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 <string.h>
|
|
|
|
|
|
|
|
#include "internal.h"
|
|
|
|
#include "language.h"
|
|
|
|
#include "type_index.h"
|
|
|
|
|
|
|
|
/* These functions compare the underlying type by reference, not by value. */
|
|
|
|
|
|
|
|
static struct hash_pair
|
|
|
|
drgn_pointer_type_hash(struct drgn_type * const *key)
|
|
|
|
{
|
|
|
|
struct drgn_qualified_type referenced_type;
|
|
|
|
size_t hash;
|
|
|
|
|
2019-04-21 10:10:33 +01:00
|
|
|
referenced_type = drgn_type_type(*key);
|
|
|
|
hash = hash_combine((uintptr_t)referenced_type.type,
|
|
|
|
referenced_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
|
|
|
return hash_pair_from_avalanching_hash(hash);
|
|
|
|
}
|
|
|
|
|
2019-04-21 10:10:33 +01:00
|
|
|
static bool drgn_pointer_type_eq(struct drgn_type * const *a,
|
|
|
|
struct drgn_type * const *b)
|
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 referenced_a, referenced_b;
|
|
|
|
|
2019-04-21 10:10:33 +01:00
|
|
|
referenced_a = drgn_type_type(*a);
|
|
|
|
referenced_b = drgn_type_type(*b);
|
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 (referenced_a.type == referenced_b.type &&
|
2019-04-21 10:10:33 +01:00
|
|
|
referenced_a.qualifiers == referenced_b.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
|
|
|
}
|
|
|
|
|
|
|
|
DEFINE_HASH_SET_FUNCTIONS(drgn_pointer_type_set, struct drgn_type *,
|
|
|
|
drgn_pointer_type_hash, drgn_pointer_type_eq)
|
|
|
|
|
|
|
|
static struct hash_pair
|
|
|
|
drgn_array_type_hash(struct drgn_type * const *key)
|
|
|
|
{
|
|
|
|
struct drgn_type *type = *key;
|
|
|
|
struct drgn_qualified_type referenced_type;
|
|
|
|
size_t hash;
|
|
|
|
|
|
|
|
referenced_type = drgn_type_type(type);
|
|
|
|
hash = hash_combine((uintptr_t)referenced_type.type,
|
|
|
|
referenced_type.qualifiers);
|
|
|
|
hash = hash_combine(hash, drgn_type_is_complete(type));
|
|
|
|
hash = hash_combine(hash, drgn_type_length(type));
|
|
|
|
return hash_pair_from_avalanching_hash(hash);
|
|
|
|
}
|
|
|
|
|
|
|
|
static bool drgn_array_type_eq(struct drgn_type * const *ap,
|
|
|
|
struct drgn_type * const *bp)
|
|
|
|
{
|
|
|
|
struct drgn_type *a = *ap, *b = *bp;
|
|
|
|
struct drgn_qualified_type referenced_a, referenced_b;
|
|
|
|
|
|
|
|
referenced_a = drgn_type_type(a);
|
|
|
|
referenced_b = drgn_type_type(b);
|
|
|
|
return (referenced_a.type == referenced_b.type &&
|
|
|
|
referenced_a.qualifiers == referenced_b.qualifiers &&
|
|
|
|
drgn_type_is_complete(a) == drgn_type_is_complete(b) &&
|
|
|
|
drgn_type_length(a) == drgn_type_length(b));
|
|
|
|
}
|
|
|
|
|
|
|
|
DEFINE_HASH_SET_FUNCTIONS(drgn_array_type_set, struct drgn_type *,
|
|
|
|
drgn_array_type_hash, drgn_array_type_eq)
|
|
|
|
|
2019-04-22 22:21:25 +01:00
|
|
|
static struct hash_pair drgn_member_hash_pair(const struct drgn_member_key *key)
|
|
|
|
{
|
|
|
|
size_t hash;
|
|
|
|
|
|
|
|
if (key->name)
|
|
|
|
hash = cityhash_size_t(key->name, key->name_len);
|
|
|
|
else
|
|
|
|
hash = 0;
|
|
|
|
hash = hash_combine((uintptr_t)key->type, hash);
|
|
|
|
return hash_pair_from_avalanching_hash(hash);
|
|
|
|
}
|
|
|
|
|
|
|
|
static bool drgn_member_eq(const struct drgn_member_key *a,
|
|
|
|
const struct drgn_member_key *b)
|
|
|
|
{
|
|
|
|
return (a->type == b->type && a->name_len == b->name_len &&
|
|
|
|
(!a->name_len || memcmp(a->name, b->name, a->name_len) == 0));
|
|
|
|
}
|
|
|
|
|
|
|
|
DEFINE_HASH_MAP_FUNCTIONS(drgn_member_map, struct drgn_member_key,
|
|
|
|
struct drgn_member_value, drgn_member_hash_pair,
|
|
|
|
drgn_member_eq)
|
|
|
|
|
|
|
|
DEFINE_HASH_SET_FUNCTIONS(drgn_type_set, struct drgn_type *,
|
|
|
|
hash_pair_ptr_type, hash_table_scalar_eq)
|
|
|
|
|
2019-05-01 16:42:32 +01:00
|
|
|
void drgn_type_index_init(struct drgn_type_index *tindex)
|
2019-04-22 20:30:15 +01:00
|
|
|
{
|
2019-04-28 16:42:28 +01:00
|
|
|
tindex->finders = NULL;
|
|
|
|
memset(tindex->primitive_types, 0, sizeof(tindex->primitive_types));
|
2019-04-22 20:30:15 +01:00
|
|
|
drgn_pointer_type_set_init(&tindex->pointer_types);
|
|
|
|
drgn_array_type_set_init(&tindex->array_types);
|
2019-04-22 22:21:25 +01:00
|
|
|
drgn_member_map_init(&tindex->members);
|
|
|
|
drgn_type_set_init(&tindex->members_cached);
|
2019-05-01 16:42:32 +01:00
|
|
|
tindex->word_size = 0;
|
2019-04-22 20:30:15 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
static void free_pointer_types(struct drgn_type_index *tindex)
|
|
|
|
{
|
|
|
|
struct drgn_pointer_type_set_pos pos;
|
|
|
|
|
|
|
|
pos = drgn_pointer_type_set_first_pos(&tindex->pointer_types);
|
|
|
|
while (pos.item) {
|
|
|
|
free(*pos.item);
|
|
|
|
drgn_pointer_type_set_next_pos(&pos);
|
|
|
|
}
|
|
|
|
drgn_pointer_type_set_deinit(&tindex->pointer_types);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void free_array_types(struct drgn_type_index *tindex)
|
|
|
|
{
|
|
|
|
struct drgn_array_type_set_pos pos;
|
|
|
|
|
|
|
|
pos = drgn_array_type_set_first_pos(&tindex->array_types);
|
|
|
|
while (pos.item) {
|
|
|
|
free(*pos.item);
|
|
|
|
drgn_array_type_set_next_pos(&pos);
|
|
|
|
}
|
|
|
|
drgn_array_type_set_deinit(&tindex->array_types);
|
|
|
|
}
|
|
|
|
|
|
|
|
void drgn_type_index_deinit(struct drgn_type_index *tindex)
|
|
|
|
{
|
2019-04-28 16:42:28 +01:00
|
|
|
struct drgn_type_finder *finder;
|
|
|
|
|
2019-04-22 22:21:25 +01:00
|
|
|
drgn_member_map_deinit(&tindex->members);
|
|
|
|
drgn_type_set_deinit(&tindex->members_cached);
|
2019-04-22 20:30:15 +01:00
|
|
|
free_array_types(tindex);
|
|
|
|
free_pointer_types(tindex);
|
2019-04-28 16:42:28 +01:00
|
|
|
|
|
|
|
finder = tindex->finders;
|
|
|
|
while (finder) {
|
|
|
|
struct drgn_type_finder *next = finder->next;
|
|
|
|
|
|
|
|
free(finder);
|
|
|
|
finder = next;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
struct drgn_error *drgn_type_index_add_finder(struct drgn_type_index *tindex,
|
|
|
|
drgn_type_find_fn fn, void *arg)
|
|
|
|
{
|
|
|
|
struct drgn_type_finder *finder;
|
|
|
|
|
|
|
|
finder = malloc(sizeof(*finder));
|
|
|
|
if (!finder)
|
|
|
|
return &drgn_enomem;
|
|
|
|
finder->fn = fn;
|
|
|
|
finder->arg = arg;
|
|
|
|
finder->next = tindex->finders;
|
|
|
|
tindex->finders = finder;
|
|
|
|
return NULL;
|
2019-04-22 20:30:15 +01:00
|
|
|
}
|
|
|
|
|
2019-04-20 00:05:19 +01:00
|
|
|
/* Default long, unsigned long, size_t, and ptrdiff_t are 64 bits. */
|
|
|
|
static struct drgn_type default_primitive_types[DRGN_PRIMITIVE_TYPE_NUM];
|
|
|
|
/* 32-bit version of long, unsigned long, size_t, and ptrdiff_t. */
|
|
|
|
static struct drgn_type default_primitive_types_32bit[4];
|
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
|
|
|
|
|
|
|
__attribute__((constructor(200)))
|
2019-04-20 00:05:19 +01:00
|
|
|
static void default_primitive_types_init(void)
|
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;
|
|
|
|
size_t i;
|
|
|
|
|
2019-04-20 00:05:19 +01:00
|
|
|
drgn_int_type_init(&default_primitive_types[DRGN_C_TYPE_CHAR],
|
|
|
|
drgn_primitive_type_spellings[DRGN_C_TYPE_CHAR][0],
|
|
|
|
1, true);
|
|
|
|
drgn_int_type_init(&default_primitive_types[DRGN_C_TYPE_SIGNED_CHAR],
|
|
|
|
drgn_primitive_type_spellings[DRGN_C_TYPE_SIGNED_CHAR][0],
|
|
|
|
1, true);
|
|
|
|
drgn_int_type_init(&default_primitive_types[DRGN_C_TYPE_UNSIGNED_CHAR],
|
|
|
|
drgn_primitive_type_spellings[DRGN_C_TYPE_UNSIGNED_CHAR][0],
|
|
|
|
1, false);
|
|
|
|
drgn_int_type_init(&default_primitive_types[DRGN_C_TYPE_SHORT],
|
|
|
|
drgn_primitive_type_spellings[DRGN_C_TYPE_SHORT][0],
|
|
|
|
2, true);
|
|
|
|
drgn_int_type_init(&default_primitive_types[DRGN_C_TYPE_UNSIGNED_SHORT],
|
|
|
|
drgn_primitive_type_spellings[DRGN_C_TYPE_UNSIGNED_SHORT][0],
|
|
|
|
2, false);
|
|
|
|
drgn_int_type_init(&default_primitive_types[DRGN_C_TYPE_INT],
|
|
|
|
drgn_primitive_type_spellings[DRGN_C_TYPE_INT][0], 4,
|
|
|
|
true);
|
|
|
|
drgn_int_type_init(&default_primitive_types[DRGN_C_TYPE_UNSIGNED_INT],
|
|
|
|
drgn_primitive_type_spellings[DRGN_C_TYPE_UNSIGNED_INT][0],
|
|
|
|
4, false);
|
|
|
|
drgn_int_type_init(&default_primitive_types[DRGN_C_TYPE_LONG],
|
|
|
|
drgn_primitive_type_spellings[DRGN_C_TYPE_LONG][0],
|
|
|
|
8, true);
|
|
|
|
drgn_int_type_init(&default_primitive_types[DRGN_C_TYPE_UNSIGNED_LONG],
|
|
|
|
drgn_primitive_type_spellings[DRGN_C_TYPE_UNSIGNED_LONG][0],
|
|
|
|
8, false);
|
|
|
|
drgn_int_type_init(&default_primitive_types[DRGN_C_TYPE_LONG_LONG],
|
|
|
|
drgn_primitive_type_spellings[DRGN_C_TYPE_LONG_LONG][0],
|
|
|
|
8, true);
|
|
|
|
drgn_int_type_init(&default_primitive_types[DRGN_C_TYPE_UNSIGNED_LONG_LONG],
|
|
|
|
drgn_primitive_type_spellings[DRGN_C_TYPE_UNSIGNED_LONG_LONG][0],
|
|
|
|
8, false);
|
|
|
|
drgn_bool_type_init(&default_primitive_types[DRGN_C_TYPE_BOOL],
|
|
|
|
drgn_primitive_type_spellings[DRGN_C_TYPE_BOOL][0],
|
|
|
|
1);
|
|
|
|
drgn_float_type_init(&default_primitive_types[DRGN_C_TYPE_FLOAT],
|
|
|
|
drgn_primitive_type_spellings[DRGN_C_TYPE_FLOAT][0],
|
|
|
|
4);
|
|
|
|
drgn_float_type_init(&default_primitive_types[DRGN_C_TYPE_DOUBLE],
|
|
|
|
drgn_primitive_type_spellings[DRGN_C_TYPE_DOUBLE][0],
|
|
|
|
8);
|
|
|
|
drgn_float_type_init(&default_primitive_types[DRGN_C_TYPE_LONG_DOUBLE],
|
|
|
|
drgn_primitive_type_spellings[DRGN_C_TYPE_LONG_DOUBLE][0],
|
|
|
|
16);
|
|
|
|
qualified_type.type = &default_primitive_types[DRGN_C_TYPE_UNSIGNED_LONG];
|
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
|
|
|
qualified_type.qualifiers = 0;
|
2019-04-20 00:05:19 +01:00
|
|
|
drgn_typedef_type_init(&default_primitive_types[DRGN_C_TYPE_SIZE_T],
|
|
|
|
drgn_primitive_type_spellings[DRGN_C_TYPE_SIZE_T][0],
|
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
|
|
|
qualified_type);
|
2019-04-20 00:05:19 +01:00
|
|
|
qualified_type.type = &default_primitive_types[DRGN_C_TYPE_LONG];
|
|
|
|
drgn_typedef_type_init(&default_primitive_types[DRGN_C_TYPE_PTRDIFF_T],
|
|
|
|
drgn_primitive_type_spellings[DRGN_C_TYPE_PTRDIFF_T][0],
|
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
|
|
|
qualified_type);
|
2019-04-20 00:05:19 +01:00
|
|
|
for (i = 0; i < ARRAY_SIZE(default_primitive_types); i++) {
|
|
|
|
if (drgn_primitive_type_kind[i] == DRGN_TYPE_VOID)
|
|
|
|
continue;
|
|
|
|
assert(drgn_type_primitive(&default_primitive_types[i]) == i);
|
|
|
|
}
|
|
|
|
|
|
|
|
drgn_int_type_init(&default_primitive_types_32bit[0],
|
|
|
|
drgn_primitive_type_spellings[DRGN_C_TYPE_LONG][0],
|
|
|
|
4, true);
|
|
|
|
drgn_int_type_init(&default_primitive_types_32bit[1],
|
|
|
|
drgn_primitive_type_spellings[DRGN_C_TYPE_UNSIGNED_LONG][0],
|
|
|
|
4, false);
|
|
|
|
qualified_type.type = &default_primitive_types_32bit[1];
|
|
|
|
drgn_typedef_type_init(&default_primitive_types_32bit[2],
|
|
|
|
drgn_primitive_type_spellings[DRGN_C_TYPE_SIZE_T][0],
|
|
|
|
qualified_type);
|
|
|
|
qualified_type.type = &default_primitive_types_32bit[0];
|
|
|
|
drgn_typedef_type_init(&default_primitive_types_32bit[3],
|
|
|
|
drgn_primitive_type_spellings[DRGN_C_TYPE_PTRDIFF_T][0],
|
|
|
|
qualified_type);
|
|
|
|
assert(drgn_type_primitive(&default_primitive_types_32bit[0]) ==
|
|
|
|
DRGN_C_TYPE_LONG);
|
|
|
|
assert(drgn_type_primitive(&default_primitive_types_32bit[1]) ==
|
|
|
|
DRGN_C_TYPE_UNSIGNED_LONG);
|
|
|
|
assert(drgn_type_primitive(&default_primitive_types_32bit[2]) ==
|
|
|
|
DRGN_C_TYPE_SIZE_T);
|
|
|
|
assert(drgn_type_primitive(&default_primitive_types_32bit[3]) ==
|
|
|
|
DRGN_C_TYPE_PTRDIFF_T);
|
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
|
|
|
}
|
|
|
|
|
2019-04-28 16:42:28 +01:00
|
|
|
/*
|
|
|
|
* Like @ref drgn_type_index_find_parsed(), but sets <tt>ret->type</tt> to NULL
|
|
|
|
* and returns @c NULL if the type is not found instead of returning a @ref
|
|
|
|
* DRGN_ERROR_LOOKUP error.
|
|
|
|
*/
|
|
|
|
static struct drgn_error *
|
|
|
|
drgn_type_index_find_parsed_internal(struct drgn_type_index *tindex,
|
|
|
|
enum drgn_type_kind kind, const char *name,
|
|
|
|
size_t name_len, const char *filename,
|
|
|
|
struct drgn_qualified_type *ret)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
struct drgn_type_finder *finder;
|
|
|
|
|
|
|
|
finder = tindex->finders;
|
|
|
|
while (finder) {
|
|
|
|
err = finder->fn(kind, name, name_len, filename, finder->arg,
|
|
|
|
ret);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
if (ret->type) {
|
|
|
|
if (drgn_type_kind(ret->type) != kind) {
|
|
|
|
return drgn_error_create(DRGN_ERROR_TYPE,
|
|
|
|
"type find callback returned wrong kind of type");
|
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
finder = finder->next;
|
|
|
|
}
|
|
|
|
ret->type = NULL;
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2019-04-22 20:30:15 +01:00
|
|
|
struct drgn_error *
|
|
|
|
drgn_type_index_find_primitive(struct drgn_type_index *tindex,
|
|
|
|
enum drgn_primitive_type type,
|
|
|
|
struct drgn_type **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
|
|
|
{
|
2019-04-22 20:30:15 +01:00
|
|
|
struct drgn_error *err;
|
|
|
|
struct drgn_qualified_type qualified_type;
|
|
|
|
enum drgn_type_kind kind;
|
|
|
|
const char * const *spellings;
|
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
|
|
|
size_t i;
|
|
|
|
|
2019-04-22 20:30:15 +01:00
|
|
|
if (tindex->primitive_types[type]) {
|
|
|
|
*ret = tindex->primitive_types[type];
|
|
|
|
return 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
|
|
|
}
|
|
|
|
|
2019-04-22 20:30:15 +01:00
|
|
|
kind = drgn_primitive_type_kind[type];
|
|
|
|
if (kind == DRGN_TYPE_VOID) {
|
|
|
|
*ret = &drgn_void_type;
|
|
|
|
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
|
|
|
}
|
|
|
|
|
2019-04-22 20:30:15 +01:00
|
|
|
spellings = drgn_primitive_type_spellings[type];
|
|
|
|
for (i = 0; spellings[i]; i++) {
|
2019-04-28 16:42:28 +01:00
|
|
|
err = drgn_type_index_find_parsed_internal(tindex, kind,
|
|
|
|
spellings[i],
|
|
|
|
strlen(spellings[i]),
|
|
|
|
NULL,
|
|
|
|
&qualified_type);
|
|
|
|
if (err) {
|
2019-04-22 20:30:15 +01:00
|
|
|
return err;
|
2019-04-28 16:42:28 +01:00
|
|
|
} else if (qualified_type.type &&
|
|
|
|
drgn_type_primitive(qualified_type.type) == type) {
|
2019-04-22 20:30:15 +01:00
|
|
|
*ret = qualified_type.type;
|
|
|
|
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
|
|
|
}
|
|
|
|
|
2019-05-01 16:42:32 +01:00
|
|
|
switch (type) {
|
|
|
|
case DRGN_C_TYPE_LONG:
|
|
|
|
i = 0;
|
|
|
|
break;
|
|
|
|
case DRGN_C_TYPE_UNSIGNED_LONG:
|
|
|
|
i = 1;
|
|
|
|
break;
|
|
|
|
case DRGN_C_TYPE_SIZE_T:
|
|
|
|
i = 2;
|
|
|
|
break;
|
|
|
|
case DRGN_C_TYPE_PTRDIFF_T:
|
|
|
|
i = 3;
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
*ret = &default_primitive_types[type];
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!tindex->word_size) {
|
|
|
|
return drgn_error_create(DRGN_ERROR_INVALID_ARGUMENT,
|
|
|
|
"word size has not been set");
|
|
|
|
}
|
|
|
|
|
|
|
|
if (tindex->word_size == 4)
|
|
|
|
*ret = &default_primitive_types_32bit[i];
|
2019-04-22 20:30:15 +01:00
|
|
|
else
|
|
|
|
*ret = &default_primitive_types[type];
|
2019-05-01 16:42:32 +01:00
|
|
|
|
2019-04-22 20:30:15 +01:00
|
|
|
out:
|
|
|
|
tindex->primitive_types[type] = *ret;
|
|
|
|
return 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
|
|
|
}
|
|
|
|
|
2019-04-28 16:42:28 +01:00
|
|
|
struct drgn_error *
|
|
|
|
drgn_type_index_find_parsed(struct drgn_type_index *tindex,
|
|
|
|
enum drgn_type_kind kind, const char *name,
|
|
|
|
size_t name_len, const char *filename,
|
|
|
|
struct drgn_qualified_type *ret)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
int precision;
|
|
|
|
|
|
|
|
err = drgn_type_index_find_parsed_internal(tindex, kind, name, name_len,
|
|
|
|
filename, ret);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
else if (ret->type)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
precision = name_len < INT_MAX ? (int)name_len : INT_MAX;
|
|
|
|
if (filename) {
|
|
|
|
return drgn_error_format(DRGN_ERROR_LOOKUP,
|
|
|
|
"could not find '%s %.*s' in '%s'",
|
|
|
|
drgn_type_kind_spelling[kind], precision, name,
|
|
|
|
filename);
|
|
|
|
} else {
|
|
|
|
return drgn_error_format(DRGN_ERROR_LOOKUP,
|
|
|
|
"could not find '%s %.*s'",
|
|
|
|
drgn_type_kind_spelling[kind], precision, name);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
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 *
|
2019-04-21 10:10:33 +01:00
|
|
|
drgn_type_index_pointer_type(struct drgn_type_index *tindex,
|
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 referenced_type,
|
|
|
|
struct drgn_type **ret)
|
|
|
|
{
|
|
|
|
struct drgn_type key, *type = &key;
|
|
|
|
struct drgn_pointer_type_set_pos pos;
|
|
|
|
struct hash_pair hp;
|
|
|
|
|
2019-05-01 16:42:32 +01:00
|
|
|
if (!tindex->word_size) {
|
|
|
|
return drgn_error_create(DRGN_ERROR_INVALID_ARGUMENT,
|
|
|
|
"word size has not been set");
|
|
|
|
}
|
|
|
|
|
2019-04-21 10:10:33 +01:00
|
|
|
drgn_pointer_type_init(type, tindex->word_size, referenced_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
|
|
|
hp = drgn_pointer_type_set_hash(&type);
|
|
|
|
pos = drgn_pointer_type_set_search_pos(&tindex->pointer_types, &type,
|
|
|
|
hp);
|
|
|
|
if (pos.item) {
|
|
|
|
*ret = *pos.item;
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
type = malloc(sizeof(*type));
|
|
|
|
if (!type)
|
|
|
|
return &drgn_enomem;
|
|
|
|
*type = key;
|
|
|
|
|
|
|
|
if (!drgn_pointer_type_set_insert_searched(&tindex->pointer_types,
|
|
|
|
&type, hp)) {
|
|
|
|
free(type);
|
|
|
|
return &drgn_enomem;
|
|
|
|
}
|
|
|
|
*ret = type;
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct drgn_error *
|
|
|
|
drgn_type_index_array_type_internal(struct drgn_type_index *tindex,
|
|
|
|
struct drgn_type *key,
|
|
|
|
struct drgn_type **ret)
|
|
|
|
{
|
|
|
|
struct drgn_type *type;
|
|
|
|
struct drgn_array_type_set_pos pos;
|
|
|
|
struct hash_pair hp;
|
|
|
|
|
|
|
|
hp = drgn_array_type_set_hash(&key);
|
|
|
|
pos = drgn_array_type_set_search_pos(&tindex->array_types, &key, hp);
|
|
|
|
if (pos.item) {
|
|
|
|
*ret = *pos.item;
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
type = malloc(sizeof(*type));
|
|
|
|
if (!type)
|
|
|
|
return &drgn_enomem;
|
|
|
|
*type = *key;
|
|
|
|
|
|
|
|
if (!drgn_array_type_set_insert_searched(&tindex->array_types, &type,
|
|
|
|
hp)) {
|
|
|
|
free(type);
|
|
|
|
return &drgn_enomem;
|
|
|
|
}
|
|
|
|
*ret = type;
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
struct drgn_error *
|
|
|
|
drgn_type_index_array_type(struct drgn_type_index *tindex, uint64_t length,
|
|
|
|
struct drgn_qualified_type element_type,
|
|
|
|
struct drgn_type **ret)
|
|
|
|
{
|
|
|
|
struct drgn_type key;
|
|
|
|
|
|
|
|
drgn_array_type_init(&key, length, element_type);
|
|
|
|
return drgn_type_index_array_type_internal(tindex, &key, ret);
|
|
|
|
}
|
|
|
|
|
|
|
|
struct drgn_error *
|
|
|
|
drgn_type_index_incomplete_array_type(struct drgn_type_index *tindex,
|
|
|
|
struct drgn_qualified_type element_type,
|
|
|
|
struct drgn_type **ret)
|
|
|
|
{
|
|
|
|
struct drgn_type key;
|
|
|
|
|
|
|
|
drgn_array_type_init_incomplete(&key, element_type);
|
|
|
|
return drgn_type_index_array_type_internal(tindex, &key, ret);
|
|
|
|
}
|
|
|
|
|
2019-04-22 22:21:25 +01:00
|
|
|
static struct drgn_error *
|
|
|
|
drgn_type_index_cache_members(struct drgn_type_index *tindex,
|
|
|
|
struct drgn_type *outer_type,
|
|
|
|
struct drgn_type *type, uint64_t bit_offset)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
struct drgn_type_member *members;
|
|
|
|
size_t num_members, i;
|
|
|
|
|
|
|
|
if (!drgn_type_has_members(type))
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
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) {
|
|
|
|
struct drgn_member_key key = {
|
|
|
|
.type = outer_type,
|
|
|
|
.name = member->name,
|
|
|
|
.name_len = strlen(member->name),
|
|
|
|
};
|
|
|
|
struct drgn_member_value value = {
|
|
|
|
.type = &member->type,
|
|
|
|
.bit_offset = bit_offset + member->bit_offset,
|
|
|
|
.bit_field_size = member->bit_field_size,
|
|
|
|
};
|
|
|
|
|
|
|
|
if (!drgn_member_map_insert(&tindex->members, &key,
|
|
|
|
&value))
|
|
|
|
return &drgn_enomem;
|
|
|
|
} else {
|
|
|
|
struct drgn_qualified_type member_type;
|
|
|
|
|
|
|
|
err = drgn_member_type(member, &member_type);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
err = drgn_type_index_cache_members(tindex, outer_type,
|
|
|
|
member_type.type,
|
|
|
|
bit_offset +
|
|
|
|
member->bit_offset);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
struct drgn_error *drgn_type_index_find_member(struct drgn_type_index *tindex,
|
|
|
|
struct drgn_type *type,
|
|
|
|
const char *member_name,
|
|
|
|
size_t member_name_len,
|
|
|
|
struct drgn_member_value **ret)
|
|
|
|
{
|
|
|
|
struct drgn_error *err;
|
|
|
|
const struct drgn_member_key key = {
|
|
|
|
.type = drgn_underlying_type(type),
|
|
|
|
.name = member_name,
|
|
|
|
.name_len = member_name_len,
|
|
|
|
};
|
|
|
|
struct hash_pair hp, cached_hp;
|
|
|
|
struct drgn_member_value *value;
|
|
|
|
|
|
|
|
hp = drgn_member_map_hash(&key);
|
|
|
|
value = drgn_member_map_search_hashed(&tindex->members, &key, hp);
|
|
|
|
if (value) {
|
|
|
|
*ret = value;
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Cache miss. One of the following is true:
|
|
|
|
*
|
|
|
|
* 1. The type isn't a structure or union, which is a type error.
|
|
|
|
* 2. The type hasn't been cached, which means we need to cache it and
|
|
|
|
* check again.
|
|
|
|
* 3. The type has already been cached, which means the member doesn't
|
|
|
|
* exist.
|
|
|
|
*/
|
|
|
|
if (!drgn_type_has_members(key.type)) {
|
|
|
|
return drgn_type_error("'%s' is not a structure or union",
|
|
|
|
type);
|
|
|
|
}
|
|
|
|
cached_hp = drgn_type_set_hash(&key.type);
|
|
|
|
if (drgn_type_set_search_hashed(&tindex->members_cached, &key.type,
|
|
|
|
cached_hp))
|
|
|
|
return drgn_error_member_not_found(type, member_name);
|
|
|
|
|
|
|
|
err = drgn_type_index_cache_members(tindex, key.type, key.type, 0);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
if (!drgn_type_set_insert_searched(&tindex->members_cached, &key.type,
|
|
|
|
cached_hp))
|
|
|
|
return &drgn_enomem;
|
|
|
|
|
|
|
|
value = drgn_member_map_search_hashed(&tindex->members, &key, hp);
|
|
|
|
if (value) {
|
|
|
|
*ret = value;
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
return drgn_error_member_not_found(type, member_name);
|
|
|
|
}
|