|
|
out_lhs: \
|
|
|
|
if (lhs == &lhs_tmp) \
|
|
|
|
drgn_object_deinit(&lhs_tmp); \
|
|
|
|
out: \
|
|
|
|
if (ret == -1) \
|
|
|
|
return NULL; \
|
|
|
|
else if (ret) \
|
|
|
|
Py_RETURN_NOTIMPLEMENTED; \
|
|
|
|
else \
|
|
|
|
return (PyObject *)res; \
|
|
|
|
}
|
|
|
|
DrgnObject_BINARY_OP(add)
|
|
|
|
DrgnObject_BINARY_OP(sub)
|
|
|
|
DrgnObject_BINARY_OP(mul)
|
|
|
|
DrgnObject_BINARY_OP(div)
|
|
|
|
DrgnObject_BINARY_OP(mod)
|
|
|
|
DrgnObject_BINARY_OP(lshift)
|
|
|
|
DrgnObject_BINARY_OP(rshift)
|
|
|
|
DrgnObject_BINARY_OP(and)
|
|
|
|
DrgnObject_BINARY_OP(or)
|
|
|
|
DrgnObject_BINARY_OP(xor)
|
|
|
|
#undef DrgnObject_BINARY_OP
|
|
|
|
|
|
|
|
#define DrgnObject_UNARY_OP(op) \
|
|
|
|
static DrgnObject *DrgnObject_##op(DrgnObject *self) \
|
|
|
|
{ \
|
|
|
|
struct drgn_error *err; \
|
|
|
|
DrgnObject *res; \
|
|
|
|
\
|
|
|
|
res = DrgnObject_alloc(DrgnObject_prog(self)); \
|
|
|
|
if (!res) \
|
|
|
|
return NULL; \
|
|
|
|
\
|
|
|
|
err = drgn_object_##op(&res->obj, &self->obj); \
|
|
|
|
if (err) { \
|
|
|
|
Py_DECREF(res); \
|
2019-07-28 08:54:45 +01:00
|
|
|
return set_drgn_error(err); \
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
} \
|
|
|
|
return res; \
|
|
|
|
}
|
|
|
|
DrgnObject_UNARY_OP(pos)
|
|
|
|
DrgnObject_UNARY_OP(neg)
|
|
|
|
DrgnObject_UNARY_OP(not)
|
|
|
|
#undef DrgnObject_UNARY_OP
|
|
|
|
|
|
|
|
static int DrgnObject_bool(DrgnObject *self)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
bool ret;
|
|
|
|
|
|
|
|
err = drgn_object_bool(&self->obj, &ret);
|
|
|
|
if (err) {
|
|
|
|
set_drgn_error(err);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static PyObject *DrgnObject_int(DrgnObject *self)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
union drgn_value value_mem;
|
|
|
|
const union drgn_value *value;
|
|
|
|
PyObject *ret;
|
|
|
|
|
|
|
|
if (!drgn_type_is_scalar(self->obj.type)) {
|
|
|
|
return set_error_type_name("cannot convert '%s' to int",
|
|
|
|
drgn_object_qualified_type(&self->obj));
|
|
|
|
}
|
|
|
|
|
|
|
|
err = drgn_object_read_value(&self->obj, &value_mem, &value);
|
2019-07-28 08:54:45 +01:00
|
|
|
if (err)
|
|
|
|
return set_drgn_error(err);
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
|
|
|
|
switch (self->obj.kind) {
|
|
|
|
case DRGN_OBJECT_SIGNED:
|
|
|
|
ret = PyLong_FromLongLong(value->svalue);
|
|
|
|
break;
|
|
|
|
case DRGN_OBJECT_UNSIGNED:
|
|
|
|
ret = PyLong_FromUnsignedLongLong(value->uvalue);
|
|
|
|
break;
|
|
|
|
case DRGN_OBJECT_FLOAT:
|
|
|
|
ret = PyLong_FromDouble(value->fvalue);
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
DRGN_UNREACHABLE();
|
|
|
|
}
|
|
|
|
drgn_object_deinit_value(&self->obj, value);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static PyObject *DrgnObject_float(DrgnObject *self)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
union drgn_value value_mem;
|
|
|
|
const union drgn_value *value;
|
|
|
|
PyObject *ret;
|
|
|
|
|
|
|
|
if (!drgn_type_is_arithmetic(self->obj.type)) {
|
|
|
|
return set_error_type_name("cannot convert '%s' to float",
|
|
|
|
drgn_object_qualified_type(&self->obj));
|
|
|
|
}
|
|
|
|
|
|
|
|
err = drgn_object_read_value(&self->obj, &value_mem, &value);
|
2019-07-28 08:54:45 +01:00
|
|
|
if (err)
|
|
|
|
return set_drgn_error(err);
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
|
|
|
|
switch (self->obj.kind) {
|
|
|
|
case DRGN_OBJECT_SIGNED:
|
|
|
|
ret = PyFloat_FromDouble(value->svalue);
|
|
|
|
break;
|
|
|
|
case DRGN_OBJECT_UNSIGNED:
|
|
|
|
ret = PyFloat_FromDouble(value->uvalue);
|
|
|
|
break;
|
|
|
|
case DRGN_OBJECT_FLOAT:
|
|
|
|
ret = PyFloat_FromDouble(value->fvalue);
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
DRGN_UNREACHABLE();
|
|
|
|
}
|
|
|
|
drgn_object_deinit_value(&self->obj, value);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static PyObject *DrgnObject_index(DrgnObject *self)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
struct drgn_type *underlying_type;
|
|
|
|
union drgn_value value_mem;
|
|
|
|
const union drgn_value *value;
|
|
|
|
PyObject *ret;
|
|
|
|
|
|
|
|
underlying_type = drgn_underlying_type(self->obj.type);
|
|
|
|
if (!drgn_type_is_integer(underlying_type) &&
|
|
|
|
drgn_type_kind(underlying_type) != DRGN_TYPE_POINTER) {
|
|
|
|
return set_error_type_name("cannot convert '%s' to index",
|
|
|
|
drgn_object_qualified_type(&self->obj));
|
|
|
|
}
|
|
|
|
|
|
|
|
err = drgn_object_read_value(&self->obj, &value_mem, &value);
|
2019-07-28 08:54:45 +01:00
|
|
|
if (err)
|
|
|
|
return set_drgn_error(err);
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
|
|
|
|
switch (self->obj.kind) {
|
|
|
|
case DRGN_OBJECT_SIGNED:
|
|
|
|
ret = PyLong_FromLongLong(value->svalue);
|
|
|
|
break;
|
|
|
|
case DRGN_OBJECT_UNSIGNED:
|
|
|
|
ret = PyLong_FromUnsignedLongLong(value->uvalue);
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
DRGN_UNREACHABLE();
|
|
|
|
}
|
|
|
|
drgn_object_deinit_value(&self->obj, value);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static PyObject *DrgnObject_round(DrgnObject *self, PyObject *args,
|
|
|
|
PyObject *kwds)
|
|
|
|
{
|
|
|
|
static char *keywords[] = {"ndigits", NULL};
|
|
|
|
PyObject *ndigits = Py_None, *value, *ret;
|
|
|
|
|
|
|
|
if (!PyArg_ParseTupleAndKeywords(args, kwds, "|O:round", keywords,
|
|
|
|
&ndigits))
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
if (!drgn_type_is_arithmetic(self->obj.type)) {
|
|
|
|
return set_error_type_name("cannot round '%s'",
|
|
|
|
drgn_object_qualified_type(&self->obj));
|
|
|
|
}
|
|
|
|
|
|
|
|
value = DrgnObject_value(self);
|
|
|
|
if (!value)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
if (ndigits == Py_None) {
|
|
|
|
ret = PyObject_CallMethod(value, "__round__", NULL);
|
|
|
|
Py_DECREF(value);
|
|
|
|
} else {
|
|
|
|
PyObject *args, *kwds, *tmp, *type;
|
|
|
|
|
|
|
|
tmp = PyObject_CallMethod(value, "__round__", "O", ndigits);
|
|
|
|
Py_DECREF(value);
|
|
|
|
if (!tmp)
|
|
|
|
return NULL;
|
|
|
|
value = tmp;
|
|
|
|
|
|
|
|
kwds = PyDict_New();
|
|
|
|
if (!kwds) {
|
|
|
|
Py_DECREF(value);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (PyDict_SetItemString(kwds, "value", value) == -1) {
|
|
|
|
Py_DECREF(value);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
Py_DECREF(value);
|
|
|
|
|
|
|
|
type = DrgnObject_get_type(self, NULL);
|
|
|
|
if (!type) {
|
|
|
|
Py_DECREF(kwds);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
args = Py_BuildValue("OO", DrgnObject_prog(self), type);
|
|
|
|
Py_DECREF(type);
|
|
|
|
if (!args) {
|
|
|
|
Py_DECREF(kwds);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = PyObject_Call((PyObject *)&DrgnObject_type, args, kwds);
|
|
|
|
Py_DECREF(args);
|
|
|
|
Py_DECREF(kwds);
|
|
|
|
}
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
#define DrgnObject_round_method(func) \
|
|
|
|
static PyObject *DrgnObject_##func(DrgnObject *self) \
|
|
|
|
{ \
|
|
|
|
struct drgn_error *err; \
|
|
|
|
union drgn_value value_mem; \
|
|
|
|
const union drgn_value *value; \
|
|
|
|
PyObject *ret; \
|
|
|
|
\
|
|
|
|
if (!drgn_type_is_arithmetic(self->obj.type)) { \
|
|
|
|
return set_error_type_name("cannot round '%s'", \
|
|
|
|
drgn_object_qualified_type(&self->obj));\
|
|
|
|
} \
|
|
|
|
\
|
|
|
|
err = drgn_object_read_value(&self->obj, &value_mem, &value); \
|
2019-07-28 08:54:45 +01:00
|
|
|
if (err) \
|
|
|
|
return set_drgn_error(err); \
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
\
|
|
|
|
switch (self->obj.kind) { \
|
|
|
|
case DRGN_OBJECT_SIGNED: \
|
|
|
|
ret = PyLong_FromLongLong(value->svalue); \
|
|
|
|
break; \
|
|
|
|
case DRGN_OBJECT_UNSIGNED: \
|
|
|
|
ret = PyLong_FromUnsignedLongLong(value->uvalue); \
|
|
|
|
break; \
|
|
|
|
case DRGN_OBJECT_FLOAT: \
|
|
|
|
ret = PyLong_FromDouble(func(value->fvalue)); \
|
|
|
|
break; \
|
|
|
|
default: \
|
|
|
|
DRGN_UNREACHABLE(); \
|
|
|
|
} \
|
|
|
|
drgn_object_deinit_value(&self->obj, value); \
|
|
|
|
return ret; \
|
|
|
|
}
|
|
|
|
DrgnObject_round_method(trunc)
|
|
|
|
DrgnObject_round_method(floor)
|
|
|
|
DrgnObject_round_method(ceil)
|
|
|
|
|
|
|
|
static PyObject *DrgnObject_richcompare(PyObject *left, PyObject *right, int op)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
struct drgn_object *lhs, lhs_tmp, *rhs, rhs_tmp;
|
|
|
|
int ret, cmp;
|
|
|
|
|
|
|
|
ret = DrgnObject_binary_operand(left, right, &lhs, &lhs_tmp);
|
|
|
|
if (ret)
|
|
|
|
goto out;
|
|
|
|
ret = DrgnObject_binary_operand(right, left, &rhs, &rhs_tmp);
|
|
|
|
if (ret)
|
|
|
|
goto out_lhs;
|
|
|
|
|
|
|
|
err = drgn_object_cmp(lhs, rhs, &cmp);
|
|
|
|
if (err) {
|
|
|
|
set_drgn_error(err);
|
|
|
|
ret = -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (rhs == &rhs_tmp)
|
|
|
|
drgn_object_deinit(&rhs_tmp);
|
|
|
|
out_lhs:
|
|
|
|
if (lhs == &lhs_tmp)
|
|
|
|
drgn_object_deinit(&lhs_tmp);
|
|
|
|
out:
|
|
|
|
if (ret == -1)
|
|
|
|
return NULL;
|
|
|
|
else if (ret)
|
|
|
|
Py_RETURN_NOTIMPLEMENTED;
|
|
|
|
else
|
|
|
|
Py_RETURN_RICHCOMPARE(cmp, 0, op);
|
|
|
|
}
|
|
|
|
|
|
|
|
static DrgnObject *DrgnObject_member(DrgnObject *self, PyObject *args,
|
|
|
|
PyObject *kwds)
|
|
|
|
{
|
|
|
|
static char *keywords[] = {"name", NULL};
|
|
|
|
struct drgn_error *err;
|
|
|
|
const char *name;
|
|
|
|
DrgnObject *res;
|
|
|
|
|
|
|
|
if (!PyArg_ParseTupleAndKeywords(args, kwds, "s:member_", keywords,
|
|
|
|
&name))
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
res = DrgnObject_alloc(DrgnObject_prog(self));
|
|
|
|
if (!res)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
if (self->obj.kind == DRGN_OBJECT_UNSIGNED) {
|
|
|
|
err = drgn_object_member_dereference(&res->obj, &self->obj,
|
|
|
|
name);
|
|
|
|
} else {
|
|
|
|
err = drgn_object_member(&res->obj, &self->obj, name);
|
|
|
|
}
|
|
|
|
if (err) {
|
|
|
|
Py_DECREF(res);
|
2019-07-28 08:54:45 +01:00
|
|
|
return set_drgn_error(err);
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
}
|
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
|
|
|
static PyObject *DrgnObject_getattro(DrgnObject *self, PyObject *attr_name)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
PyObject *attr;
|
|
|
|
PyObject *exc_type, *exc_value, *exc_traceback;
|
|
|
|
const char *name;
|
|
|
|
DrgnObject *res;
|
|
|
|
|
|
|
|
attr = PyObject_GenericGetAttr((PyObject *)self, attr_name);
|
|
|
|
if (attr || !PyErr_ExceptionMatches(PyExc_AttributeError))
|
|
|
|
return attr;
|
|
|
|
PyErr_Fetch(&exc_type, &exc_value, &exc_traceback);
|
|
|
|
|
|
|
|
name = PyUnicode_AsUTF8(attr_name);
|
|
|
|
if (!name) {
|
|
|
|
res = NULL;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
res = DrgnObject_alloc(DrgnObject_prog(self));
|
|
|
|
if (!res)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
if (self->obj.kind == DRGN_OBJECT_UNSIGNED) {
|
|
|
|
err = drgn_object_member_dereference(&res->obj, &self->obj,
|
|
|
|
name);
|
|
|
|
} else {
|
|
|
|
err = drgn_object_member(&res->obj, &self->obj, name);
|
|
|
|
}
|
|
|
|
if (err) {
|
|
|
|
Py_CLEAR(res);
|
|
|
|
if (err->code == DRGN_ERROR_TYPE) {
|
|
|
|
/*
|
2019-11-15 01:12:47 +00:00
|
|
|
* If the object doesn't have a compound type,
|
|
|
|
* raise the original AttributeError.
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
*/
|
|
|
|
PyErr_Restore(exc_type, exc_value, exc_traceback);
|
|
|
|
return NULL;
|
|
|
|
} else if (err->code == DRGN_ERROR_LOOKUP) {
|
|
|
|
PyErr_SetString(PyExc_AttributeError, err->message);
|
|
|
|
drgn_error_destroy(err);
|
|
|
|
} else {
|
|
|
|
set_drgn_error(err);
|
|
|
|
}
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
out:
|
|
|
|
Py_XDECREF(exc_traceback);
|
|
|
|
Py_XDECREF(exc_value);
|
|
|
|
Py_DECREF(exc_type);
|
|
|
|
return (PyObject *)res;
|
|
|
|
}
|
|
|
|
|
|
|
|
static Py_ssize_t DrgnObject_length(DrgnObject *self)
|
|
|
|
{
|
|
|
|
struct drgn_type *underlying_type;
|
|
|
|
uint64_t length;
|
|
|
|
|
|
|
|
underlying_type = drgn_underlying_type(self->obj.type);
|
|
|
|
if (drgn_type_kind(underlying_type) != DRGN_TYPE_ARRAY ||
|
|
|
|
!drgn_type_is_complete(underlying_type)) {
|
|
|
|
set_error_type_name("'%s' has no len()",
|
|
|
|
drgn_object_qualified_type(&self->obj));
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
length = drgn_type_length(underlying_type);
|
|
|
|
if (length > PY_SSIZE_T_MAX) {
|
|
|
|
PyErr_SetString(PyExc_OverflowError, "length is too large");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
return length;
|
|
|
|
}
|
|
|
|
|
|
|
|
static DrgnObject *DrgnObject_subscript_impl(DrgnObject *self,
|
2019-11-30 04:50:52 +00:00
|
|
|
int64_t index)
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
DrgnObject *res;
|
|
|
|
|
|
|
|
res = DrgnObject_alloc(DrgnObject_prog(self));
|
|
|
|
if (!res)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
err = drgn_object_subscript(&res->obj, &self->obj, index);
|
|
|
|
if (err) {
|
|
|
|
Py_DECREF(res);
|
2019-07-28 08:54:45 +01:00
|
|
|
return set_drgn_error(err);
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
}
|
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
|
|
|
static DrgnObject *DrgnObject_subscript(DrgnObject *self, PyObject *key)
|
|
|
|
{
|
2019-11-30 04:50:52 +00:00
|
|
|
struct index_arg index = { .is_signed = true };
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
|
2019-10-16 02:03:21 +01:00
|
|
|
if (!index_converter(key, &index))
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
return NULL;
|
2019-11-30 04:50:52 +00:00
|
|
|
return DrgnObject_subscript_impl(self, index.svalue);
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static ObjectIterator *DrgnObject_iter(DrgnObject *self)
|
|
|
|
{
|
|
|
|
struct drgn_type *underlying_type;
|
|
|
|
ObjectIterator *it;
|
|
|
|
|
|
|
|
underlying_type = drgn_underlying_type(self->obj.type);
|
|
|
|
if (drgn_type_kind(underlying_type) != DRGN_TYPE_ARRAY ||
|
|
|
|
!drgn_type_is_complete(underlying_type)) {
|
|
|
|
set_error_type_name("'%s' is not iterable",
|
|
|
|
drgn_object_qualified_type(&self->obj));
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
it = (ObjectIterator *)ObjectIterator_type.tp_alloc(&ObjectIterator_type,
|
|
|
|
0);
|
|
|
|
if (!it)
|
|
|
|
return NULL;
|
|
|
|
it->obj = self;
|
|
|
|
Py_INCREF(self);
|
|
|
|
it->length = drgn_type_length(underlying_type);
|
|
|
|
return it;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int add_to_dir(PyObject *dir, struct drgn_type *type)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
struct drgn_type_member *members;
|
|
|
|
size_t num_members, i;
|
|
|
|
|
|
|
|
type = drgn_underlying_type(type);
|
|
|
|
if (!drgn_type_has_members(type))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
members = drgn_type_members(type);
|
|
|
|
num_members = drgn_type_num_members(type);
|
|
|
|
for (i = 0; i < num_members; i++) {
|
|
|
|
struct drgn_type_member *member;
|
|
|
|
|
|
|
|
member = &members[i];
|
|
|
|
if (member->name) {
|
|
|
|
PyObject *str;
|
|
|
|
|
|
|
|
str = PyUnicode_FromString(member->name);
|
|
|
|
if (!str)
|
|
|
|
return -1;
|
|
|
|
if (PyList_Append(dir, str) == -1) {
|
|
|
|
Py_DECREF(str);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
Py_DECREF(str);
|
|
|
|
} else {
|
|
|
|
struct drgn_qualified_type member_type;
|
|
|
|
|
|
|
|
err = drgn_member_type(member, &member_type);
|
|
|
|
if (err) {
|
|
|
|
set_drgn_error(err);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
if (add_to_dir(dir, member_type.type) == -1)
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static PyObject *DrgnObject_dir(DrgnObject *self)
|
|
|
|
{
|
|
|
|
_Py_IDENTIFIER(__dir__);
|
|
|
|
PyObject *method, *dir;
|
|
|
|
struct drgn_type *type;
|
|
|
|
|
|
|
|
method = _PyObject_GetAttrId((PyObject *)Py_TYPE(self)->tp_base,
|
|
|
|
&PyId___dir__);
|
|
|
|
if (!method)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
dir = PyObject_CallFunctionObjArgs(method, self, NULL);
|
|
|
|
Py_DECREF(method);
|
|
|
|
if (!dir)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
type = drgn_underlying_type(self->obj.type);
|
|
|
|
if (drgn_type_kind(type) == DRGN_TYPE_POINTER)
|
|
|
|
type = drgn_type_type(type).type;
|
|
|
|
if (add_to_dir(dir, type) == -1) {
|
|
|
|
Py_DECREF(dir);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
return dir;
|
|
|
|
}
|
|
|
|
|
|
|
|
static PyGetSetDef DrgnObject_getset[] = {
|
2019-04-04 09:27:23 +01:00
|
|
|
{"prog_", (getter)DrgnObject_get_prog, NULL, drgn_Object_prog__DOC},
|
|
|
|
{"type_", (getter)DrgnObject_get_type, NULL, drgn_Object_type__DOC},
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
{"address_", (getter)DrgnObject_get_address, NULL,
|
2019-04-04 09:27:23 +01:00
|
|
|
drgn_Object_address__DOC},
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
{"byteorder_", (getter)DrgnObject_get_byteorder, NULL,
|
2019-04-04 09:27:23 +01:00
|
|
|
drgn_Object_byteorder__DOC},
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
{"bit_offset_", (getter)DrgnObject_get_bit_offset, NULL,
|
2019-04-04 09:27:23 +01:00
|
|
|
drgn_Object_bit_offset__DOC},
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
{"bit_field_size_", (getter)DrgnObject_get_bit_field_size, NULL,
|
2019-04-04 09:27:23 +01:00
|
|
|
drgn_Object_bit_field_size__DOC},
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
{},
|
|
|
|
};
|
|
|
|
|
|
|
|
static PyMethodDef DrgnObject_methods[] = {
|
|
|
|
{"__getitem__", (PyCFunction)DrgnObject_subscript,
|
2019-04-04 09:27:23 +01:00
|
|
|
METH_O | METH_COEXIST, drgn_Object___getitem___DOC},
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
{"value_", (PyCFunction)DrgnObject_value, METH_NOARGS,
|
2019-04-04 09:27:23 +01:00
|
|
|
drgn_Object_value__DOC},
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
{"string_", (PyCFunction)DrgnObject_string, METH_NOARGS,
|
2019-04-04 09:27:23 +01:00
|
|
|
drgn_Object_string__DOC},
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
{"member_", (PyCFunction)DrgnObject_member,
|
2019-04-04 09:27:23 +01:00
|
|
|
METH_VARARGS | METH_KEYWORDS, drgn_Object_member__DOC},
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
{"address_of_", (PyCFunction)DrgnObject_address_of, METH_NOARGS,
|
2019-04-04 09:27:23 +01:00
|
|
|
drgn_Object_address_of__DOC},
|
|
|
|
{"read_", (PyCFunction)DrgnObject_read, METH_NOARGS,
|
|
|
|
drgn_Object_read__DOC},
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
{"__round__", (PyCFunction)DrgnObject_round,
|
|
|
|
METH_VARARGS | METH_KEYWORDS},
|
|
|
|
{"__trunc__", (PyCFunction)DrgnObject_trunc, METH_NOARGS},
|
|
|
|
{"__floor__", (PyCFunction)DrgnObject_floor, METH_NOARGS},
|
|
|
|
{"__ceil__", (PyCFunction)DrgnObject_ceil, METH_NOARGS},
|
2019-04-04 09:27:23 +01:00
|
|
|
{"__dir__", (PyCFunction)DrgnObject_dir, METH_NOARGS,
|
2019-11-15 01:12:47 +00:00
|
|
|
"dir() implementation which includes structure, union, and class members."},
|
2019-04-04 09:27:23 +01:00
|
|
|
{"__format__", (PyCFunction)DrgnObject_format, METH_O,
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
"Object formatter."},
|
|
|
|
{},
|
|
|
|
};
|
|
|
|
|
|
|
|
static PyNumberMethods DrgnObject_as_number = {
|
2019-11-15 07:16:29 +00:00
|
|
|
.nb_add = (binaryfunc)DrgnObject_add,
|
|
|
|
.nb_subtract = (binaryfunc)DrgnObject_sub,
|
|
|
|
.nb_multiply = (binaryfunc)DrgnObject_mul,
|
|
|
|
.nb_remainder = (binaryfunc)DrgnObject_mod,
|
|
|
|
.nb_negative = (unaryfunc)DrgnObject_neg,
|
|
|
|
.nb_positive = (unaryfunc)DrgnObject_pos,
|
|
|
|
.nb_bool = (inquiry)DrgnObject_bool,
|
|
|
|
.nb_invert = (unaryfunc)DrgnObject_not,
|
|
|
|
.nb_lshift = (binaryfunc)DrgnObject_lshift,
|
|
|
|
.nb_rshift = (binaryfunc)DrgnObject_rshift,
|
|
|
|
.nb_and = (binaryfunc)DrgnObject_and,
|
|
|
|
.nb_xor = (binaryfunc)DrgnObject_xor,
|
|
|
|
.nb_or = (binaryfunc)DrgnObject_or,
|
|
|
|
.nb_int = (unaryfunc)DrgnObject_int,
|
|
|
|
.nb_float = (unaryfunc)DrgnObject_float,
|
|
|
|
.nb_true_divide = (binaryfunc)DrgnObject_div,
|
|
|
|
.nb_index = (unaryfunc)DrgnObject_index,
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
};
|
|
|
|
|
|
|
|
static PyMappingMethods DrgnObject_as_mapping = {
|
2019-11-15 07:16:29 +00:00
|
|
|
.mp_length = (lenfunc)DrgnObject_length,
|
|
|
|
.mp_subscript = (binaryfunc)DrgnObject_subscript,
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
};
|
|
|
|
|
|
|
|
PyTypeObject DrgnObject_type = {
|
|
|
|
PyVarObject_HEAD_INIT(NULL, 0)
|
2019-11-15 07:16:29 +00:00
|
|
|
.tp_name = "_drgn.Object",
|
|
|
|
.tp_basicsize = sizeof(DrgnObject),
|
|
|
|
.tp_dealloc = (destructor)DrgnObject_dealloc,
|
|
|
|
.tp_repr = (reprfunc)DrgnObject_repr,
|
|
|
|
.tp_as_number = &DrgnObject_as_number,
|
|
|
|
.tp_as_mapping = &DrgnObject_as_mapping,
|
|
|
|
.tp_str = (reprfunc)DrgnObject_str,
|
|
|
|
.tp_getattro = (getattrofunc)DrgnObject_getattro,
|
|
|
|
.tp_flags = Py_TPFLAGS_DEFAULT,
|
|
|
|
.tp_doc = drgn_Object_DOC,
|
|
|
|
.tp_richcompare = DrgnObject_richcompare,
|
|
|
|
.tp_iter = (getiterfunc)DrgnObject_iter,
|
|
|
|
.tp_methods = DrgnObject_methods,
|
|
|
|
.tp_getset = DrgnObject_getset,
|
|
|
|
.tp_init = (initproc)DrgnObject_init,
|
|
|
|
.tp_new = (newfunc)DrgnObject_new,
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
};
|
|
|
|
|
2019-04-04 09:27:23 +01:00
|
|
|
PyObject *DrgnObject_NULL(PyObject *self, PyObject *args, PyObject *kwds)
|
|
|
|
{
|
|
|
|
static char *keywords[] = {"prog", "type", NULL};
|
|
|
|
PyObject *prog_obj, *type_obj;
|
|
|
|
PyObject *a, *k, *ret;
|
|
|
|
|
|
|
|
if (!PyArg_ParseTupleAndKeywords(args, kwds, "OO:NULL", keywords,
|
|
|
|
&prog_obj, &type_obj))
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
a = Py_BuildValue("OO", prog_obj, type_obj);
|
|
|
|
if (!a)
|
|
|
|
return NULL;
|
|
|
|
k = Py_BuildValue("{s:i}", "value", 0);
|
|
|
|
if (!k) {
|
|
|
|
Py_DECREF(a);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
ret = PyObject_Call((PyObject *)&DrgnObject_type, a, k);
|
|
|
|
Py_DECREF(k);
|
|
|
|
Py_DECREF(a);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
DrgnObject *cast(PyObject *self, PyObject *args, PyObject *kwds)
|
|
|
|
{
|
|
|
|
static char *keywords[] = {"type", "obj", NULL};
|
|
|
|
struct drgn_error *err;
|
|
|
|
struct drgn_qualified_type qualified_type;
|
|
|
|
PyObject *type_obj;
|
|
|
|
DrgnObject *obj, *res;
|
|
|
|
|
|
|
|
if (!PyArg_ParseTupleAndKeywords(args, kwds, "OO!:cast", keywords,
|
|
|
|
&type_obj, &DrgnObject_type, &obj))
|
|
|
|
return NULL;
|
|
|
|
|
2019-04-12 07:26:39 +01:00
|
|
|
if (Program_type_arg(DrgnObject_prog(obj), type_obj, false,
|
|
|
|
&qualified_type) == -1)
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
return NULL;
|
|
|
|
|
|
|
|
res = DrgnObject_alloc(DrgnObject_prog(obj));
|
|
|
|
if (!res)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
err = drgn_object_cast(&res->obj, qualified_type, &obj->obj);
|
|
|
|
if (err) {
|
|
|
|
Py_DECREF(res);
|
2019-07-28 08:54:45 +01:00
|
|
|
return set_drgn_error(err);
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
}
|
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
|
|
|
DrgnObject *reinterpret(PyObject *self, PyObject *args, PyObject *kwds)
|
|
|
|
{
|
|
|
|
static char *keywords[] = {"type", "obj", "byteorder", NULL};
|
|
|
|
struct drgn_error *err;
|
|
|
|
PyObject *type_obj;
|
|
|
|
struct drgn_qualified_type qualified_type;
|
2019-10-16 05:21:21 +01:00
|
|
|
struct byteorder_arg byteorder = {
|
|
|
|
.allow_none = true,
|
|
|
|
.is_none = true,
|
|
|
|
.value = DRGN_PROGRAM_ENDIAN,
|
|
|
|
};
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
DrgnObject *obj, *res;
|
|
|
|
|
2019-10-16 05:21:21 +01:00
|
|
|
if (!PyArg_ParseTupleAndKeywords(args, kwds, "OO!|O&:reinterpret",
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
keywords, &type_obj, &DrgnObject_type,
|
2019-10-16 05:21:21 +01:00
|
|
|
&obj, byteorder_converter,
|
|
|
|
&byteorder))
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
return NULL;
|
|
|
|
|
2019-04-12 07:26:39 +01:00
|
|
|
if (Program_type_arg(DrgnObject_prog(obj), type_obj, false,
|
|
|
|
&qualified_type) == -1)
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
return NULL;
|
|
|
|
|
|
|
|
res = DrgnObject_alloc(DrgnObject_prog(obj));
|
|
|
|
if (!res)
|
|
|
|
return NULL;
|
|
|
|
|
2019-10-16 05:21:21 +01:00
|
|
|
err = drgn_object_reinterpret(&res->obj, qualified_type,
|
|
|
|
byteorder.value, &obj->obj);
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
if (err) {
|
|
|
|
Py_DECREF(res);
|
2019-07-28 08:54:45 +01:00
|
|
|
return set_drgn_error(err);
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
}
|
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
|
|
|
DrgnObject *DrgnObject_container_of(PyObject *self, PyObject *args,
|
|
|
|
PyObject *kwds)
|
|
|
|
{
|
|
|
|
static char *keywords[] = {"ptr", "type", "member", NULL};
|
|
|
|
struct drgn_error *err;
|
|
|
|
DrgnObject *obj, *res;
|
|
|
|
PyObject *type_obj;
|
|
|
|
struct drgn_qualified_type qualified_type;
|
|
|
|
const char *member_designator;
|
|
|
|
|
|
|
|
if (!PyArg_ParseTupleAndKeywords(args, kwds, "O!Os:container_of",
|
|
|
|
keywords, &DrgnObject_type, &obj,
|
|
|
|
&type_obj, &member_designator))
|
|
|
|
return NULL;
|
|
|
|
|
2019-04-12 07:26:39 +01:00
|
|
|
if (Program_type_arg(DrgnObject_prog(obj), type_obj, false,
|
|
|
|
&qualified_type) == -1)
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
return NULL;
|
|
|
|
|
|
|
|
res = DrgnObject_alloc(DrgnObject_prog(obj));
|
|
|
|
if (!res)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
err = drgn_object_container_of(&res->obj, &obj->obj, qualified_type,
|
|
|
|
member_designator);
|
|
|
|
if (err) {
|
|
|
|
Py_DECREF(res);
|
2019-07-28 08:54:45 +01:00
|
|
|
return set_drgn_error(err);
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
}
|
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void ObjectIterator_dealloc(ObjectIterator *self)
|
|
|
|
{
|
|
|
|
Py_DECREF(self->obj);
|
|
|
|
Py_TYPE(self)->tp_free((PyObject *)self);
|
|
|
|
}
|
|
|
|
|
|
|
|
static DrgnObject *ObjectIterator_next(ObjectIterator *self)
|
|
|
|
{
|
|
|
|
if (self->index >= self->length)
|
|
|
|
return NULL;
|
|
|
|
return DrgnObject_subscript_impl(self->obj, self->index++);
|
|
|
|
}
|
|
|
|
|
|
|
|
static PyObject *ObjectIterator_length_hint(ObjectIterator *self)
|
|
|
|
{
|
|
|
|
return PyLong_FromUnsignedLongLong(self->length);
|
|
|
|
}
|
|
|
|
|
|
|
|
static PyMethodDef ObjectIterator_methods[] = {
|
|
|
|
{"__length_hint__", (PyCFunction)ObjectIterator_length_hint,
|
|
|
|
METH_NOARGS},
|
|
|
|
{},
|
|
|
|
};
|
|
|
|
|
|
|
|
PyTypeObject ObjectIterator_type = {
|
|
|
|
PyVarObject_HEAD_INIT(NULL, 0)
|
2019-11-15 07:16:29 +00:00
|
|
|
.tp_name = "_drgn._ObjectIterator",
|
|
|
|
.tp_basicsize = sizeof(ObjectIterator),
|
|
|
|
.tp_dealloc = (destructor)ObjectIterator_dealloc,
|
|
|
|
.tp_flags = Py_TPFLAGS_DEFAULT,
|
|
|
|
.tp_iter = PyObject_SelfIter,
|
|
|
|
.tp_iternext = (iternextfunc)ObjectIterator_next,
|
|
|
|
.tp_methods = ObjectIterator_methods,
|
Rewrite drgn core in C
The current mixed Python/C implementation works well, but it has a
couple of important limitations:
- It's too slow for some common use cases, like iterating over large
data structures.
- It can't be reused in utilities written in other languages.
This replaces the internals with a new library written in C, libdrgn. It
includes Python bindings with mostly the same public interface as
before, with some important improvements:
- Types are now represented by a single Type class rather than the messy
polymorphism in the Python implementation.
- Qualifiers are a bitmask instead of a set of strings.
- Bit fields are not considered a separate type.
- The lvalue/rvalue terminology is replaced with reference/value.
- Structure, union, and array values are better supported.
- Function objects are supported.
- Program distinguishes between lookups of variables, constants, and
functions.
The C rewrite is about 6x as fast as the original Python when using the
Python bindings, and about 8x when using the C API directly.
Currently, the exposed API in C is fairly conservative. In the future,
the memory reader, type index, and object index APIs will probably be
exposed for more flexibility.
2019-03-22 23:27:46 +00:00
|
|
|
};
|