gecko-dev/dom/base/nsIMessageManager.idl

542 lines
20 KiB
Plaintext

/* -*- Mode: IDL; tab-width: 2; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* 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/. */
#include "nsISupports.idl"
interface mozIDOMWindowProxy;
interface nsIDocShell;
interface nsIContent;
interface nsIFrameLoader;
interface nsIPrincipal;
/**
* Message managers provide a way for chrome-privileged JS code to
* communicate with each other, even across process boundaries.
*
* Message managers are separated into "parent side" and "child side".
* These don't always correspond to process boundaries, but can. For
* each child-side message manager, there is always exactly one
* corresponding parent-side message manager that it sends messages
* to. However, for each parent-side message manager, there may be
* either one or many child-side managers it can message.
*
* Message managers that always have exactly one "other side" are of
* type nsIMessageSender. Parent-side message managers that have many
* "other sides" are of type nsIMessageBroadcaster.
*
* Child-side message managers can send synchronous messages to their
* parent side, but not the other way around.
*
* There are two realms of message manager hierarchies. One realm
* approximately corresponds to DOM elements, the other corresponds to
* process boundaries.
*
* Message managers corresponding to DOM elements
* ==============================================
*
* In this realm of message managers, there are
* - "frame message managers" which correspond to frame elements
* - "window message managers" which correspond to top-level chrome
* windows
* - "group message managers" which correspond to named message
* managers with a specific window MM as the parent
* - the "global message manager", on the parent side. See below.
*
* The DOM-realm message managers can communicate in the ways shown by
* the following diagram. The parent side and child side can
* correspond to process boundaries, but don't always.
*
* Parent side Child side
* ------------- ------------
* global MMg
* |
* +-->window MMw1
* | |
* | +-->frame MMp1_1<------------>frame MMc1_1
* | |
* | +-->frame MMp1_2<------------>frame MMc1_2
* | |
* | +-->group MMgr1
* | | |
* | | +-->frame MMp2_1<------->frame MMc2_1
* | | |
* | | +-->frame MMp2_2<------->frame MMc2_2
* | |
* | +-->group MMgr2
* | | ...
* | |
* | ...
* |
* +-->window MMw2
* ...
*
* For example: a message sent from MMc1_1, from the child side, is
* sent only to MMp1_1 on the parent side. However, note that all
* message managers in the hierarchy above MMp1_1, in this diagram
* MMw1 and MMg, will also notify their message listeners when the
* message arrives.
*
* A message sent from MMc2_1 will be sent to MMp2_1 and also notify
* all message managers in the hierarchy above that, including the
* group message manager MMgr1.
* For example: a message broadcast through the global MMg on the
* parent side would be broadcast to MMw1, which would transitively
* broadcast it to MMp1_1, MM1p_2. The message would next be
* broadcast to MMgr1, which would broadcast it to MMp2_1 and MMp2_2.
* After that it would broadcast to MMgr2 and then to MMw2, and so
* on down the hierarchy.
*
* ***** PERFORMANCE AND SECURITY WARNING *****
* Messages broadcast through the global MM and window or group MMs
* can result in messages being dispatched across many OS processes,
* and to many processes with different permissions. Great care
* should be taken when broadcasting.
*
* Interfaces
* ----------
*
* The global MMg and window MMw's are message broadcasters implementing
* nsIMessageBroadcaster while the frame MMp's are simple message senders
* (nsIMessageSender). Their counterparts in the content processes are
* message senders implementing nsIContentFrameMessageManager.
*
* nsIMessageListenerManager
* / \
* nsIMessageSender nsIMessageBroadcaster
* |
* nsISyncMessageSender (content process/in-process only)
* |
* nsIContentFrameMessageManager (content process/in-process only)
* |
* nsIInProcessContentFrameMessageManager (in-process only)
*
*
* Message managers in the chrome process can also be QI'ed to nsIFrameScriptLoader.
*
*
* Message managers corresponding to process boundaries
* ====================================================
*
* The second realm of message managers is the "process message
* managers". With one exception, these always correspond to process
* boundaries. The picture looks like
*
* Parent process Child processes
* ---------------- -----------------
* global (GPPMM)
* |
* +-->parent in-process PIPMM<-->child in-process CIPPMM
* |
* +-->parent (PPMM1)<------------------>child (CPMM1)
* |
* +-->parent (PPMM2)<------------------>child (CPMM2)
* ...
*
* Note, PIPMM and CIPPMM both run in the parent process.
*
* For example: the parent-process PPMM1 sends messages to the
* child-process CPMM1.
*
* For example: CPMM1 sends messages directly to PPMM1. The global GPPMM
* will also notify their message listeners when the message arrives.
*
* For example: messages sent through the global GPPMM will be
* dispatched to the listeners of the same-process, CIPPMM, CPMM1,
* CPMM2, etc.
*
* ***** PERFORMANCE AND SECURITY WARNING *****
* Messages broadcast through the GPPMM can result in messages
* being dispatched across many OS processes, and to many processes
* with different permissions. Great care should be taken when
* broadcasting.
*
* Requests sent to parent-process message listeners should usually
* have replies scoped to the requesting CPMM. The following pattern
* is common
*
* const ParentProcessListener = {
* receiveMessage: function(aMessage) {
* let childMM = aMessage.target.QueryInterface(Ci.nsIMessageSender);
* switch (aMessage.name) {
* case "Foo:Request":
* // service request
* childMM.sendAsyncMessage("Foo:Response", { data });
* }
* }
* };
*/
[scriptable, function, uuid(2b44eb57-a9c6-4773-9a1e-fe0818739a4c)]
interface nsIMessageListener : nsISupports
{
/**
* This is for JS only.
* receiveMessage is called with one parameter, which has the following
* properties:
* {
* target: %the target of the message. Either an element owning
* the message manager, or message manager itself if no
* element owns it%
* name: %message name%,
* sync: %true or false%.
* data: %structured clone of the sent message data%,
* json: %same as .data, deprecated%,
* objects: %named table of jsvals/objects, or null%
* principal: %principal for the window app
* }
*
* Each listener is invoked with its own copy of the message
* parameter.
*
* When the listener is called, 'this' value is the target of the message.
*
* If the message is synchronous, the possible return value is
* returned as JSON (will be changed to use structured clones).
* When there are multiple listeners to sync messages, each
* listener's return value is sent back as an array. |undefined|
* return values show up as undefined values in the array.
*/
void receiveMessage();
};
[scriptable, builtinclass, uuid(b949bfec-bb7d-47bc-b387-ac6a9b655072)]
interface nsIMessageListenerManager : nsISupports
{
/**
* Register |listener| to receive |messageName|. All listener
* callbacks for a particular message are invoked when that message
* is received.
*
* The message manager holds a strong ref to |listener|.
*
* If the same listener registers twice for the same message, the
* second registration is ignored.
*
* Pass true for listenWhenClosed if you want to receive messages
* during the short period after a frame has been removed from the
* DOM and before its frame script has finished unloading. This
* parameter only has an effect for frame message managers in
* the main process. Default is false.
*/
void addMessageListener(in AString messageName,
in nsIMessageListener listener,
[optional] in boolean listenWhenClosed);
/**
* Undo an |addMessageListener| call -- that is, calling this causes us to no
* longer invoke |listener| when |messageName| is received.
*
* removeMessageListener does not remove a message listener added via
* addWeakMessageListener; use removeWeakMessageListener for that.
*/
void removeMessageListener(in AString messageName,
in nsIMessageListener listener);
/**
* This is just like addMessageListener, except the message manager holds a
* weak ref to |listener|.
*
* If you have two weak message listeners for the same message, they may be
* called in any order.
*/
void addWeakMessageListener(in AString messageName,
in nsIMessageListener listener);
/**
* This undoes an |addWeakMessageListener| call.
*/
void removeWeakMessageListener(in AString messageName,
in nsIMessageListener listener);
[notxpcom] boolean markForCC();
};
/**
* Message "senders" have a single "other side" to which messages are
* sent. For example, a child-process message manager will send
* messages that are only delivered to its one parent-process message
* manager.
*/
[scriptable, builtinclass, uuid(bb5d79e4-e73c-45e7-9651-4d718f4b994c)]
interface nsIMessageSender : nsIMessageListenerManager
{
/**
* Send |messageName| and |obj| to the "other side" of this message
* manager. This invokes listeners who registered for
* |messageName|.
*
* See nsIMessageListener::receiveMessage() for the format of the
* data delivered to listeners.
* @throws NS_ERROR_NOT_INITIALIZED if the sender is not initialized. For
* example, we will throw NS_ERROR_NOT_INITIALIZED if we try to send
* a message to a cross-process frame but the other process has not
* yet been set up.
* @throws NS_ERROR_FAILURE when the message receiver cannot be found. For
* example, we will throw NS_ERROR_FAILURE if we try to send a message
* to a cross-process frame whose process has crashed.
*/
[implicit_jscontext, optional_argc]
void sendAsyncMessage([optional] in AString messageName,
[optional] in jsval obj,
[optional] in jsval objects,
[optional] in nsIPrincipal principal,
[optional] in jsval transfers);
};
/**
* Message "broadcasters" don't have a single "other side" that they
* send messages to, but rather a set of subordinate message managers.
* For example, broadcasting a message through a window message
* manager will broadcast the message to all frame message managers
* within its window.
*/
[scriptable, builtinclass, uuid(4d7d62ad-4725-4f39-86cf-8fb22bf9c1d8)]
interface nsIMessageBroadcaster : nsIMessageListenerManager
{
/**
* Like |sendAsyncMessage()|, but also broadcasts this message to
* all "child" message managers of this message manager. See long
* comment above for details.
*
* WARNING: broadcasting messages can be very expensive and leak
* sensitive data. Use with extreme caution.
*/
[implicit_jscontext, optional_argc]
void broadcastAsyncMessage([optional] in AString messageName,
[optional] in jsval obj,
[optional] in jsval objects);
/**
* Number of subordinate message managers.
*/
readonly attribute unsigned long childCount;
/**
* Return a single subordinate message manager.
*/
nsIMessageListenerManager getChildAt(in unsigned long aIndex);
};
[scriptable, builtinclass, uuid(0e602c9e-1977-422a-a8e4-fe0d4a4f78d0)]
interface nsISyncMessageSender : nsIMessageSender
{
/**
* Like |sendAsyncMessage()|, except blocks the sender until all
* listeners of the message have been invoked. Returns an array
* containing return values from each listener invoked.
*/
[implicit_jscontext, optional_argc]
jsval sendSyncMessage([optional] in AString messageName,
[optional] in jsval obj,
[optional] in jsval objects,
[optional] in nsIPrincipal principal);
/**
* Like |sendSyncMessage()|, except re-entrant. New RPC messages may be
* issued even if, earlier on the call stack, we are waiting for a reply
* to an earlier sendRpcMessage() call.
*
* Both sendSyncMessage and sendRpcMessage will block until a reply is
* received, but they may be temporarily interrupted to process an urgent
* incoming message (such as a CPOW request).
*/
[implicit_jscontext, optional_argc]
jsval sendRpcMessage([optional] in AString messageName,
[optional] in jsval obj,
[optional] in jsval objects,
[optional] in nsIPrincipal principal);
};
[scriptable, builtinclass, uuid(13f3555f-769e-44ea-b607-5239230c3162)]
interface nsIMessageManagerGlobal : nsISyncMessageSender
{
/**
* Print a string to stdout.
*/
void dump(in DOMString aStr);
/**
* If leak detection is enabled, print a note to the leak log that this
* process will intentionally crash.
*/
void privateNoteIntentionalCrash();
/**
* Ascii base64 data to binary data and vice versa
*/
DOMString atob(in DOMString aAsciiString);
DOMString btoa(in DOMString aBase64Data);
};
[scriptable, builtinclass, uuid(694e367c-aa25-4446-8499-2c527c4bd838)]
interface nsIContentFrameMessageManager : nsIMessageManagerGlobal
{
/**
* The current top level window in the frame or null.
*/
readonly attribute mozIDOMWindowProxy content;
/**
* The top level docshell or null.
*/
readonly attribute nsIDocShell docShell;
};
[uuid(b39a3324-b574-4f85-8cdb-274d04f807ef)]
interface nsIInProcessContentFrameMessageManager : nsIContentFrameMessageManager
{
[notxpcom] nsIContent getOwnerContent();
[notxpcom] void cacheFrameLoader(in nsIFrameLoader aFrameLoader);
};
[scriptable, builtinclass, uuid(6d12e467-2446-46db-9965-e4e93cb87ca5)]
interface nsIContentProcessMessageManager : nsIMessageManagerGlobal
{
/**
* Read out a copy of the object that was initialized in the parent
* process via nsIProcessScriptLoader.initialProcessData.
*/
[implicit_jscontext]
readonly attribute jsval initialProcessData;
};
[scriptable, builtinclass, uuid(bf61446b-ba24-4b1d-88c7-4f94724b9ce1)]
interface nsIFrameScriptLoader : nsISupports
{
/**
* Load a script in the (remote) frame. aURL must be the absolute URL.
* data: URLs are also supported. For example data:,dump("foo\n");
* If aAllowDelayedLoad is true, script will be loaded when the
* remote frame becomes available. Otherwise the script will be loaded
* only if the frame is already available.
*/
void loadFrameScript(in AString aURL, in boolean aAllowDelayedLoad,
[optional] in boolean aRunInGlobalScope);
/**
* Removes aURL from the list of scripts which support delayed load.
*/
void removeDelayedFrameScript(in AString aURL);
/**
* Returns all delayed scripts that will be loaded once a (remote)
* frame becomes available. The return value is a list of pairs
* [<URL>, <WasLoadedInGlobalScope>].
*/
[implicit_jscontext]
jsval getDelayedFrameScripts();
};
[scriptable, builtinclass, uuid(7e1e1a20-b24f-11e4-ab27-0800200c9a66)]
interface nsIProcessScriptLoader : nsISupports
{
/**
* Load a script in the (possibly remote) process. aURL must be the absolute URL.
* data: URLs are also supported. For example data:,dump("foo\n");
* If aAllowDelayedLoad is true, script will be loaded when the
* remote frame becomes available. Otherwise the script will be loaded
* only if the frame is already available.
*/
void loadProcessScript(in AString aURL, in boolean aAllowDelayedLoad);
/**
* Removes aURL from the list of scripts which support delayed load.
*/
void removeDelayedProcessScript(in AString aURL);
/**
* Returns all delayed scripts that will be loaded once a (remote)
* frame becomes available. The return value is a list of URLs.
*/
[implicit_jscontext]
jsval getDelayedProcessScripts();
};
[scriptable, builtinclass, uuid(5b390753-abb3-49b0-ae3b-b803dab58144)]
interface nsIGlobalProcessScriptLoader : nsIProcessScriptLoader
{
/**
* Allows the parent process to set the initial process data for
* new, not-yet-created child processes. This attribute should only
* be used by the global parent process message manager. When a new
* process is created, it gets a copy of this data (via structured
* cloning). It can access the data via the initialProcessData
* attribute of its childprocessmessagemanager.
*
* This value will always be a JS object. Different users are
* expected to set properties on this object. The property name
* should be unique enough that other Gecko consumers won't
* accidentally choose it.
*/
[implicit_jscontext]
readonly attribute jsval initialProcessData;
};
[scriptable, builtinclass, uuid(637e8538-4f8f-4a3d-8510-e74386233e19)]
interface nsIProcessChecker : nsISupports
{
bool killChild();
/**
* Return true if the "remote" process has |aPermission|. This is
* intended to be used by JS implementations of cross-process DOM
* APIs, like so
*
* recvFooRequest: function(message) {
* if (!message.target.assertPermission("foo")) {
* return false;
* }
* // service foo request
*
* This interface only returns meaningful data when our content is
* in a separate process. If it shares the same OS process as us,
* then applying this permission check doesn't add any security,
* though it doesn't hurt anything either.
*
* Note: If the remote content process does *not* have |aPermission|,
* it will be killed as a precaution.
*/
boolean assertPermission(in DOMString aPermission);
/**
* Return true if the "remote" process has |aManifestURL|. This is
* intended to be used by JS implementations of cross-process DOM
* APIs, like so
*
* recvFooRequest: function(message) {
* if (!message.target.assertContainApp("foo")) {
* return false;
* }
* // service foo request
*
* This interface only returns meaningful data when our content is
* in a separate process. If it shares the same OS process as us,
* then applying this manifest URL check doesn't add any security,
* though it doesn't hurt anything either.
*
* Note: If the remote content process does *not* contain |aManifestURL|,
* it will be killed as a precaution.
*/
boolean assertContainApp(in DOMString aManifestURL);
boolean assertAppHasPermission(in DOMString aPermission);
/**
* Return true if the "remote" process' principal has an appStatus equal to
* |aStatus|.
*
* This interface only returns meaningful data when our content is
* in a separate process. If it shares the same OS process as us,
* then applying this permission check doesn't add any security,
* though it doesn't hurt anything either.
*
* Note: If the remote content process does *not* has the |aStatus|,
* it will be killed as a precaution.
*/
boolean assertAppHasStatus(in unsigned short aStatus);
};