gecko-dev/dom/ipc/RefMessageBodyService.h

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

108 строки
3.9 KiB
C
Исходник Обычный вид История

/* -*- 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 mozilla_dom_RefMessageBodyService_h
#define mozilla_dom_RefMessageBodyService_h
#include "mozilla/ErrorResult.h"
#include "mozilla/dom/StructuredCloneHolder.h"
#include "mozilla/Maybe.h"
#include "mozilla/StaticMutex.h"
#include "nsRefPtrHashtable.h"
namespace mozilla {
namespace dom {
/**
* At the time a BroadcastChannel or MessagePort sends messages, we don't know
* which process is going to receive it. Because of this, we need to check if
* the message is able to cross the process boundary.
* If the message contains objects such as SharedArrayBuffers, WASM modules or
* ImageBitmaps, it can be delivered on the current process only.
* Instead of sending the whole message via IPC, we send a unique ID, while the
* message is kept alive by RefMessageBodyService, on the current process using
* a ref-counted RefMessageBody object.
* When the receiver obtains the message ID, it checks if the local
* RefMessageBodyService knows that ID. If yes, the sender and the receiver are
* on the same process and the delivery can be completed; if not, a
* messageerror event has to be dispatched instead.
*
* For MessagePort communication is 1-to-1 and because of this, the
* receiver takes ownership of the message (RefMessageBodyService::Steal()).
* If the receiver port is on a different process, RefMessageBodyService will
* return a nullptr and a messageerror event will be dispatched.
* For BroadcastChannel, the life-time of a message is a bit different than for
* MessagePort. It has a 1-to-many communication and we could have multiple
* receivers. Each one needs to deliver the message without taking full
* ownership of it.
* In order to support this feature, BroadcastChannel needs to call
* RefMessageBodyService::SetMaxCount() to inform how many ports are allowed to
* retrieve the current message, on the current process. Receivers on other
* processes are not kept in consideration because they will not be able to
* retrieve the message from RefMessageBodyService. When the last allowed
* port has called RefMessageBodyService::GetAndCount(), the message is
* released.
*/
class RefMessageBody final {
friend class RefMessageBodyService;
public:
NS_INLINE_DECL_THREADSAFE_REFCOUNTING(RefMessageBody)
RefMessageBody(const nsID& aPortID,
UniquePtr<ipc::StructuredCloneData>&& aCloneData)
: mPortID(aPortID),
mCloneData(std::move(aCloneData)),
mMaxCount(Nothing()),
mCount(0) {}
const nsID& PortID() const { return mPortID; }
ipc::StructuredCloneData* CloneData() { return mCloneData.get(); }
private:
~RefMessageBody() = default;
const nsID mPortID;
UniquePtr<ipc::StructuredCloneData> mCloneData;
// When mCount reaches mMaxCount, this object is released by the service.
Maybe<uint32_t> mMaxCount;
uint32_t mCount;
};
class RefMessageBodyService final {
public:
NS_INLINE_DECL_THREADSAFE_REFCOUNTING(RefMessageBodyService)
static already_AddRefed<RefMessageBodyService> GetOrCreate();
void ForgetPort(const nsID& aPortID);
const nsID Register(already_AddRefed<RefMessageBody> aBody, ErrorResult& aRv);
already_AddRefed<RefMessageBody> Steal(const nsID& aID);
already_AddRefed<RefMessageBody> GetAndCount(const nsID& aID);
void SetMaxCount(const nsID& aID, uint32_t aMaxCount);
private:
RefMessageBodyService();
~RefMessageBodyService();
static RefMessageBodyService* GetOrCreateInternal(
const StaticMutexAutoLock& aProofOfLock);
nsRefPtrHashtable<nsIDHashKey, RefMessageBody> mMessages;
};
} // namespace dom
} // namespace mozilla
#endif // mozilla_dom_RefMessageBodyService_h