Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
// Copyright 2019 the V8 project authors. All rights reserved.
|
|
|
|
// Use of this source code is governed by a BSD-style license that can be
|
|
|
|
// found in the LICENSE file.
|
|
|
|
|
|
|
|
#include <sstream>
|
|
|
|
|
|
|
|
#include "debug-helper-internal.h"
|
|
|
|
#include "heap-constants.h"
|
|
|
|
#include "include/v8-internal.h"
|
|
|
|
#include "src/common/ptr-compr-inl.h"
|
2019-08-21 15:37:47 +00:00
|
|
|
#include "src/objects/string-inl.h"
|
2019-08-26 21:58:03 +00:00
|
|
|
#include "src/strings/unicode-inl.h"
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
#include "torque-generated/class-debug-readers-tq.h"
|
|
|
|
|
|
|
|
namespace i = v8::internal;
|
|
|
|
|
2020-02-27 18:14:53 +00:00
|
|
|
namespace v8 {
|
|
|
|
namespace internal {
|
|
|
|
namespace debug_helper_internal {
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
|
2019-11-22 16:52:09 +00:00
|
|
|
constexpr char kObject[] = "v8::internal::Object";
|
|
|
|
constexpr char kTaggedValue[] = "v8::internal::TaggedValue";
|
|
|
|
constexpr char kSmi[] = "v8::internal::Smi";
|
|
|
|
constexpr char kHeapObject[] = "v8::internal::HeapObject";
|
|
|
|
#ifdef V8_COMPRESS_POINTERS
|
|
|
|
constexpr char kObjectAsStoredInHeap[] = "v8::internal::TaggedValue";
|
|
|
|
#else
|
|
|
|
constexpr char kObjectAsStoredInHeap[] = "v8::internal::Object";
|
|
|
|
#endif
|
|
|
|
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
std::string AppendAddressAndType(const std::string& brief, uintptr_t address,
|
|
|
|
const char* type) {
|
|
|
|
std::stringstream brief_stream;
|
|
|
|
brief_stream << "0x" << std::hex << address << " <" << type << ">";
|
|
|
|
return brief.empty() ? brief_stream.str()
|
|
|
|
: brief + " (" + brief_stream.str() + ")";
|
|
|
|
}
|
|
|
|
|
2019-11-22 16:52:09 +00:00
|
|
|
std::string JoinWithSpace(const std::string& a, const std::string& b) {
|
|
|
|
return a.empty() || b.empty() ? a + b : a + " " + b;
|
|
|
|
}
|
|
|
|
|
2019-08-26 21:58:03 +00:00
|
|
|
struct TypedObject {
|
|
|
|
TypedObject(d::TypeCheckResult type_check_result,
|
|
|
|
std::unique_ptr<TqObject> object)
|
|
|
|
: type_check_result(type_check_result), object(std::move(object)) {}
|
2019-09-19 20:23:39 +00:00
|
|
|
|
|
|
|
// How we discovered the object's type, or why we failed to do so.
|
2019-08-21 15:37:47 +00:00
|
|
|
d::TypeCheckResult type_check_result;
|
2019-09-19 20:23:39 +00:00
|
|
|
|
|
|
|
// Pointer to some TqObject subclass, representing the most specific known
|
|
|
|
// type for the object.
|
2019-08-26 21:58:03 +00:00
|
|
|
std::unique_ptr<TqObject> object;
|
2019-09-19 20:23:39 +00:00
|
|
|
|
|
|
|
// Collection of other guesses at more specific types than the one represented
|
|
|
|
// by |object|.
|
|
|
|
std::vector<TypedObject> possible_types;
|
2019-08-21 15:37:47 +00:00
|
|
|
};
|
|
|
|
|
2019-08-26 21:58:03 +00:00
|
|
|
TypedObject GetTypedObjectByHint(uintptr_t address,
|
|
|
|
std::string type_hint_string) {
|
2019-09-10 10:12:00 +00:00
|
|
|
#define TYPE_NAME_CASE(ClassName, ...) \
|
|
|
|
if (type_hint_string == "v8::internal::" #ClassName) { \
|
|
|
|
return {d::TypeCheckResult::kUsedTypeHint, \
|
|
|
|
std::make_unique<Tq##ClassName>(address)}; \
|
2019-08-21 15:37:47 +00:00
|
|
|
}
|
|
|
|
|
[torque] Use generated instance types, part 1
This change begins making use of the fact that Torque now knows about
the relationship between classes and instance types, to replace a few
repetitive lists:
- Instance type checkers (single and range), defined in
src/objects/instance-type.h
- Verification dispatch in src/diagnostics/objects-debug.cc
- Printer dispatch in src/diagnostics/objects-printer.cc
- Postmortem object type detection in
tools/debug_helper/get-object-properties.cc
Torque is updated to generate four macro lists for the instance types,
representing all of the classes separated in two dimensions: classes
that correspond to a single instance type versus those that have a
range, and classes that are fully defined in Torque (with fields and
methods inside '{}') versus those that are only declared. The latter
distinction is useful because fully-defined classes are guaranteed to
correspond to real C++ classes, whereas only-declared classes are not.
A few other changes were required to make the lists above work:
- Renamed IsFiller to IsFreeSpaceOrFiller to better reflect what it does
and avoid conflicts with the new macro-generated IsFiller method. This
is the part I'm most worried about: I think the new name is an
improvement for clarity and consistency, but I could imagine someone
typing IsFiller out of habit and introducing a bug. If we'd prefer to
keep the name IsFiller, my other idea is to rename FreeSpace to
VariableSizeFiller and Filler to FixedSizeFiller.
- Made Tuple3 extend from Struct, not Tuple2, because IsTuple2 is
expected to check for only TUPLE2_TYPE and not include TUPLE3_TYPE.
- Normalized the dispatched behavior for BigIntBase and HeapNumber.
- Added a few new object printers.
Bug: v8:7793
Change-Id: I5462bb105f8a314baa59bd6ab6ab6215df6f313c
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1860314
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Reviewed-by: Dan Elphick <delphick@chromium.org>
Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Cr-Commit-Position: refs/heads/master@{#64597}
2019-10-28 17:42:41 +00:00
|
|
|
TORQUE_INSTANCE_CHECKERS_SINGLE_FULLY_DEFINED(TYPE_NAME_CASE)
|
|
|
|
TORQUE_INSTANCE_CHECKERS_RANGE_FULLY_DEFINED(TYPE_NAME_CASE)
|
2019-09-19 20:23:39 +00:00
|
|
|
STRING_CLASS_TYPES(TYPE_NAME_CASE)
|
2019-08-21 15:37:47 +00:00
|
|
|
|
|
|
|
#undef TYPE_NAME_CASE
|
|
|
|
|
2019-08-26 21:58:03 +00:00
|
|
|
return {d::TypeCheckResult::kUnknownTypeHint,
|
2019-09-10 10:12:00 +00:00
|
|
|
std::make_unique<TqHeapObject>(address)};
|
2019-08-21 15:37:47 +00:00
|
|
|
}
|
|
|
|
|
2019-09-19 20:23:39 +00:00
|
|
|
TypedObject GetTypedObjectForString(uintptr_t address, i::InstanceType type,
|
|
|
|
d::TypeCheckResult type_source) {
|
2019-08-21 15:37:47 +00:00
|
|
|
class StringGetDispatcher : public i::AllStatic {
|
|
|
|
public:
|
2019-09-19 20:23:39 +00:00
|
|
|
#define DEFINE_METHOD(ClassName) \
|
|
|
|
static inline TypedObject Handle##ClassName( \
|
|
|
|
uintptr_t address, d::TypeCheckResult type_source) { \
|
|
|
|
return {type_source, std::make_unique<Tq##ClassName>(address)}; \
|
2019-08-21 15:37:47 +00:00
|
|
|
}
|
|
|
|
STRING_CLASS_TYPES(DEFINE_METHOD)
|
|
|
|
#undef DEFINE_METHOD
|
2019-09-19 20:23:39 +00:00
|
|
|
static inline TypedObject HandleInvalidString(
|
|
|
|
uintptr_t address, d::TypeCheckResult type_source) {
|
2019-08-26 21:58:03 +00:00
|
|
|
return {d::TypeCheckResult::kUnknownInstanceType,
|
2019-09-10 10:12:00 +00:00
|
|
|
std::make_unique<TqString>(address)};
|
2019-08-21 15:37:47 +00:00
|
|
|
}
|
|
|
|
};
|
|
|
|
|
|
|
|
return i::StringShape(type)
|
2019-08-26 21:58:03 +00:00
|
|
|
.DispatchToSpecificTypeWithoutCast<StringGetDispatcher, TypedObject>(
|
2019-09-19 20:23:39 +00:00
|
|
|
address, type_source);
|
|
|
|
}
|
|
|
|
|
|
|
|
TypedObject GetTypedObjectByInstanceType(uintptr_t address,
|
|
|
|
i::InstanceType type,
|
|
|
|
d::TypeCheckResult type_source) {
|
|
|
|
switch (type) {
|
|
|
|
#define INSTANCE_TYPE_CASE(ClassName, INSTANCE_TYPE) \
|
|
|
|
case i::INSTANCE_TYPE: \
|
|
|
|
return {type_source, std::make_unique<Tq##ClassName>(address)};
|
[torque] Use generated instance types, part 1
This change begins making use of the fact that Torque now knows about
the relationship between classes and instance types, to replace a few
repetitive lists:
- Instance type checkers (single and range), defined in
src/objects/instance-type.h
- Verification dispatch in src/diagnostics/objects-debug.cc
- Printer dispatch in src/diagnostics/objects-printer.cc
- Postmortem object type detection in
tools/debug_helper/get-object-properties.cc
Torque is updated to generate four macro lists for the instance types,
representing all of the classes separated in two dimensions: classes
that correspond to a single instance type versus those that have a
range, and classes that are fully defined in Torque (with fields and
methods inside '{}') versus those that are only declared. The latter
distinction is useful because fully-defined classes are guaranteed to
correspond to real C++ classes, whereas only-declared classes are not.
A few other changes were required to make the lists above work:
- Renamed IsFiller to IsFreeSpaceOrFiller to better reflect what it does
and avoid conflicts with the new macro-generated IsFiller method. This
is the part I'm most worried about: I think the new name is an
improvement for clarity and consistency, but I could imagine someone
typing IsFiller out of habit and introducing a bug. If we'd prefer to
keep the name IsFiller, my other idea is to rename FreeSpace to
VariableSizeFiller and Filler to FixedSizeFiller.
- Made Tuple3 extend from Struct, not Tuple2, because IsTuple2 is
expected to check for only TUPLE2_TYPE and not include TUPLE3_TYPE.
- Normalized the dispatched behavior for BigIntBase and HeapNumber.
- Added a few new object printers.
Bug: v8:7793
Change-Id: I5462bb105f8a314baa59bd6ab6ab6215df6f313c
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1860314
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Reviewed-by: Dan Elphick <delphick@chromium.org>
Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Cr-Commit-Position: refs/heads/master@{#64597}
2019-10-28 17:42:41 +00:00
|
|
|
TORQUE_INSTANCE_CHECKERS_SINGLE_FULLY_DEFINED(INSTANCE_TYPE_CASE)
|
2020-03-03 15:53:40 +00:00
|
|
|
TORQUE_INSTANCE_CHECKERS_MULTIPLE_FULLY_DEFINED(INSTANCE_TYPE_CASE)
|
2019-09-19 20:23:39 +00:00
|
|
|
#undef INSTANCE_TYPE_CASE
|
|
|
|
|
|
|
|
default:
|
|
|
|
|
|
|
|
// Special case: concrete subtypes of String are not included in the
|
|
|
|
// main instance type list because they use the low bits of the instance
|
|
|
|
// type enum as flags.
|
|
|
|
if (type <= i::LAST_STRING_TYPE) {
|
|
|
|
return GetTypedObjectForString(address, type, type_source);
|
|
|
|
}
|
|
|
|
|
|
|
|
#define INSTANCE_RANGE_CASE(ClassName, FIRST_TYPE, LAST_TYPE) \
|
|
|
|
if (type >= i::FIRST_TYPE && type <= i::LAST_TYPE) { \
|
|
|
|
return {type_source, std::make_unique<Tq##ClassName>(address)}; \
|
|
|
|
}
|
[torque] Use generated instance types, part 1
This change begins making use of the fact that Torque now knows about
the relationship between classes and instance types, to replace a few
repetitive lists:
- Instance type checkers (single and range), defined in
src/objects/instance-type.h
- Verification dispatch in src/diagnostics/objects-debug.cc
- Printer dispatch in src/diagnostics/objects-printer.cc
- Postmortem object type detection in
tools/debug_helper/get-object-properties.cc
Torque is updated to generate four macro lists for the instance types,
representing all of the classes separated in two dimensions: classes
that correspond to a single instance type versus those that have a
range, and classes that are fully defined in Torque (with fields and
methods inside '{}') versus those that are only declared. The latter
distinction is useful because fully-defined classes are guaranteed to
correspond to real C++ classes, whereas only-declared classes are not.
A few other changes were required to make the lists above work:
- Renamed IsFiller to IsFreeSpaceOrFiller to better reflect what it does
and avoid conflicts with the new macro-generated IsFiller method. This
is the part I'm most worried about: I think the new name is an
improvement for clarity and consistency, but I could imagine someone
typing IsFiller out of habit and introducing a bug. If we'd prefer to
keep the name IsFiller, my other idea is to rename FreeSpace to
VariableSizeFiller and Filler to FixedSizeFiller.
- Made Tuple3 extend from Struct, not Tuple2, because IsTuple2 is
expected to check for only TUPLE2_TYPE and not include TUPLE3_TYPE.
- Normalized the dispatched behavior for BigIntBase and HeapNumber.
- Added a few new object printers.
Bug: v8:7793
Change-Id: I5462bb105f8a314baa59bd6ab6ab6215df6f313c
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1860314
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Reviewed-by: Dan Elphick <delphick@chromium.org>
Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Cr-Commit-Position: refs/heads/master@{#64597}
2019-10-28 17:42:41 +00:00
|
|
|
TORQUE_INSTANCE_CHECKERS_RANGE_FULLY_DEFINED(INSTANCE_RANGE_CASE)
|
2019-09-19 20:23:39 +00:00
|
|
|
#undef INSTANCE_RANGE_CASE
|
|
|
|
|
|
|
|
return {d::TypeCheckResult::kUnknownInstanceType,
|
|
|
|
std::make_unique<TqHeapObject>(address)};
|
|
|
|
}
|
2019-08-21 15:37:47 +00:00
|
|
|
}
|
|
|
|
|
2019-08-26 21:58:03 +00:00
|
|
|
TypedObject GetTypedHeapObject(uintptr_t address, d::MemoryAccessor accessor,
|
2019-09-19 20:23:39 +00:00
|
|
|
const char* type_hint,
|
|
|
|
const d::HeapAddresses& heap_addresses) {
|
2019-09-10 10:12:00 +00:00
|
|
|
auto heap_object = std::make_unique<TqHeapObject>(address);
|
2019-08-26 21:58:03 +00:00
|
|
|
Value<uintptr_t> map_ptr = heap_object->GetMapValue(accessor);
|
|
|
|
|
|
|
|
if (map_ptr.validity != d::MemoryAccessResult::kOk) {
|
2019-09-19 20:23:39 +00:00
|
|
|
// If we can't read the Map pointer from the object, then we likely can't
|
|
|
|
// read anything else, so there's not any point in attempting to use the
|
|
|
|
// type hint. Just return a failure.
|
2019-08-26 21:58:03 +00:00
|
|
|
return {map_ptr.validity == d::MemoryAccessResult::kAddressNotValid
|
|
|
|
? d::TypeCheckResult::kObjectPointerInvalid
|
|
|
|
: d::TypeCheckResult::kObjectPointerValidButInaccessible,
|
|
|
|
std::move(heap_object)};
|
|
|
|
}
|
2019-09-19 20:23:39 +00:00
|
|
|
|
2019-08-26 21:58:03 +00:00
|
|
|
Value<i::InstanceType> type =
|
|
|
|
TqMap(map_ptr.value).GetInstanceTypeValue(accessor);
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
if (type.validity == d::MemoryAccessResult::kOk) {
|
2019-09-19 20:23:39 +00:00
|
|
|
return GetTypedObjectByInstanceType(address, type.value,
|
|
|
|
d::TypeCheckResult::kUsedMap);
|
|
|
|
}
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
|
2019-09-19 20:23:39 +00:00
|
|
|
// We can't read the Map, so check whether it is in the list of known Maps,
|
|
|
|
// as another way to get its instance type.
|
|
|
|
KnownInstanceType known_map_type =
|
2020-01-23 00:06:59 +00:00
|
|
|
FindKnownMapInstanceTypes(map_ptr.value, heap_addresses);
|
2019-09-19 20:23:39 +00:00
|
|
|
if (known_map_type.confidence == KnownInstanceType::Confidence::kHigh) {
|
|
|
|
DCHECK_EQ(known_map_type.types.size(), 1);
|
|
|
|
return GetTypedObjectByInstanceType(address, known_map_type.types[0],
|
|
|
|
d::TypeCheckResult::kKnownMapPointer);
|
|
|
|
}
|
2019-08-21 15:37:47 +00:00
|
|
|
|
2019-09-19 20:23:39 +00:00
|
|
|
// Create a basic result that says that the object is a HeapObject and we
|
|
|
|
// couldn't read its Map.
|
|
|
|
TypedObject result = {
|
|
|
|
type.validity == d::MemoryAccessResult::kAddressNotValid
|
|
|
|
? d::TypeCheckResult::kMapPointerInvalid
|
|
|
|
: d::TypeCheckResult::kMapPointerValidButInaccessible,
|
|
|
|
std::move(heap_object)};
|
|
|
|
|
|
|
|
// If a type hint is available, it may give us something more specific than
|
|
|
|
// HeapObject. However, a type hint of Object would be even less specific, so
|
|
|
|
// we'll only use the type hint if it's a subclass of HeapObject.
|
|
|
|
if (type_hint != nullptr) {
|
|
|
|
TypedObject hint_result = GetTypedObjectByHint(address, type_hint);
|
|
|
|
if (result.object->IsSuperclassOf(hint_result.object.get())) {
|
|
|
|
result = std::move(hint_result);
|
|
|
|
}
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
}
|
|
|
|
|
2019-09-19 20:23:39 +00:00
|
|
|
// If low-confidence results are available from known Maps, include them only
|
|
|
|
// if they don't contradict the primary type and would provide some additional
|
|
|
|
// specificity.
|
|
|
|
for (const i::InstanceType type_guess : known_map_type.types) {
|
|
|
|
TypedObject guess_result = GetTypedObjectByInstanceType(
|
|
|
|
address, type_guess, d::TypeCheckResult::kKnownMapPointer);
|
|
|
|
if (result.object->IsSuperclassOf(guess_result.object.get())) {
|
|
|
|
result.possible_types.push_back(std::move(guess_result));
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
}
|
2019-08-26 21:58:03 +00:00
|
|
|
}
|
2019-09-19 20:23:39 +00:00
|
|
|
|
|
|
|
return result;
|
2019-08-26 21:58:03 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
// An object visitor that accumulates the first few characters of a string.
|
|
|
|
class ReadStringVisitor : public TqObjectVisitor {
|
|
|
|
public:
|
2020-01-23 00:06:59 +00:00
|
|
|
static v8::base::Optional<std::string> Visit(
|
|
|
|
d::MemoryAccessor accessor, const d::HeapAddresses& heap_addresses,
|
|
|
|
const TqString* object) {
|
|
|
|
ReadStringVisitor visitor(accessor, heap_addresses);
|
|
|
|
object->Visit(&visitor);
|
|
|
|
return visitor.GetString();
|
|
|
|
}
|
2019-08-26 21:58:03 +00:00
|
|
|
|
|
|
|
// Returns the result as UTF-8 once visiting is complete.
|
2020-01-23 00:06:59 +00:00
|
|
|
v8::base::Optional<std::string> GetString() {
|
|
|
|
if (failed_) return {};
|
2019-08-26 21:58:03 +00:00
|
|
|
std::vector<char> result(
|
|
|
|
string_.size() * unibrow::Utf16::kMaxExtraUtf8BytesForOneUtf16CodeUnit);
|
|
|
|
unsigned write_index = 0;
|
|
|
|
int prev_char = unibrow::Utf16::kNoPreviousCharacter;
|
|
|
|
for (size_t read_index = 0; read_index < string_.size(); ++read_index) {
|
|
|
|
uint16_t character = string_[read_index];
|
|
|
|
write_index +=
|
|
|
|
unibrow::Utf8::Encode(result.data() + write_index, character,
|
|
|
|
prev_char, /*replace_invalid=*/true);
|
|
|
|
prev_char = character;
|
|
|
|
}
|
2020-01-23 00:06:59 +00:00
|
|
|
return std::string(result.data(), write_index);
|
2019-08-26 21:58:03 +00:00
|
|
|
}
|
|
|
|
|
2020-01-23 00:06:59 +00:00
|
|
|
template <typename TChar>
|
|
|
|
Value<TChar> ReadCharacter(uintptr_t data_address, int32_t index) {
|
|
|
|
TChar value{};
|
|
|
|
d::MemoryAccessResult validity =
|
|
|
|
accessor_(data_address + index * sizeof(TChar),
|
|
|
|
reinterpret_cast<uint8_t*>(&value), sizeof(value));
|
|
|
|
return {validity, value};
|
|
|
|
}
|
|
|
|
|
|
|
|
template <typename TChar>
|
|
|
|
void ReadStringCharacters(const TqString* object, uintptr_t data_address) {
|
2019-08-26 21:58:03 +00:00
|
|
|
int32_t length = GetOrFinish(object->GetLengthValue(accessor_));
|
|
|
|
for (; index_ < length && index_ < limit_ && !done_; ++index_) {
|
2020-01-23 00:06:59 +00:00
|
|
|
STATIC_ASSERT(sizeof(TChar) <= sizeof(char16_t));
|
2019-08-26 21:58:03 +00:00
|
|
|
char16_t c = static_cast<char16_t>(
|
2020-01-23 00:06:59 +00:00
|
|
|
GetOrFinish(ReadCharacter<TChar>(data_address, index_)));
|
2019-08-26 21:58:03 +00:00
|
|
|
if (!done_) AddCharacter(c);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2020-01-23 00:06:59 +00:00
|
|
|
template <typename TChar, typename TString>
|
|
|
|
void ReadSeqString(const TString* object) {
|
|
|
|
ReadStringCharacters<TChar>(object, object->GetCharsAddress());
|
|
|
|
}
|
|
|
|
|
2019-08-26 21:58:03 +00:00
|
|
|
void VisitSeqOneByteString(const TqSeqOneByteString* object) override {
|
2020-01-23 00:06:59 +00:00
|
|
|
ReadSeqString<char>(object);
|
2019-08-26 21:58:03 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
void VisitSeqTwoByteString(const TqSeqTwoByteString* object) override {
|
2020-01-23 00:06:59 +00:00
|
|
|
ReadSeqString<char16_t>(object);
|
2019-08-26 21:58:03 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
void VisitConsString(const TqConsString* object) override {
|
|
|
|
uintptr_t first_address = GetOrFinish(object->GetFirstValue(accessor_));
|
|
|
|
if (done_) return;
|
2019-09-19 20:23:39 +00:00
|
|
|
auto first =
|
|
|
|
GetTypedHeapObject(first_address, accessor_, nullptr, heap_addresses_)
|
|
|
|
.object;
|
2019-08-26 21:58:03 +00:00
|
|
|
first->Visit(this);
|
|
|
|
if (done_) return;
|
|
|
|
int32_t first_length = GetOrFinish(
|
|
|
|
static_cast<TqString*>(first.get())->GetLengthValue(accessor_));
|
|
|
|
uintptr_t second = GetOrFinish(object->GetSecondValue(accessor_));
|
|
|
|
if (done_) return;
|
|
|
|
IndexModifier modifier(this, -first_length, -first_length);
|
2019-09-19 20:23:39 +00:00
|
|
|
GetTypedHeapObject(second, accessor_, nullptr, heap_addresses_)
|
|
|
|
.object->Visit(this);
|
2019-08-26 21:58:03 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
void VisitSlicedString(const TqSlicedString* object) override {
|
|
|
|
uintptr_t parent = GetOrFinish(object->GetParentValue(accessor_));
|
|
|
|
int32_t length = GetOrFinish(object->GetLengthValue(accessor_));
|
|
|
|
int32_t offset = i::PlatformSmiTagging::SmiToInt(
|
|
|
|
GetOrFinish(object->GetOffsetValue(accessor_)));
|
|
|
|
if (done_) return;
|
|
|
|
int32_t limit_adjust = offset + length - limit_;
|
|
|
|
IndexModifier modifier(this, offset, limit_adjust < 0 ? limit_adjust : 0);
|
2019-09-19 20:23:39 +00:00
|
|
|
GetTypedHeapObject(parent, accessor_, nullptr, heap_addresses_)
|
|
|
|
.object->Visit(this);
|
2019-08-26 21:58:03 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
void VisitThinString(const TqThinString* object) override {
|
|
|
|
uintptr_t actual = GetOrFinish(object->GetActualValue(accessor_));
|
|
|
|
if (done_) return;
|
2019-09-19 20:23:39 +00:00
|
|
|
GetTypedHeapObject(actual, accessor_, nullptr, heap_addresses_)
|
|
|
|
.object->Visit(this);
|
2019-08-26 21:58:03 +00:00
|
|
|
}
|
|
|
|
|
2020-01-23 00:06:59 +00:00
|
|
|
bool IsExternalStringCached(const TqExternalString* object) {
|
|
|
|
// The safest way to get the instance type is to use known map pointers, in
|
|
|
|
// case the map data is not available.
|
|
|
|
uintptr_t map = GetOrFinish(object->GetMapValue(accessor_));
|
|
|
|
if (done_) return false;
|
|
|
|
auto instance_types = FindKnownMapInstanceTypes(map, heap_addresses_);
|
|
|
|
// Exactly one of the matched instance types should be a string type,
|
|
|
|
// because all maps for string types are in the same space (read-only
|
|
|
|
// space). The "uncached" flag on that instance type tells us whether it's
|
|
|
|
// safe to read the cached data.
|
|
|
|
for (const auto& type : instance_types.types) {
|
|
|
|
if ((type & i::kIsNotStringMask) == i::kStringTag &&
|
|
|
|
(type & i::kStringRepresentationMask) == i::kExternalStringTag) {
|
|
|
|
return (type & i::kUncachedExternalStringMask) !=
|
|
|
|
i::kUncachedExternalStringTag;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// If for some reason we can't find an external string type here (maybe the
|
|
|
|
// caller provided an external string type as the type hint, but it doesn't
|
|
|
|
// actually match the in-memory map pointer), then we can't safely use the
|
|
|
|
// cached data.
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
template <typename TChar>
|
|
|
|
void ReadExternalString(const TqExternalString* object) {
|
|
|
|
// Cached external strings are easy to read; uncached external strings
|
|
|
|
// require knowledge of the embedder. For now, we only read cached external
|
|
|
|
// strings.
|
|
|
|
if (IsExternalStringCached(object)) {
|
|
|
|
uintptr_t data_address = reinterpret_cast<uintptr_t>(
|
|
|
|
GetOrFinish(object->GetResourceDataValue(accessor_)));
|
|
|
|
if (done_) return;
|
|
|
|
ReadStringCharacters<TChar>(object, data_address);
|
|
|
|
} else {
|
|
|
|
// TODO(v8:9376): Come up with some way that a caller with full knowledge
|
|
|
|
// of a particular embedder could provide a callback function for getting
|
|
|
|
// uncached string data.
|
|
|
|
AddEllipsisAndFinish();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void VisitExternalOneByteString(
|
|
|
|
const TqExternalOneByteString* object) override {
|
|
|
|
ReadExternalString<char>(object);
|
|
|
|
}
|
|
|
|
|
|
|
|
void VisitExternalTwoByteString(
|
|
|
|
const TqExternalTwoByteString* object) override {
|
|
|
|
ReadExternalString<char16_t>(object);
|
2019-08-26 21:58:03 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
void VisitObject(const TqObject* object) override {
|
|
|
|
// If we fail to find a specific type for a sub-object within a cons string,
|
|
|
|
// sliced string, or thin string, we will end up here.
|
|
|
|
AddEllipsisAndFinish();
|
|
|
|
}
|
|
|
|
|
|
|
|
private:
|
2020-01-23 00:06:59 +00:00
|
|
|
ReadStringVisitor(d::MemoryAccessor accessor,
|
|
|
|
const d::HeapAddresses& heap_addresses)
|
|
|
|
: accessor_(accessor),
|
|
|
|
heap_addresses_(heap_addresses),
|
|
|
|
index_(0),
|
|
|
|
limit_(INT32_MAX),
|
|
|
|
done_(false),
|
|
|
|
failed_(false) {}
|
|
|
|
|
2019-08-26 21:58:03 +00:00
|
|
|
// Unpacks a value that was fetched from the debuggee. If the value indicates
|
|
|
|
// that it couldn't successfully fetch memory, then prevents further work.
|
|
|
|
template <typename T>
|
|
|
|
T GetOrFinish(Value<T> value) {
|
|
|
|
if (value.validity != d::MemoryAccessResult::kOk) {
|
|
|
|
AddEllipsisAndFinish();
|
|
|
|
}
|
|
|
|
return value.value;
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
}
|
|
|
|
|
2019-08-26 21:58:03 +00:00
|
|
|
void AddEllipsisAndFinish() {
|
|
|
|
if (!done_) {
|
|
|
|
done_ = true;
|
2020-01-23 00:06:59 +00:00
|
|
|
if (string_.empty()) {
|
|
|
|
failed_ = true;
|
|
|
|
} else {
|
|
|
|
string_ += u"...";
|
|
|
|
}
|
2019-08-26 21:58:03 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void AddCharacter(char16_t c) {
|
|
|
|
if (string_.size() >= kMaxCharacters) {
|
|
|
|
AddEllipsisAndFinish();
|
|
|
|
} else {
|
|
|
|
string_.push_back(c);
|
|
|
|
}
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
}
|
|
|
|
|
2019-08-26 21:58:03 +00:00
|
|
|
// Temporarily adds offsets to both index_ and limit_, to handle ConsString
|
|
|
|
// and SlicedString.
|
|
|
|
class IndexModifier {
|
|
|
|
public:
|
|
|
|
IndexModifier(ReadStringVisitor* that, int32_t index_adjust,
|
|
|
|
int32_t limit_adjust)
|
|
|
|
: that_(that),
|
|
|
|
index_adjust_(index_adjust),
|
|
|
|
limit_adjust_(limit_adjust) {
|
|
|
|
that_->index_ += index_adjust_;
|
|
|
|
that_->limit_ += limit_adjust_;
|
|
|
|
}
|
|
|
|
~IndexModifier() {
|
|
|
|
that_->index_ -= index_adjust_;
|
|
|
|
that_->limit_ -= limit_adjust_;
|
|
|
|
}
|
|
|
|
|
|
|
|
private:
|
|
|
|
ReadStringVisitor* that_;
|
|
|
|
int32_t index_adjust_;
|
|
|
|
int32_t limit_adjust_;
|
|
|
|
DISALLOW_COPY_AND_ASSIGN(IndexModifier);
|
|
|
|
};
|
|
|
|
|
|
|
|
static constexpr int kMaxCharacters = 80; // How many characters to print.
|
|
|
|
|
|
|
|
std::u16string string_; // Result string.
|
|
|
|
d::MemoryAccessor accessor_;
|
2019-09-19 20:23:39 +00:00
|
|
|
const d::HeapAddresses& heap_addresses_;
|
2019-08-26 21:58:03 +00:00
|
|
|
int32_t index_; // Index of next char to read.
|
|
|
|
int32_t limit_; // Don't read past this index (set by SlicedString).
|
|
|
|
bool done_; // Whether to stop further work.
|
2020-01-23 00:06:59 +00:00
|
|
|
bool failed_; // Whether an error was encountered before any valid data.
|
2019-08-26 21:58:03 +00:00
|
|
|
};
|
|
|
|
|
2019-11-22 16:52:09 +00:00
|
|
|
// An object visitor that supplies extra information for some types.
|
2019-08-26 21:58:03 +00:00
|
|
|
class AddInfoVisitor : public TqObjectVisitor {
|
|
|
|
public:
|
2019-11-22 16:52:09 +00:00
|
|
|
// Returns a descriptive string and a list of properties for the given object.
|
|
|
|
// Both may be empty, and are meant as an addition to, not a replacement for,
|
|
|
|
// the Torque-generated data about the object.
|
|
|
|
static std::pair<std::string, std::vector<std::unique_ptr<ObjectProperty>>>
|
|
|
|
Visit(const TqObject* object, d::MemoryAccessor accessor,
|
|
|
|
const d::HeapAddresses& heap_addresses) {
|
|
|
|
AddInfoVisitor visitor(accessor, heap_addresses);
|
|
|
|
object->Visit(&visitor);
|
|
|
|
return {std::move(visitor.brief_), std::move(visitor.properties_)};
|
|
|
|
}
|
2019-08-26 21:58:03 +00:00
|
|
|
|
|
|
|
void VisitString(const TqString* object) override {
|
2020-01-23 00:06:59 +00:00
|
|
|
auto str = ReadStringVisitor::Visit(accessor_, heap_addresses_, object);
|
|
|
|
if (str.has_value()) {
|
|
|
|
brief_ = "\"" + *str + "\"";
|
|
|
|
}
|
2019-11-22 16:52:09 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
void VisitJSObject(const TqJSObject* object) override {
|
|
|
|
// JSObject and its subclasses can be followed directly by an array of
|
|
|
|
// property values. The start and end offsets of those values are described
|
|
|
|
// by a pair of values in its Map.
|
|
|
|
auto map_ptr = object->GetMapValue(accessor_);
|
|
|
|
if (map_ptr.validity != d::MemoryAccessResult::kOk) {
|
|
|
|
return; // Can't read the JSObject. Nothing useful to do.
|
|
|
|
}
|
|
|
|
TqMap map(map_ptr.value);
|
|
|
|
|
|
|
|
// On JSObject instances, this value is the start of in-object properties.
|
|
|
|
// The constructor function index option is only for primitives.
|
|
|
|
auto start_offset =
|
|
|
|
map.GetInObjectPropertiesStartOrConstructorFunctionIndexValue(
|
|
|
|
accessor_);
|
|
|
|
|
|
|
|
// The total size of the object in memory. This may include over-allocated
|
|
|
|
// expansion space that doesn't correspond to any user-accessible property.
|
|
|
|
auto instance_size = map.GetInstanceSizeInWordsValue(accessor_);
|
|
|
|
|
|
|
|
if (start_offset.validity != d::MemoryAccessResult::kOk ||
|
|
|
|
instance_size.validity != d::MemoryAccessResult::kOk) {
|
|
|
|
return; // Can't read the Map. Nothing useful to do.
|
|
|
|
}
|
|
|
|
int num_properties = instance_size.value - start_offset.value;
|
|
|
|
if (num_properties > 0) {
|
|
|
|
properties_.push_back(std::make_unique<ObjectProperty>(
|
|
|
|
"in-object properties", kObjectAsStoredInHeap, kObject,
|
|
|
|
object->GetMapAddress() + start_offset.value * i::kTaggedSize,
|
|
|
|
num_properties, i::kTaggedSize,
|
|
|
|
std::vector<std::unique_ptr<StructProperty>>(),
|
|
|
|
d::PropertyKind::kArrayOfKnownSize));
|
|
|
|
}
|
2019-08-26 21:58:03 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
private:
|
2019-11-22 16:52:09 +00:00
|
|
|
AddInfoVisitor(d::MemoryAccessor accessor,
|
|
|
|
const d::HeapAddresses& heap_addresses)
|
|
|
|
: accessor_(accessor), heap_addresses_(heap_addresses) {}
|
|
|
|
|
|
|
|
// Inputs used by this visitor:
|
|
|
|
|
2019-08-26 21:58:03 +00:00
|
|
|
d::MemoryAccessor accessor_;
|
2019-09-19 20:23:39 +00:00
|
|
|
const d::HeapAddresses& heap_addresses_;
|
2019-11-22 16:52:09 +00:00
|
|
|
|
|
|
|
// Outputs generated by this visitor:
|
|
|
|
|
|
|
|
// A brief description of the object.
|
|
|
|
std::string brief_;
|
|
|
|
// A list of extra properties to append after the automatic ones that are
|
|
|
|
// created for all Torque-defined class fields.
|
|
|
|
std::vector<std::unique_ptr<ObjectProperty>> properties_;
|
2019-08-26 21:58:03 +00:00
|
|
|
};
|
|
|
|
|
2019-09-19 20:23:39 +00:00
|
|
|
std::unique_ptr<ObjectPropertiesResult> GetHeapObjectPropertiesNotCompressed(
|
2019-08-26 21:58:03 +00:00
|
|
|
uintptr_t address, d::MemoryAccessor accessor, const char* type_hint,
|
2019-09-19 20:23:39 +00:00
|
|
|
const d::HeapAddresses& heap_addresses) {
|
|
|
|
// Regardless of whether we can read the object itself, maybe we can find its
|
|
|
|
// pointer in the list of known objects.
|
|
|
|
std::string brief = FindKnownObject(address, heap_addresses);
|
|
|
|
|
|
|
|
TypedObject typed =
|
|
|
|
GetTypedHeapObject(address, accessor, type_hint, heap_addresses);
|
2019-11-22 16:52:09 +00:00
|
|
|
auto props = typed.object->GetProperties(accessor);
|
2019-08-26 21:58:03 +00:00
|
|
|
|
2019-11-22 16:52:09 +00:00
|
|
|
// Use the AddInfoVisitor to get any extra properties or descriptive text that
|
|
|
|
// can't be directly derived from Torque class definitions.
|
|
|
|
auto extra_info =
|
|
|
|
AddInfoVisitor::Visit(typed.object.get(), accessor, heap_addresses);
|
|
|
|
brief = JoinWithSpace(brief, extra_info.first);
|
|
|
|
props.insert(props.end(), std::make_move_iterator(extra_info.second.begin()),
|
|
|
|
std::make_move_iterator(extra_info.second.end()));
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
|
2019-08-26 21:58:03 +00:00
|
|
|
brief = AppendAddressAndType(brief, address, typed.object->GetName());
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
|
2019-09-19 20:23:39 +00:00
|
|
|
// Convert the low-confidence guessed types to a list of strings as expected
|
|
|
|
// for the response.
|
|
|
|
std::vector<std::string> guessed_types;
|
|
|
|
for (const auto& guess : typed.possible_types) {
|
|
|
|
guessed_types.push_back(guess.object->GetName());
|
|
|
|
}
|
|
|
|
|
2019-09-10 10:12:00 +00:00
|
|
|
return std::make_unique<ObjectPropertiesResult>(
|
2019-11-22 16:52:09 +00:00
|
|
|
typed.type_check_result, brief, typed.object->GetName(), std::move(props),
|
|
|
|
std::move(guessed_types));
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
}
|
|
|
|
|
2019-09-19 20:23:39 +00:00
|
|
|
std::unique_ptr<ObjectPropertiesResult> GetHeapObjectPropertiesMaybeCompressed(
|
2019-09-05 18:50:33 +00:00
|
|
|
uintptr_t address, d::MemoryAccessor memory_accessor,
|
|
|
|
d::HeapAddresses heap_addresses, const char* type_hint) {
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
// Try to figure out the heap range, for pointer compression (this is unused
|
|
|
|
// if pointer compression is disabled).
|
|
|
|
uintptr_t any_uncompressed_ptr = 0;
|
|
|
|
if (!IsPointerCompressed(address)) any_uncompressed_ptr = address;
|
2019-09-05 18:50:33 +00:00
|
|
|
if (any_uncompressed_ptr == 0)
|
|
|
|
any_uncompressed_ptr = heap_addresses.any_heap_pointer;
|
|
|
|
if (any_uncompressed_ptr == 0)
|
|
|
|
any_uncompressed_ptr = heap_addresses.map_space_first_page;
|
|
|
|
if (any_uncompressed_ptr == 0)
|
|
|
|
any_uncompressed_ptr = heap_addresses.old_space_first_page;
|
|
|
|
if (any_uncompressed_ptr == 0)
|
|
|
|
any_uncompressed_ptr = heap_addresses.read_only_space_first_page;
|
|
|
|
FillInUnknownHeapAddresses(&heap_addresses, any_uncompressed_ptr);
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
if (any_uncompressed_ptr == 0) {
|
|
|
|
// We can't figure out the heap range. Just check for known objects.
|
2019-09-05 18:50:33 +00:00
|
|
|
std::string brief = FindKnownObject(address, heap_addresses);
|
2019-11-22 16:52:09 +00:00
|
|
|
brief = AppendAddressAndType(brief, address, kTaggedValue);
|
2019-09-10 10:12:00 +00:00
|
|
|
return std::make_unique<ObjectPropertiesResult>(
|
2019-11-22 16:52:09 +00:00
|
|
|
d::TypeCheckResult::kUnableToDecompress, brief, kTaggedValue);
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
}
|
|
|
|
|
2019-09-19 20:23:39 +00:00
|
|
|
address = EnsureDecompressed(address, any_uncompressed_ptr);
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
|
2019-09-19 20:23:39 +00:00
|
|
|
return GetHeapObjectPropertiesNotCompressed(address, memory_accessor,
|
|
|
|
type_hint, heap_addresses);
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
}
|
|
|
|
|
2019-09-19 20:23:39 +00:00
|
|
|
std::unique_ptr<ObjectPropertiesResult> GetObjectProperties(
|
2019-09-05 18:50:33 +00:00
|
|
|
uintptr_t address, d::MemoryAccessor memory_accessor,
|
|
|
|
const d::HeapAddresses& heap_addresses, const char* type_hint) {
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
if (static_cast<uint32_t>(address) == i::kClearedWeakHeapObjectLower32) {
|
2019-09-10 10:12:00 +00:00
|
|
|
return std::make_unique<ObjectPropertiesResult>(
|
2019-11-22 16:52:09 +00:00
|
|
|
d::TypeCheckResult::kWeakRef, "cleared weak ref", kHeapObject);
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
}
|
|
|
|
bool is_weak = (address & i::kHeapObjectTagMask) == i::kWeakHeapObjectTag;
|
|
|
|
if (is_weak) {
|
|
|
|
address &= ~i::kWeakHeapObjectMask;
|
|
|
|
}
|
|
|
|
if (i::Internals::HasHeapObjectTag(address)) {
|
2019-09-19 20:23:39 +00:00
|
|
|
std::unique_ptr<ObjectPropertiesResult> result =
|
|
|
|
GetHeapObjectPropertiesMaybeCompressed(address, memory_accessor,
|
|
|
|
heap_addresses, type_hint);
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
if (is_weak) {
|
|
|
|
result->Prepend("weak ref to ");
|
|
|
|
}
|
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
|
|
|
// For smi values, construct a response with a description representing the
|
|
|
|
// untagged value.
|
|
|
|
int32_t value = i::PlatformSmiTagging::SmiToInt(address);
|
|
|
|
std::stringstream stream;
|
|
|
|
stream << value << " (0x" << std::hex << value << ")";
|
2019-11-22 16:52:09 +00:00
|
|
|
return std::make_unique<ObjectPropertiesResult>(d::TypeCheckResult::kSmi,
|
|
|
|
stream.str(), kSmi);
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
}
|
|
|
|
|
2020-02-27 18:14:53 +00:00
|
|
|
} // namespace debug_helper_internal
|
|
|
|
} // namespace internal
|
|
|
|
} // namespace v8
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
|
2020-02-27 18:14:53 +00:00
|
|
|
namespace di = v8::internal::debug_helper_internal;
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
|
|
|
|
extern "C" {
|
|
|
|
V8_DEBUG_HELPER_EXPORT d::ObjectPropertiesResult*
|
|
|
|
_v8_debug_helper_GetObjectProperties(uintptr_t object,
|
|
|
|
d::MemoryAccessor memory_accessor,
|
2019-09-05 18:50:33 +00:00
|
|
|
const d::HeapAddresses& heap_addresses,
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
const char* type_hint) {
|
2019-09-19 20:23:39 +00:00
|
|
|
return di::GetObjectProperties(object, memory_accessor, heap_addresses,
|
|
|
|
type_hint)
|
Reland "Add postmortem debugging helper library"
This is a reland of 517ab73fd7e3fdb70220b9699bca4c69a32e212e
Updates since original: now compressed pointers passed to the function
GetObjectProperties are required to be sign-extended. Previously, the
function allowed zero-extended values, but that led to ambiguity on
pointers like 0x88044919: is it compressed or is the heap range actually
centered on 0x100000000?
Original change's description:
> Add postmortem debugging helper library
>
> This change begins to implement the functionality described in
> https://docs.google.com/document/d/1evHnb1uLlSbvHAAsmOXyc25x3uh1DjgNa8u1RHvwVhk/edit#
> for investigating V8 state in crash dumps.
>
> This change adds a new library, v8_debug_helper, for providing platform-
> agnostic assistance with postmortem debugging. This library can be used
> by extensions built for debuggers such as WinDbg or lldb. Its public API
> is described by debug-helper.h; currently the only method it exposes is
> GetObjectProperties, but we'd like to add more functionality over time.
> The API surface is restricted to plain C-style structs and pointers, so
> that it's easy to link from a debugger extension built with a different
> toolchain.
>
> This change also adds a new cctest file to exercise some basic
> interaction with the new library.
>
> The API function GetObjectProperties takes an object pointer (which
> could be compressed, or weak, or a SMI), and returns a string
> description of the object and a list of properties the object contains.
> For now, the list of properties is entirely based on Torque object
> definitions, but we expect to add custom properties in future updates so
> that it can be easier to make sense of complex data structures such as
> dictionaries.
>
> GetObjectProperties does several things that are intended to generate
> somewhat useful results even in cases where memory may be corrupt or
> unavailable:
> - The caller may optionally provide a type string which will be used if
> the memory for the object's Map is inaccessible.
> - All object pointers are compared against the list of known objects
> generated by mkgrokdump. The caller may optionally provide the
> pointers for the first pages of various heap spaces, to avoid spurious
> matches. If those pointers are not provided, then any matches are
> prefixed with "maybe" in the resulting description string, such as
> "maybe UndefinedValue (0x4288000341 <Oddball>)".
>
> Bug: v8:9376
>
> Change-Id: Iebf3cc2dea3133c7811bcefcdf38d9458b02fded
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1628012
> Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
> Reviewed-by: Yang Guo <yangguo@chromium.org>
> Reviewed-by: Michael Stanton <mvstanton@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#62882}
Bug: v8:9376
Change-Id: I866a1cc9d4c34bfe10c7b98462451fe69763cf3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1717090
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/master@{#63008}
2019-07-30 14:38:15 +00:00
|
|
|
.release()
|
|
|
|
->GetPublicView();
|
|
|
|
}
|
|
|
|
V8_DEBUG_HELPER_EXPORT void _v8_debug_helper_Free_ObjectPropertiesResult(
|
|
|
|
d::ObjectPropertiesResult* result) {
|
|
|
|
std::unique_ptr<di::ObjectPropertiesResult> ptr(
|
|
|
|
static_cast<di::ObjectPropertiesResultExtended*>(result)->base);
|
|
|
|
}
|
|
|
|
}
|