v8/tools/debug_helper/get-object-properties.cc

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

800 lines
32 KiB
C++
Raw Normal View History

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/external-pointer.h"
#include "src/execution/frame-constants.h"
#include "src/execution/frames.h"
#include "src/execution/isolate-utils.h"
#include "src/objects/string-inl.h"
#include "src/strings/unicode-inl.h"
Reland "[torque] refactor: use -tq only in filenames derived from .tq files" This is a reland of 64caf2b0b26cb6c955fbbb81365ac54820de51a0 Original change's description: > [torque] refactor: use -tq only in filenames derived from .tq files > > This is to establish a naming rule for Torque-generated files: > - If the file is called foo/bar-tq..., then it is derived from a > file foo/bar.tq > - Otherwise it doesn't belong to a specific .tq file. > > So far, we attached -tq to all Torque-generated file names, where it > sometimes corresponded to a .tq file name and sometimes not. > It is not necessary to add -tq to file names to indicate that they are > Torque-generated, since they are already in a directory called > torque-generated, and we always refer to them as > "torque-generated/filename", so there is no confusion even though some > files now have the same name as a corresponding hand-written file, for > example factory.cc. > > TBR: hpayer@chromium.org > Bug: v8:7793 > Change-Id: Ie172babad1fc7422fd1059c48f5dafaa53e50c8b > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2414218 > Commit-Queue: Tobias Tebbi <tebbi@chromium.org> > Reviewed-by: Jakob Gruber <jgruber@chromium.org> > Cr-Commit-Position: refs/heads/master@{#70060} Bug: v8:7793 TBR: hpayer@chromium.org jgruber@chromium.org Change-Id: I6c492bc64aee1ff167e7ef401825eca9097a7f38 Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2431565 Commit-Queue: Tobias Tebbi <tebbi@chromium.org> Reviewed-by: Tobias Tebbi <tebbi@chromium.org> Cr-Commit-Position: refs/heads/master@{#70137}
2020-09-25 14:11:33 +00:00
#include "torque-generated/class-debug-readers.h"
#include "torque-generated/debug-macros.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
namespace i = v8::internal;
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
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() + ")";
}
std::string JoinWithSpace(const std::string& a, const std::string& b) {
return a.empty() || b.empty() ? a + b : a + " " + b;
}
struct TypedObject {
TypedObject(d::TypeCheckResult type_check_result,
std::unique_ptr<TqObject> object)
: type_check_result(type_check_result), object(std::move(object)) {}
// How we discovered the object's type, or why we failed to do so.
d::TypeCheckResult type_check_result;
// Pointer to some TqObject subclass, representing the most specific known
// type for the object.
std::unique_ptr<TqObject> object;
// Collection of other guesses at more specific types than the one represented
// by |object|.
std::vector<TypedObject> possible_types;
};
TypedObject GetTypedObjectByHint(uintptr_t address,
std::string type_hint_string) {
#define TYPE_NAME_CASE(ClassName, ...) \
if (type_hint_string == "v8::internal::" #ClassName) { \
return {d::TypeCheckResult::kUsedTypeHint, \
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(TYPE_NAME_CASE)
TORQUE_INSTANCE_CHECKERS_RANGE_FULLY_DEFINED(TYPE_NAME_CASE)
STRING_CLASS_TYPES(TYPE_NAME_CASE)
#undef TYPE_NAME_CASE
return {d::TypeCheckResult::kUnknownTypeHint,
std::make_unique<TqHeapObject>(address)};
}
TypedObject GetTypedObjectForString(uintptr_t address, i::InstanceType type,
d::TypeCheckResult type_source) {
class StringGetDispatcher : public i::AllStatic {
public:
#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)}; \
}
STRING_CLASS_TYPES(DEFINE_METHOD)
#undef DEFINE_METHOD
static inline TypedObject HandleInvalidString(
uintptr_t address, d::TypeCheckResult type_source) {
return {d::TypeCheckResult::kUnknownInstanceType,
std::make_unique<TqString>(address)};
}
};
return i::StringShape(type)
.DispatchToSpecificTypeWithoutCast<StringGetDispatcher, TypedObject>(
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)
#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)
#undef INSTANCE_RANGE_CASE
return {d::TypeCheckResult::kUnknownInstanceType,
std::make_unique<TqHeapObject>(address)};
}
}
bool IsTypedHeapObjectInstanceTypeOf(uintptr_t address,
d::MemoryAccessor accessor,
i::InstanceType instance_type) {
auto heap_object = std::make_unique<TqHeapObject>(address);
Value<uintptr_t> map_ptr = heap_object->GetMapValue(accessor);
if (map_ptr.validity == d::MemoryAccessResult::kOk) {
Value<i::InstanceType> type =
TqMap(map_ptr.value).GetInstanceTypeValue(accessor);
if (type.validity == d::MemoryAccessResult::kOk) {
return instance_type == type.value;
}
}
return false;
}
TypedObject GetTypedHeapObject(uintptr_t address, d::MemoryAccessor accessor,
const char* type_hint,
const d::HeapAddresses& heap_addresses) {
auto heap_object = std::make_unique<TqHeapObject>(address);
Value<uintptr_t> map_ptr = heap_object->GetMapValue(accessor);
if (map_ptr.validity != d::MemoryAccessResult::kOk) {
// 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.
return {map_ptr.validity == d::MemoryAccessResult::kAddressNotValid
? d::TypeCheckResult::kObjectPointerInvalid
: d::TypeCheckResult::kObjectPointerValidButInaccessible,
std::move(heap_object)};
}
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) {
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
// 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 =
FindKnownMapInstanceTypes(map_ptr.value, heap_addresses);
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);
}
// 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
}
// 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
}
}
return result;
}
// An object visitor that accumulates the first few characters of a string.
class ReadStringVisitor : public TqObjectVisitor {
public:
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();
}
// Returns the result as UTF-8 once visiting is complete.
v8::base::Optional<std::string> GetString() {
if (failed_) return {};
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;
}
return std::string(result.data(), write_index);
}
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) {
int32_t length = GetOrFinish(object->GetLengthValue(accessor_));
for (; index_ < length && index_ < limit_ && !done_; ++index_) {
STATIC_ASSERT(sizeof(TChar) <= sizeof(char16_t));
char16_t c = static_cast<char16_t>(
GetOrFinish(ReadCharacter<TChar>(data_address, index_)));
if (!done_) AddCharacter(c);
}
}
template <typename TChar, typename TString>
void ReadSeqString(const TString* object) {
ReadStringCharacters<TChar>(object, object->GetCharsAddress());
}
void VisitSeqOneByteString(const TqSeqOneByteString* object) override {
ReadSeqString<char>(object);
}
void VisitSeqTwoByteString(const TqSeqTwoByteString* object) override {
ReadSeqString<char16_t>(object);
}
void VisitConsString(const TqConsString* object) override {
uintptr_t first_address = GetOrFinish(object->GetFirstValue(accessor_));
if (done_) return;
auto first =
GetTypedHeapObject(first_address, accessor_, nullptr, heap_addresses_)
.object;
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);
GetTypedHeapObject(second, accessor_, nullptr, heap_addresses_)
.object->Visit(this);
}
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);
GetTypedHeapObject(parent, accessor_, nullptr, heap_addresses_)
.object->Visit(this);
}
void VisitThinString(const TqThinString* object) override {
uintptr_t actual = GetOrFinish(object->GetActualValue(accessor_));
if (done_) return;
GetTypedHeapObject(actual, accessor_, nullptr, heap_addresses_)
.object->Visit(this);
}
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)) {
ExternalPointer_t resource_data =
GetOrFinish(object->GetResourceDataValue(accessor_));
#ifdef V8_COMPRESS_POINTERS
uintptr_t data_address = static_cast<uintptr_t>(
DecodeExternalPointer(GetPtrComprCageBaseFromOnHeapAddress(
heap_addresses_.any_heap_pointer),
resource_data, kExternalStringResourceDataTag));
#else
[cpu-profiler] Ensure sampled thread has Isolate lock under Windows While the sampler checked if the sampled thread had the Isolate locked (if locks are being used) under Linux, the check was not done under Windows (or Fuchsia) which meant that in a multi-threading application under Windows, thread locking was not checked making it prone to seg faults and the like as the profiler would be using isolate->js_entry_sp to determine the stack to walk but isolate->js_entry_sp is the stack pointer for the thread that currently has the Isolate lock so, if the sampled thread does not have the lock, the sampler woud be iterating over the wrong stack, one that might actually be actively changing on another thread. The fix was to move the lock check into CpuSampler and Ticker (--prof) so all OSes would do the correct check. The basic concept is that on all operating systems a CpuProfiler, and so its corresponding CpuCampler, the profiler is tied to a thread. This is not based on first principles or anything, it's simply the way it works in V8, though it is a useful conceit as it makes visualization and interpretation of profile data much easier. To collect a sample on a thread associated with a profiler the thread must be stopped for obvious reasons -- walking the stack of a running thread is a formula for disaster. The mechanism for stopping a thread is OS-specific and is done in sample.cc. There are currently three basic approaches, one for Linux/Unix variants, one for Windows and one for Fuchsia. The approaches vary as to which thread actually collects the sample -- under Linux the sample is actually collected on the (interrupted) sampled thread whereas under Fuchsia/Windows it's on a separate thread. However, in a multi-threaded environment (where Locker is used), it's not sufficient for the sampled thread to be stopped. Because the stack walk involves looking in the Isolate heap, no other thread can be messing with the heap while the sample is collected. The only ways to ensure this would be to either stop all threads whenever collecting a sample, or to ensure that the thread being sampled holds the Isolate lock so prevents other threads from messing with the heap. While there might be something to be said for the "stop all threads" approach, the current approach in V8 is to only stop the sampled thread so, if in a multi-threaded environment, the profiler must check if the thread being sampled holds the Isolate lock. Since this check must be done, independent of which thread the sample is being collected on (since it varies from OS to OS), the approach is to save the thread id of the thread to be profiled/sampled when the CpuSampler is instantiated (on all OSes it is instantiated on the sampled thread) and then check that thread id against the Isolate lock holder thread id before collecting a sample. If it matches, we know sample.cc has stop the sampled thread, one way or another, and we know that no other thread can mess with the heap (since the stopped thread holds the Isolate lock) so it's safe to walk the stack and collect data from the heap so the sample can be taken. It it doesn't match, we can't safely collect the sample so we don't. Bug: v8:10850 Change-Id: Iba6cabcd3e11a19c261c004103e37e806934dc6f Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2411343 Reviewed-by: Peter Marshall <petermarshall@chromium.org> Commit-Queue: Peter Marshall <petermarshall@chromium.org> Cr-Commit-Position: refs/heads/master@{#69952}
2020-09-16 12:00:30 +00:00
uintptr_t data_address = static_cast<uintptr_t>(resource_data);
#endif // V8_COMPRESS_POINTERS
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);
}
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:
ReadStringVisitor(d::MemoryAccessor accessor,
const d::HeapAddresses& heap_addresses)
: accessor_(accessor),
heap_addresses_(heap_addresses),
index_(0),
limit_(INT32_MAX),
done_(false),
failed_(false) {}
// 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
}
void AddEllipsisAndFinish() {
if (!done_) {
done_ = true;
if (string_.empty()) {
failed_ = true;
} else {
string_ += u"...";
}
}
}
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
}
// 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(const IndexModifier&) = delete;
IndexModifier& operator=(const IndexModifier&) = delete;
~IndexModifier() {
that_->index_ -= index_adjust_;
that_->limit_ -= limit_adjust_;
}
private:
ReadStringVisitor* that_;
int32_t index_adjust_;
int32_t limit_adjust_;
};
static constexpr int kMaxCharacters = 80; // How many characters to print.
std::u16string string_; // Result string.
d::MemoryAccessor accessor_;
const d::HeapAddresses& heap_addresses_;
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.
bool failed_; // Whether an error was encountered before any valid data.
};
// An object visitor that supplies extra information for some types.
class AddInfoVisitor : public TqObjectVisitor {
public:
// Returns a descriptive string and a list of properties for the given object.
// Both may be empty, and are meant as an addition or 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_)};
}
void VisitString(const TqString* object) override {
auto str = ReadStringVisitor::Visit(accessor_, heap_addresses_, object);
if (str.has_value()) {
brief_ = "\"" + *str + "\"";
}
}
void VisitExternalString(const TqExternalString* object) override {
VisitString(object);
// Cast resource field to v8::String::ExternalStringResourceBase* would add
// more info.
properties_.push_back(std::make_unique<ObjectProperty>(
"resource",
CheckTypeName<v8::String::ExternalStringResourceBase*>(
"v8::String::ExternalStringResourceBase*"),
CheckTypeName<v8::String::ExternalStringResourceBase*>(
"v8::String::ExternalStringResourceBase*"),
object->GetResourceAddress(), 1,
sizeof(v8::String::ExternalStringResourceBase*),
std::vector<std::unique_ptr<StructProperty>>(),
d::PropertyKind::kSingle));
}
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));
}
}
private:
AddInfoVisitor(d::MemoryAccessor accessor,
const d::HeapAddresses& heap_addresses)
: accessor_(accessor), heap_addresses_(heap_addresses) {}
// Inputs used by this visitor:
d::MemoryAccessor accessor_;
const d::HeapAddresses& heap_addresses_;
// 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_;
};
std::unique_ptr<ObjectPropertiesResult> GetHeapObjectPropertiesNotCompressed(
uintptr_t address, d::MemoryAccessor accessor, const char* type_hint,
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);
auto props = typed.object->GetProperties(accessor);
// 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);
// Overwrite existing properties if they have the same name.
for (size_t i = 0; i < extra_info.second.size(); i++) {
bool overwrite = false;
for (size_t j = 0; j < props.size(); j++) {
if (strcmp(props[j]->GetPublicView()->name,
extra_info.second[i]->GetPublicView()->name) == 0) {
props[j] = std::move(extra_info.second[i]);
overwrite = true;
break;
}
}
if (overwrite) continue;
props.push_back(std::move(extra_info.second[i]));
}
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
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
// 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());
}
return std::make_unique<ObjectPropertiesResult>(
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
}
std::unique_ptr<ObjectPropertiesResult> GetHeapObjectPropertiesMaybeCompressed(
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;
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.
std::string brief = FindKnownObject(address, heap_addresses);
brief = AppendAddressAndType(brief, address, kTaggedValue);
return std::make_unique<ObjectPropertiesResult>(
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
}
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
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
}
std::unique_ptr<ObjectPropertiesResult> GetObjectProperties(
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) {
return std::make_unique<ObjectPropertiesResult>(
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)) {
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 << ")";
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
}
std::unique_ptr<StackFrameResult> GetStackFrame(
uintptr_t frame_pointer, d::MemoryAccessor memory_accessor) {
// Read the data at frame_pointer + kContextOrFrameTypeOffset.
intptr_t context_or_frame_type = 0;
d::MemoryAccessResult validity = memory_accessor(
frame_pointer + CommonFrameConstants::kContextOrFrameTypeOffset,
reinterpret_cast<void*>(&context_or_frame_type), sizeof(intptr_t));
auto props = std::vector<std::unique_ptr<ObjectProperty>>();
if (validity == d::MemoryAccessResult::kOk) {
// If it is context, not frame marker then add new property
// "currently_executing_function".
if (!StackFrame::IsTypeMarker(context_or_frame_type)) {
props.push_back(std::make_unique<ObjectProperty>(
"currently_executing_jsfunction",
CheckTypeName<v8::internal::JSFunction>("v8::internal::JSFunction"),
CheckTypeName<v8::internal::JSFunction*>("v8::internal::JSFunction"),
frame_pointer + StandardFrameConstants::kFunctionOffset, 1,
sizeof(v8::internal::JSFunction),
std::vector<std::unique_ptr<StructProperty>>(),
d::PropertyKind::kSingle));
// Add more items in the Locals pane representing the JS function name,
// source file name, and line & column numbers within the source file, so
// that the user doesnt need to dig through the shared_function_info to
// find them.
intptr_t js_function_ptr = 0;
validity = memory_accessor(
frame_pointer + StandardFrameConstants::kFunctionOffset,
reinterpret_cast<void*>(&js_function_ptr), sizeof(intptr_t));
if (validity == d::MemoryAccessResult::kOk) {
TqJSFunction js_function(js_function_ptr);
auto shared_function_info_ptr =
js_function.GetSharedFunctionInfoValue(memory_accessor);
if (shared_function_info_ptr.validity == d::MemoryAccessResult::kOk) {
TqSharedFunctionInfo shared_function_info(
shared_function_info_ptr.value);
auto script_or_debug_info_ptr =
shared_function_info.GetScriptOrDebugInfoValue(memory_accessor);
if (script_or_debug_info_ptr.validity == d::MemoryAccessResult::kOk) {
// Make sure script_or_debug_info_ptr is script.
auto address = script_or_debug_info_ptr.value;
if (IsTypedHeapObjectInstanceTypeOf(address, memory_accessor,
i::InstanceType::SCRIPT_TYPE)) {
TqScript script(script_or_debug_info_ptr.value);
props.push_back(std::make_unique<ObjectProperty>(
"script_name", kObjectAsStoredInHeap, kObject,
script.GetNameAddress(), 1, i::kTaggedSize,
std::vector<std::unique_ptr<StructProperty>>(),
d::PropertyKind::kSingle));
props.push_back(std::make_unique<ObjectProperty>(
"script_source", kObjectAsStoredInHeap, kObject,
script.GetSourceAddress(), 1, i::kTaggedSize,
std::vector<std::unique_ptr<StructProperty>>(),
d::PropertyKind::kSingle));
}
}
auto name_or_scope_info_ptr =
shared_function_info.GetNameOrScopeInfoValue(memory_accessor);
if (name_or_scope_info_ptr.validity == d::MemoryAccessResult::kOk) {
auto scope_info_address = name_or_scope_info_ptr.value;
// Make sure name_or_scope_info_ptr is scope info.
if (IsTypedHeapObjectInstanceTypeOf(
scope_info_address, memory_accessor,
i::InstanceType::SCOPE_INFO_TYPE)) {
auto indexed_field_slice_function_variable_info =
TqDebugFieldSliceScopeInfoFunctionVariableInfo(
memory_accessor, scope_info_address);
if (indexed_field_slice_function_variable_info.validity ==
d::MemoryAccessResult::kOk) {
props.push_back(std::make_unique<ObjectProperty>(
"function_name", kObjectAsStoredInHeap, kObject,
scope_info_address - i::kHeapObjectTag +
std::get<1>(
indexed_field_slice_function_variable_info.value),
std::get<2>(
indexed_field_slice_function_variable_info.value),
i::kTaggedSize,
std::vector<std::unique_ptr<StructProperty>>(),
d::PropertyKind::kSingle));
}
std::vector<std::unique_ptr<StructProperty>>
position_info_struct_field_list;
position_info_struct_field_list.push_back(
std::make_unique<StructProperty>(
"start", kObjectAsStoredInHeap, kObject, 0, 0, 0));
position_info_struct_field_list.push_back(
std::make_unique<StructProperty>("end", kObjectAsStoredInHeap,
kObject, 4, 0, 0));
auto indexed_field_slice_position_info =
TqDebugFieldSliceScopeInfoPositionInfo(memory_accessor,
scope_info_address);
if (indexed_field_slice_position_info.validity ==
d::MemoryAccessResult::kOk) {
props.push_back(std::make_unique<ObjectProperty>(
"function_character_offset", "", "",
scope_info_address - i::kHeapObjectTag +
std::get<1>(indexed_field_slice_position_info.value),
std::get<2>(indexed_field_slice_position_info.value),
i::kTaggedSize, std::move(position_info_struct_field_list),
d::PropertyKind::kSingle));
}
}
}
}
}
}
}
return std::make_unique<StackFrameResult>(std::move(props));
}
} // 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
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,
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) {
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);
}
V8_DEBUG_HELPER_EXPORT d::StackFrameResult* _v8_debug_helper_GetStackFrame(
uintptr_t frame_pointer, d::MemoryAccessor memory_accessor) {
return di::GetStackFrame(frame_pointer, memory_accessor)
.release()
->GetPublicView();
}
V8_DEBUG_HELPER_EXPORT void _v8_debug_helper_Free_StackFrameResult(
d::StackFrameResult* result) {
std::unique_ptr<di::StackFrameResult> ptr(
static_cast<di::StackFrameResultExtended*>(result)->base);
}
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
}