summaryrefslogtreecommitdiffstats
path: root/js/public/StructuredClone.h
diff options
context:
space:
mode:
Diffstat (limited to 'js/public/StructuredClone.h')
-rw-r--r--js/public/StructuredClone.h782
1 files changed, 782 insertions, 0 deletions
diff --git a/js/public/StructuredClone.h b/js/public/StructuredClone.h
new file mode 100644
index 0000000000..55c5c91fe1
--- /dev/null
+++ b/js/public/StructuredClone.h
@@ -0,0 +1,782 @@
+/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*-
+ * vim: set ts=8 sts=2 et sw=2 tw=80:
+ * This Source Code Form is subject to the terms of the Mozilla Public
+ * License, v. 2.0. If a copy of the MPL was not distributed with this
+ * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
+
+#ifndef js_StructuredClone_h
+#define js_StructuredClone_h
+
+#include "mozilla/Attributes.h"
+#include "mozilla/BufferList.h"
+#include "mozilla/MemoryReporting.h"
+
+#include <stdint.h>
+#include <utility>
+
+#include "jstypes.h"
+
+#include "js/AllocPolicy.h"
+#include "js/RootingAPI.h"
+#include "js/TypeDecls.h"
+#include "js/Vector.h"
+
+/*
+ * API for safe passing of structured data, HTML 2018 Feb 21 section 2.7.
+ * <https://html.spec.whatwg.org/multipage/structured-data.html>
+ *
+ * This is a serialization scheme for JS values, somewhat like JSON. It
+ * preserves some aspects of JS objects (strings, numbers, own data properties
+ * with string keys, array elements) but not others (methods, getters and
+ * setters, prototype chains). Unlike JSON, structured data:
+ *
+ * - can contain cyclic references.
+ *
+ * - handles Maps, Sets, and some other object types.
+ *
+ * - supports *transferring* objects of certain types from one realm to
+ * another, rather than cloning them.
+ *
+ * - is specified by a living standard, and continues to evolve.
+ *
+ * - is encoded in a nonstandard binary format, and is never exposed to Web
+ * content in its serialized form. It's used internally by the browser to
+ * send data from one thread/realm/domain to another, not across the
+ * network.
+ */
+
+struct JSStructuredCloneReader;
+struct JSStructuredCloneWriter;
+
+/**
+ * The structured-clone serialization format version number.
+ *
+ * When serialized data is stored as bytes, e.g. in your Firefox profile, later
+ * versions of the engine may have to read it. When you upgrade Firefox, we
+ * don't crawl through your whole profile converting all saved data from the
+ * previous version of the serialization format to the latest version. So it is
+ * normal to have data in old formats stored in your profile.
+ *
+ * The JS engine can *write* data only in the current format version.
+ *
+ * It can *read* any data written in the current version, and data written for
+ * DifferentProcess scope in earlier versions.
+ *
+ *
+ * ## When to bump this version number
+ *
+ * When making a change so drastic that the JS engine needs to know whether
+ * it's reading old or new serialized data in order to handle both correctly,
+ * increment this version number. Make sure the engine can still read all
+ * old data written with previous versions.
+ *
+ * If StructuredClone.cpp doesn't contain code that distinguishes between
+ * version 8 and version 9, there should not be a version 9.
+ *
+ * Do not increment for changes that only affect SameProcess encoding.
+ *
+ * Increment only for changes that would otherwise break old serialized data.
+ * Do not increment for new data types. (Rationale: Modulo bugs, older versions
+ * of the JS engine can already correctly throw errors when they encounter new,
+ * unrecognized features. A version number bump does not actually help them.)
+ */
+#define JS_STRUCTURED_CLONE_VERSION 8
+
+namespace JS {
+
+/**
+ * Indicates the "scope of validity" of serialized data.
+ *
+ * Writing plain JS data produces an array of bytes that can be copied and
+ * read in another process or whatever. The serialized data is Plain Old Data.
+ * However, HTML also supports `Transferable` objects, which, when cloned, can
+ * be moved from the source object into the clone, like when you take a
+ * photograph of someone and it steals their soul.
+ * See <https://developer.mozilla.org/en-US/docs/Web/API/Transferable>.
+ * We support cloning and transferring objects of many types.
+ *
+ * For example, when we transfer an ArrayBuffer (within a process), we "detach"
+ * the ArrayBuffer, embed the raw buffer pointer in the serialized data, and
+ * later install it in a new ArrayBuffer in the destination realm. Ownership
+ * of that buffer memory is transferred from the original ArrayBuffer to the
+ * serialized data and then to the clone.
+ *
+ * This only makes sense within a single address space. When we transfer an
+ * ArrayBuffer to another process, the contents of the buffer must be copied
+ * into the serialized data. (The original ArrayBuffer is still detached,
+ * though, for consistency; in some cases the caller shouldn't know or care if
+ * the recipient is in the same process.)
+ *
+ * ArrayBuffers are actually a lucky case; some objects (like MessagePorts)
+ * can't reasonably be stored by value in serialized data -- it's pointers or
+ * nothing.
+ *
+ * So there is a tradeoff between scope of validity -- how far away the
+ * serialized data may be sent and still make sense -- and efficiency or
+ * features. The read and write algorithms therefore take an argument of this
+ * type, allowing the user to control those trade-offs.
+ */
+enum class StructuredCloneScope : uint32_t {
+ /**
+ * The most restrictive scope, with greatest efficiency and features.
+ *
+ * When writing, this means: The caller promises that the serialized data
+ * will **not** be shipped off to a different process or stored in a
+ * database. However, it may be shipped to another thread. It's OK to
+ * produce serialized data that contains pointers to data that is safe to
+ * send across threads, such as array buffers. In Rust terms, the
+ * serialized data will be treated as `Send` but not `Copy`.
+ *
+ * When reading, this means: Accept transferred objects and buffers
+ * (pointers). The caller promises that the serialized data was written
+ * using this API (otherwise, the serialized data may contain bogus
+ * pointers, leading to undefined behavior).
+ *
+ * Starts from 1 because there used to be a SameProcessSameThread enum value
+ * of 0 and these values are encoded into the structured serialization format
+ * as part of the SCTAG_HEADER, and IndexedDB persists the representation to
+ * disk.
+ */
+ SameProcess = 1,
+
+ /**
+ * When writing, this means we're writing for an audience in a different
+ * process. Produce serialized data that can be sent to other processes,
+ * bitwise copied, or even stored as bytes in a database and read by later
+ * versions of Firefox years from now. The HTML5 spec refers to this as
+ * "ForStorage" as in StructuredSerializeForStorage, though we use
+ * DifferentProcess for IPC as well as storage.
+ *
+ * Transferable objects are limited to ArrayBuffers, whose contents are
+ * copied into the serialized data (rather than just writing a pointer).
+ *
+ * When reading, this means: Do not accept pointers.
+ */
+ DifferentProcess,
+
+ /**
+ * Handle a backwards-compatibility case with IndexedDB (bug 1434308): when
+ * reading, this means to treat legacy SameProcess data as if it were
+ * DifferentProcess.
+ *
+ * Do not use this for writing; use DifferentProcess instead.
+ */
+ DifferentProcessForIndexedDB,
+
+ /**
+ * Existing code wants to be able to create an uninitialized
+ * JSStructuredCloneData without knowing the scope, then populate it with
+ * data (at which point the scope *is* known.)
+ */
+ Unassigned,
+
+ /**
+ * This scope is used when the deserialization context is unknown. When
+ * writing, DifferentProcess or SameProcess scope is chosen based on the
+ * nature of the object.
+ */
+ UnknownDestination,
+};
+
+/** Values used to describe the ownership individual Transferables.
+ *
+ * Note that these *can* show up in DifferentProcess clones, since
+ * DifferentProcess ArrayBuffers can be Transferred. In that case, this will
+ * distinguish the specific ownership mechanism: is it a malloc pointer or a
+ * memory mapping? */
+enum TransferableOwnership {
+ /** Transferable data has not been filled in yet. */
+ SCTAG_TMO_UNFILLED = 0,
+
+ /** Structured clone buffer does not yet own the data. */
+ SCTAG_TMO_UNOWNED = 1,
+
+ /** All enum values at least this large are owned by the clone buffer. */
+ SCTAG_TMO_FIRST_OWNED = 2,
+
+ /** Data is a pointer that can be freed. */
+ SCTAG_TMO_ALLOC_DATA = SCTAG_TMO_FIRST_OWNED,
+
+ /** Data is a memory mapped pointer. */
+ SCTAG_TMO_MAPPED_DATA = 3,
+
+ /**
+ * Data is embedding-specific. The engine can free it by calling the
+ * freeTransfer op. */
+ SCTAG_TMO_CUSTOM = 4,
+
+ /**
+ * Same as SCTAG_TMO_CUSTOM, but the embedding can also use
+ * SCTAG_TMO_USER_MIN and greater, up to 2^32-1, to distinguish specific
+ * ownership variants.
+ */
+ SCTAG_TMO_USER_MIN
+};
+
+class CloneDataPolicy {
+ bool allowIntraClusterClonableSharedObjects_;
+ bool allowSharedMemoryObjects_;
+
+ public:
+ // The default is to deny all policy-controlled aspects.
+
+ CloneDataPolicy()
+ : allowIntraClusterClonableSharedObjects_(false),
+ allowSharedMemoryObjects_(false) {}
+
+ // SharedArrayBuffers and WASM modules can only be cloned intra-process
+ // because the shared memory areas are allocated in process-private memory or
+ // because there are security issues of sharing them cross agent clusters.
+ // y default, we don't allow shared-memory and intra-cluster objects. Clients
+ // should therefore enable these 2 clone features when needed.
+
+ void allowIntraClusterClonableSharedObjects() {
+ allowIntraClusterClonableSharedObjects_ = true;
+ }
+ bool areIntraClusterClonableSharedObjectsAllowed() const {
+ return allowIntraClusterClonableSharedObjects_;
+ }
+
+ void allowSharedMemoryObjects() { allowSharedMemoryObjects_ = true; }
+ bool areSharedMemoryObjectsAllowed() const {
+ return allowSharedMemoryObjects_;
+ }
+};
+
+} /* namespace JS */
+
+/**
+ * Read structured data from the reader r. This hook is used to read a value
+ * previously serialized by a call to the WriteStructuredCloneOp hook.
+ *
+ * tag and data are the pair of uint32_t values from the header. The callback
+ * may use the JS_Read* APIs to read any other relevant parts of the object
+ * from the reader r. closure is any value passed to the JS_ReadStructuredClone
+ * function.
+ *
+ * Return the new object on success, or raise an exception and return nullptr on
+ * error.
+ */
+typedef JSObject* (*ReadStructuredCloneOp)(
+ JSContext* cx, JSStructuredCloneReader* r,
+ const JS::CloneDataPolicy& cloneDataPolicy, uint32_t tag, uint32_t data,
+ void* closure);
+
+/**
+ * Structured data serialization hook. The engine can write primitive values,
+ * Objects, Arrays, Dates, RegExps, TypedArrays, ArrayBuffers, Sets, Maps,
+ * and SharedTypedArrays. Any other type of object requires application support.
+ * This callback must first use the JS_WriteUint32Pair API to write an object
+ * header, passing a value greater than JS_SCTAG_USER to the tag parameter.
+ * Then it can use the JS_Write* APIs to write any other relevant parts of
+ * the value v to the writer w. closure is any value passed to the
+ * JS_WriteStructuredClone function.
+ *
+ * Return true on success, false on error. On error, an exception should
+ * normally be set.
+ */
+typedef bool (*WriteStructuredCloneOp)(JSContext* cx,
+ JSStructuredCloneWriter* w,
+ JS::HandleObject obj,
+ bool* sameProcessScopeRequired,
+ void* closure);
+
+/**
+ * This is called when serialization or deserialization encounters an error.
+ * To follow HTML5, the application must throw a DATA_CLONE_ERR DOMException
+ * with error set to one of the JS_SCERR_* values.
+ *
+ * Note that if the .reportError field of the JSStructuredCloneCallbacks is
+ * set (to a function with this signature), then an exception will *not* be
+ * set on the JSContext when an error is encountered. The clone operation
+ * will still be aborted and will return false, however, so it is up to the
+ * embedding to do what it needs to for the error.
+ *
+ * Example: for the DOM, mozilla::dom::StructuredCloneHolder will save away
+ * the error message during its reportError callback. Then when the overall
+ * operation fails, it will clear any exception that might have been set
+ * from other ways to fail and pass the saved error message to
+ * ErrorResult::ThrowDataCloneError().
+ */
+typedef void (*StructuredCloneErrorOp)(JSContext* cx, uint32_t errorid,
+ void* closure, const char* errorMessage);
+
+/**
+ * This is called when JS_ReadStructuredClone receives a transferable object
+ * not known to the engine. If this hook does not exist or returns false, the
+ * JS engine calls the reportError op if set, otherwise it throws a
+ * DATA_CLONE_ERR DOM Exception. This method is called before any other
+ * callback and must return a non-null object in returnObject on success.
+ *
+ * If this readTransfer() hook is called and produces an object, then the
+ * read() hook will *not* be called for the same object, since the main data
+ * will only contain a backreference to the already-read object.
+ */
+typedef bool (*ReadTransferStructuredCloneOp)(
+ JSContext* cx, JSStructuredCloneReader* r, uint32_t tag, void* content,
+ uint64_t extraData, void* closure, JS::MutableHandleObject returnObject);
+
+/**
+ * Called when JS_WriteStructuredClone receives a transferable object not
+ * handled by the engine. If this hook does not exist or returns false, the JS
+ * engine will call the reportError hook or fall back to throwing a
+ * DATA_CLONE_ERR DOM Exception. This method is called before any other
+ * callback.
+ *
+ * tag: indicates what type of transferable this is. Must be greater than
+ * 0xFFFF0201 (value of the internal SCTAG_TRANSFER_MAP_PENDING_ENTRY)
+ *
+ * ownership: see TransferableOwnership, above. Used to communicate any needed
+ * ownership info to the FreeTransferStructuredCloneOp.
+ *
+ * content, extraData: what the ReadTransferStructuredCloneOp will receive
+ */
+typedef bool (*TransferStructuredCloneOp)(JSContext* cx,
+ JS::Handle<JSObject*> obj,
+ void* closure,
+ // Output:
+ uint32_t* tag,
+ JS::TransferableOwnership* ownership,
+ void** content, uint64_t* extraData);
+
+/**
+ * Called when freeing a transferable handled by the embedding. Note that it
+ * should never trigger a garbage collection (and will assert in a
+ * debug build if it does.)
+ *
+ * This callback will be used to release ownership in three situations:
+ *
+ * 1. During serialization: an object is Transferred from, then an error is
+ * encountered later and the incomplete serialization is discarded.
+ *
+ * 2. During deserialization: before an object is Transferred to, an error
+ * is encountered and the incompletely deserialized clone is discarded.
+ *
+ * 3. Serialized data that includes Transferring is never deserialized (eg when
+ * the receiver disappears before reading in the message), and the clone data
+ * is destroyed.
+ *
+ */
+typedef void (*FreeTransferStructuredCloneOp)(
+ uint32_t tag, JS::TransferableOwnership ownership, void* content,
+ uint64_t extraData, void* closure);
+
+/**
+ * Called when the transferring objects are checked. If this function returns
+ * false, the serialization ends throwing a DataCloneError exception.
+ */
+typedef bool (*CanTransferStructuredCloneOp)(JSContext* cx,
+ JS::Handle<JSObject*> obj,
+ bool* sameProcessScopeRequired,
+ void* closure);
+
+/**
+ * Called when a SharedArrayBuffer (including one owned by a Wasm memory object)
+ * has been processed in context `cx` by structured cloning. If `receiving` is
+ * true then the SAB has been received from a channel and a new SAB object has
+ * been created; if false then an existing SAB has been serialized onto a
+ * channel.
+ *
+ * If the callback returns false then the clone operation (read or write) will
+ * signal a failure.
+ */
+typedef bool (*SharedArrayBufferClonedOp)(JSContext* cx, bool receiving,
+ void* closure);
+
+struct JSStructuredCloneCallbacks {
+ ReadStructuredCloneOp read;
+ WriteStructuredCloneOp write;
+ StructuredCloneErrorOp reportError;
+ ReadTransferStructuredCloneOp readTransfer;
+ TransferStructuredCloneOp writeTransfer;
+ FreeTransferStructuredCloneOp freeTransfer;
+ CanTransferStructuredCloneOp canTransfer;
+ SharedArrayBufferClonedOp sabCloned;
+};
+
+enum OwnTransferablePolicy {
+ /**
+ * The buffer owns any Transferables that it might contain, and should
+ * properly release them upon destruction.
+ */
+ OwnsTransferablesIfAny,
+
+ /**
+ * Do not free any Transferables within this buffer when deleting it. This
+ * is used to mark a clone buffer as containing data from another process,
+ * and so it can't legitimately contain pointers. If the buffer claims to
+ * have transferables, it's a bug or an attack. This is also used for
+ * abandon(), where a buffer still contains raw data but the ownership has
+ * been given over to some other entity.
+ */
+ IgnoreTransferablesIfAny,
+
+ /**
+ * A buffer that cannot contain Transferables at all. This usually means
+ * the buffer is empty (not yet filled in, or having been cleared).
+ */
+ NoTransferables
+};
+
+namespace js {
+class SharedArrayRawBuffer;
+
+class SharedArrayRawBufferRefs {
+ public:
+ SharedArrayRawBufferRefs() = default;
+ SharedArrayRawBufferRefs(SharedArrayRawBufferRefs&& other) = default;
+ SharedArrayRawBufferRefs& operator=(SharedArrayRawBufferRefs&& other);
+ ~SharedArrayRawBufferRefs();
+
+ [[nodiscard]] bool acquire(JSContext* cx, SharedArrayRawBuffer* rawbuf);
+ [[nodiscard]] bool acquireAll(JSContext* cx,
+ const SharedArrayRawBufferRefs& that);
+ void takeOwnership(SharedArrayRawBufferRefs&&);
+ void releaseAll();
+
+ private:
+ js::Vector<js::SharedArrayRawBuffer*, 0, js::SystemAllocPolicy> refs_;
+};
+
+template <typename T, typename AllocPolicy>
+struct BufferIterator;
+} // namespace js
+
+/**
+ * JSStructuredCloneData represents structured clone data together with the
+ * information needed to read/write/transfer/free the records within it, in the
+ * form of a set of callbacks.
+ */
+class MOZ_NON_MEMMOVABLE JS_PUBLIC_API JSStructuredCloneData {
+ public:
+ using BufferList = mozilla::BufferList<js::SystemAllocPolicy>;
+ using Iterator = BufferList::IterImpl;
+
+ private:
+ static const size_t kStandardCapacity = 4096;
+
+ BufferList bufList_;
+
+ // The (address space, thread) scope within which this clone is valid. Note
+ // that this must be either set during construction, or start out as
+ // Unassigned and transition once to something else.
+ JS::StructuredCloneScope scope_;
+
+ const JSStructuredCloneCallbacks* callbacks_ = nullptr;
+ void* closure_ = nullptr;
+ OwnTransferablePolicy ownTransferables_ =
+ OwnTransferablePolicy::NoTransferables;
+ js::SharedArrayRawBufferRefs refsHeld_;
+
+ friend struct JSStructuredCloneWriter;
+ friend class JS_PUBLIC_API JSAutoStructuredCloneBuffer;
+ template <typename T, typename AllocPolicy>
+ friend struct js::BufferIterator;
+
+ public:
+ // The constructor must be infallible but SystemAllocPolicy is not, so both
+ // the initial size and initial capacity of the BufferList must be zero.
+ explicit JSStructuredCloneData(JS::StructuredCloneScope scope)
+ : bufList_(0, 0, kStandardCapacity, js::SystemAllocPolicy()),
+ scope_(scope) {}
+
+ // Steal the raw data from a BufferList. In this case, we don't know the
+ // scope and none of the callback info is assigned yet.
+ JSStructuredCloneData(BufferList&& buffers, JS::StructuredCloneScope scope,
+ OwnTransferablePolicy ownership)
+ : bufList_(std::move(buffers)),
+ scope_(scope),
+ ownTransferables_(ownership) {}
+ JSStructuredCloneData(JSStructuredCloneData&& other) = default;
+ JSStructuredCloneData& operator=(JSStructuredCloneData&& other) = default;
+ ~JSStructuredCloneData();
+
+ void setCallbacks(const JSStructuredCloneCallbacks* callbacks, void* closure,
+ OwnTransferablePolicy policy) {
+ callbacks_ = callbacks;
+ closure_ = closure;
+ ownTransferables_ = policy;
+ }
+
+ [[nodiscard]] bool Init(size_t initialCapacity = 0) {
+ return bufList_.Init(0, initialCapacity);
+ }
+
+ JS::StructuredCloneScope scope() const {
+ if (scope_ == JS::StructuredCloneScope::UnknownDestination) {
+ return JS::StructuredCloneScope::DifferentProcess;
+ }
+ return scope_;
+ }
+
+ void sameProcessScopeRequired() {
+ if (scope_ == JS::StructuredCloneScope::UnknownDestination) {
+ scope_ = JS::StructuredCloneScope::SameProcess;
+ }
+ }
+
+ void initScope(JS::StructuredCloneScope newScope) {
+ MOZ_ASSERT(Size() == 0, "initScope() of nonempty JSStructuredCloneData");
+ if (scope() != JS::StructuredCloneScope::Unassigned) {
+ MOZ_ASSERT(scope() == newScope,
+ "Cannot change scope after it has been initialized");
+ }
+ scope_ = newScope;
+ }
+
+ size_t Size() const { return bufList_.Size(); }
+
+ const Iterator Start() const { return bufList_.Iter(); }
+
+ [[nodiscard]] bool Advance(Iterator& iter, size_t distance) const {
+ return iter.AdvanceAcrossSegments(bufList_, distance);
+ }
+
+ [[nodiscard]] bool ReadBytes(Iterator& iter, char* buffer,
+ size_t size) const {
+ return bufList_.ReadBytes(iter, buffer, size);
+ }
+
+ // Append new data to the end of the buffer.
+ [[nodiscard]] bool AppendBytes(const char* data, size_t size) {
+ MOZ_ASSERT(scope() != JS::StructuredCloneScope::Unassigned);
+ return bufList_.WriteBytes(data, size);
+ }
+
+ // Update data stored within the existing buffer. There must be at least
+ // 'size' bytes between the position of 'iter' and the end of the buffer.
+ [[nodiscard]] bool UpdateBytes(Iterator& iter, const char* data,
+ size_t size) const {
+ MOZ_ASSERT(scope() != JS::StructuredCloneScope::Unassigned);
+ while (size > 0) {
+ size_t remaining = iter.RemainingInSegment();
+ size_t nbytes = std::min(remaining, size);
+ memcpy(iter.Data(), data, nbytes);
+ data += nbytes;
+ size -= nbytes;
+ iter.Advance(bufList_, nbytes);
+ }
+ return true;
+ }
+
+ char* AllocateBytes(size_t maxSize, size_t* size) {
+ return bufList_.AllocateBytes(maxSize, size);
+ }
+
+ void Clear() {
+ discardTransferables();
+ bufList_.Clear();
+ }
+
+ // Return a new read-only JSStructuredCloneData that "borrows" the contents
+ // of |this|. Its lifetime should not exceed the donor's. This is only
+ // allowed for DifferentProcess clones, so finalization of the borrowing
+ // clone will do nothing.
+ JSStructuredCloneData Borrow(Iterator& iter, size_t size,
+ bool* success) const {
+ MOZ_ASSERT(scope() == JS::StructuredCloneScope::DifferentProcess);
+ return JSStructuredCloneData(
+ bufList_.Borrow<js::SystemAllocPolicy>(iter, size, success), scope(),
+ IgnoreTransferablesIfAny);
+ }
+
+ // Iterate over all contained data, one BufferList segment's worth at a
+ // time, and invoke the given FunctionToApply with the data pointer and
+ // size. The function should return a bool value, and this loop will exit
+ // with false if the function ever returns false.
+ template <typename FunctionToApply>
+ bool ForEachDataChunk(FunctionToApply&& function) const {
+ Iterator iter = bufList_.Iter();
+ while (!iter.Done()) {
+ if (!function(iter.Data(), iter.RemainingInSegment())) {
+ return false;
+ }
+ iter.Advance(bufList_, iter.RemainingInSegment());
+ }
+ return true;
+ }
+
+ // Append the entire contents of other's bufList_ to our own.
+ [[nodiscard]] bool Append(const JSStructuredCloneData& other) {
+ MOZ_ASSERT(scope() == other.scope());
+ return other.ForEachDataChunk(
+ [&](const char* data, size_t size) { return AppendBytes(data, size); });
+ }
+
+ size_t SizeOfExcludingThis(mozilla::MallocSizeOf mallocSizeOf) {
+ return bufList_.SizeOfExcludingThis(mallocSizeOf);
+ }
+
+ void discardTransferables();
+
+ private:
+ // This internal method exposes the real value of scope_. It's meant to be
+ // used only when starting the writing.
+ JS::StructuredCloneScope scopeForInternalWriting() const { return scope_; }
+};
+
+/**
+ * Implements StructuredDeserialize and StructuredDeserializeWithTransfer.
+ *
+ * Note: If `data` contains transferable objects, it can be read only once.
+ */
+JS_PUBLIC_API bool JS_ReadStructuredClone(
+ JSContext* cx, const JSStructuredCloneData& data, uint32_t version,
+ JS::StructuredCloneScope scope, JS::MutableHandleValue vp,
+ const JS::CloneDataPolicy& cloneDataPolicy,
+ const JSStructuredCloneCallbacks* optionalCallbacks, void* closure);
+
+/**
+ * Implements StructuredSerialize, StructuredSerializeForStorage, and
+ * StructuredSerializeWithTransfer.
+ *
+ * Note: If the scope is DifferentProcess then the cloneDataPolicy must deny
+ * shared-memory objects, or an error will be signaled if a shared memory object
+ * is seen.
+ */
+JS_PUBLIC_API bool JS_WriteStructuredClone(
+ JSContext* cx, JS::HandleValue v, JSStructuredCloneData* data,
+ JS::StructuredCloneScope scope, const JS::CloneDataPolicy& cloneDataPolicy,
+ const JSStructuredCloneCallbacks* optionalCallbacks, void* closure,
+ JS::HandleValue transferable);
+
+JS_PUBLIC_API bool JS_StructuredCloneHasTransferables(
+ JSStructuredCloneData& data, bool* hasTransferable);
+
+JS_PUBLIC_API bool JS_StructuredClone(
+ JSContext* cx, JS::HandleValue v, JS::MutableHandleValue vp,
+ const JSStructuredCloneCallbacks* optionalCallbacks, void* closure);
+
+/**
+ * The C-style API calls to read and write structured clones are fragile --
+ * they rely on the caller to properly handle ownership of the clone data, and
+ * the handling of the input data as well as the interpretation of the contents
+ * of the clone buffer are dependent on the callbacks passed in. If you
+ * serialize and deserialize with different callbacks, the results are
+ * questionable.
+ *
+ * JSAutoStructuredCloneBuffer wraps things up in an RAII class for data
+ * management, and uses the same callbacks for both writing and reading
+ * (serializing and deserializing).
+ */
+class JS_PUBLIC_API JSAutoStructuredCloneBuffer {
+ JSStructuredCloneData data_;
+ uint32_t version_;
+
+ public:
+ JSAutoStructuredCloneBuffer(JS::StructuredCloneScope scope,
+ const JSStructuredCloneCallbacks* callbacks,
+ void* closure)
+ : data_(scope), version_(JS_STRUCTURED_CLONE_VERSION) {
+ data_.setCallbacks(callbacks, closure,
+ OwnTransferablePolicy::NoTransferables);
+ }
+
+ JSAutoStructuredCloneBuffer(JSAutoStructuredCloneBuffer&& other);
+ JSAutoStructuredCloneBuffer& operator=(JSAutoStructuredCloneBuffer&& other);
+
+ ~JSAutoStructuredCloneBuffer() { clear(); }
+
+ JSStructuredCloneData& data() { return data_; }
+ bool empty() const { return !data_.Size(); }
+
+ void clear();
+
+ JS::StructuredCloneScope scope() const { return data_.scope(); }
+
+ /**
+ * Adopt some memory. It will be automatically freed by the destructor.
+ * data must have been allocated by the JS engine (e.g., extracted via
+ * JSAutoStructuredCloneBuffer::steal).
+ */
+ void adopt(JSStructuredCloneData&& data,
+ uint32_t version = JS_STRUCTURED_CLONE_VERSION,
+ const JSStructuredCloneCallbacks* callbacks = nullptr,
+ void* closure = nullptr);
+
+ /**
+ * Release ownership of the buffer and assign it and ownership of it to
+ * `data`.
+ */
+ void giveTo(JSStructuredCloneData* data);
+
+ bool read(JSContext* cx, JS::MutableHandleValue vp,
+ const JS::CloneDataPolicy& cloneDataPolicy = JS::CloneDataPolicy(),
+ const JSStructuredCloneCallbacks* optionalCallbacks = nullptr,
+ void* closure = nullptr);
+
+ bool write(JSContext* cx, JS::HandleValue v,
+ const JSStructuredCloneCallbacks* optionalCallbacks = nullptr,
+ void* closure = nullptr);
+
+ bool write(JSContext* cx, JS::HandleValue v, JS::HandleValue transferable,
+ const JS::CloneDataPolicy& cloneDataPolicy,
+ const JSStructuredCloneCallbacks* optionalCallbacks = nullptr,
+ void* closure = nullptr);
+
+ size_t sizeOfExcludingThis(mozilla::MallocSizeOf mallocSizeOf) {
+ return data_.SizeOfExcludingThis(mallocSizeOf);
+ }
+
+ size_t sizeOfIncludingThis(mozilla::MallocSizeOf mallocSizeOf) {
+ return mallocSizeOf(this) + sizeOfExcludingThis(mallocSizeOf);
+ }
+
+ private:
+ // Copy and assignment are not supported.
+ JSAutoStructuredCloneBuffer(const JSAutoStructuredCloneBuffer& other) =
+ delete;
+ JSAutoStructuredCloneBuffer& operator=(
+ const JSAutoStructuredCloneBuffer& other) = delete;
+};
+
+// The range of tag values the application may use for its own custom object
+// types.
+#define JS_SCTAG_USER_MIN ((uint32_t)0xFFFF8000)
+#define JS_SCTAG_USER_MAX ((uint32_t)0xFFFFFFFF)
+
+#define JS_SCERR_RECURSION 0
+#define JS_SCERR_TRANSFERABLE 1
+#define JS_SCERR_DUP_TRANSFERABLE 2
+#define JS_SCERR_UNSUPPORTED_TYPE 3
+#define JS_SCERR_SHMEM_TRANSFERABLE 4
+#define JS_SCERR_TYPED_ARRAY_DETACHED 5
+#define JS_SCERR_WASM_NO_TRANSFER 6
+#define JS_SCERR_NOT_CLONABLE 7
+#define JS_SCERR_NOT_CLONABLE_WITH_COOP_COEP 8
+
+JS_PUBLIC_API bool JS_ReadUint32Pair(JSStructuredCloneReader* r, uint32_t* p1,
+ uint32_t* p2);
+
+JS_PUBLIC_API bool JS_ReadBytes(JSStructuredCloneReader* r, void* p,
+ size_t len);
+
+JS_PUBLIC_API bool JS_ReadString(JSStructuredCloneReader* r,
+ JS::MutableHandleString str);
+
+JS_PUBLIC_API bool JS_ReadDouble(JSStructuredCloneReader* r, double* v);
+
+JS_PUBLIC_API bool JS_ReadTypedArray(JSStructuredCloneReader* r,
+ JS::MutableHandleValue vp);
+
+JS_PUBLIC_API bool JS_WriteUint32Pair(JSStructuredCloneWriter* w, uint32_t tag,
+ uint32_t data);
+
+JS_PUBLIC_API bool JS_WriteBytes(JSStructuredCloneWriter* w, const void* p,
+ size_t len);
+
+JS_PUBLIC_API bool JS_WriteString(JSStructuredCloneWriter* w,
+ JS::HandleString str);
+
+JS_PUBLIC_API bool JS_WriteDouble(JSStructuredCloneWriter* w, double v);
+
+JS_PUBLIC_API bool JS_WriteTypedArray(JSStructuredCloneWriter* w,
+ JS::HandleValue v);
+
+JS_PUBLIC_API bool JS_ObjectNotWritten(JSStructuredCloneWriter* w,
+ JS::HandleObject obj);
+
+JS_PUBLIC_API JS::StructuredCloneScope JS_GetStructuredCloneScope(
+ JSStructuredCloneWriter* w);
+
+#endif /* js_StructuredClone_h */