mozilla
/
gecko-dev
зеркало из https://github.com/mozilla/gecko-dev.git
1
0
Форкнуть 0
Код Задачи Пакеты Проекты Релизы Вики Активность

Bug 1669556 - Simplify PreCreate handling WrapperFactory. r=mccr8 

Differential Revision: https://phabricator.services.mozilla.com/D92682
This commit is contained in:
Bobby Holley 2020-10-07 20:06:06 +00:00
Родитель 347101914e
Коммит ca99f17aad
1 изменённых файлов: 27 добавлений и 79 удалений
Показать статистику Скачать Patch файл Скачать Diff файл Показать все файлы Свернуть все файлы

106
js/xpconnect/wrappers/WrapperFactory.cpp
Просмотреть файл

@ -246,87 +246,35 @@ void WrapperFactory::PrepareForWrapping(JSContext* cx, HandleObject scope,
XPCCallContext ccx(cx, obj);
RootedObject wrapScope(cx, scope);
{
if (ccx.GetScriptable() && ccx.GetScriptable()->WantPreCreate()) {
// We have a precreate hook. This object might enforce that we only
// ever create JS object for it.
if (ccx.GetScriptable() && ccx.GetScriptable()->WantPreCreate()) {
// We have a precreate hook. This object might enforce that we only
// ever create JS object for it.
// Note: this penalizes objects that only have one wrapper, but are
// being accessed across compartments. We would really prefer to
// replace the above code with a test that says "do you only have one
// wrapper?"
nsresult rv = wn->GetScriptable()->PreCreate(wn->Native(), cx, scope,
wrapScope.address());
if (NS_FAILED(rv)) {
retObj.set(waive ? WaiveXray(cx, obj) : obj);
return;
}
// If the handed back scope differs from the passed-in scope and is in
// a separate compartment, then this object is explicitly requesting
// that we don't create a second JS object for it: create a security
// wrapper.
if (JS::GetCompartment(scope) != JS::GetCompartment(wrapScope)) {
retObj.set(waive ? WaiveXray(cx, obj) : obj);
return;
}
RootedObject currentScope(cx, JS::GetNonCCWObjectGlobal(obj));
if (MOZ_UNLIKELY(wrapScope != currentScope)) {
// The wrapper claims it wants to be in the new scope, but
// currently has a reflection that lives in the old scope. This
// can mean one of two things, both of which are rare:
//
// 1 - The object has a PreCreate hook (we checked for it above),
// but is deciding to request one-wrapper-per-scope (rather than
// one-wrapper-per-native) for some reason. Usually, a PreCreate
// hook indicates one-wrapper-per-native. In this case we want to
// make a new wrapper in the new scope.
//
// 2 - We're midway through wrapper reparenting. The document has
// moved to a new scope, but |wn| hasn't been moved yet, and
// we ended up calling JS_WrapObject() on its JS object. In this
// case, we want to return the existing wrapper.
//
// So we do a trick: call PreCreate _again_, but say that we're
// wrapping for the old scope, rather than the new one. If (1) is
// the case, then PreCreate will return the scope we pass to it
// (the old scope). If (2) is the case, PreCreate will return the
// scope of the document (the new scope).
RootedObject probe(cx);
rv = wn->GetScriptable()->PreCreate(wn->Native(), cx, currentScope,
probe.address());
// Check for case (2).
if (probe != currentScope) {
MOZ_ASSERT(probe == wrapScope);
retObj.set(waive ? WaiveXray(cx, obj) : obj);
return;
}
// Ok, must be case (1). Fall through and create a new wrapper.
}
// Nasty hack for late-breaking bug 781476. This will confuse identity
// checks, but it's probably better than any of our alternatives.
//
// Note: We have to ignore domain here. The JS engine assumes that, given
// a compartment c, if c->wrap(x) returns a cross-compartment wrapper at
// time t0, it will also return a cross-compartment wrapper for any time
// t1 > t0 unless an explicit transplant is performed. In particular,
// wrapper recomputation assumes that recomputing a wrapper will always
// result in a wrapper.
//
// This doesn't actually pose a security issue, because we'll still
// compute the correct (opaque) wrapper for the object below given the
// security characteristics of the two compartments.
if (!AccessCheck::isChrome(JS::GetCompartment(wrapScope)) &&
CompartmentOriginInfo::Subsumes(JS::GetCompartment(wrapScope),
JS::GetCompartment(obj))) {
retObj.set(waive ? WaiveXray(cx, obj) : obj);
return;
}
// Note: this penalizes objects that only have one wrapper, but are
// being accessed across compartments. We would really prefer to
// replace the above code with a test that says "do you only have one
// wrapper?"
nsresult rv = wn->GetScriptable()->PreCreate(wn->Native(), cx, scope,
wrapScope.address());
if (NS_FAILED(rv)) {
retObj.set(waive ? WaiveXray(cx, obj) : obj);
return;
}
// If the handed back scope differs from the passed-in scope and is in
// a separate compartment, then this object is explicitly requesting
// that we don't create a second JS object for it: create a security
// wrapper.
//
// Note: The only two objects that still use PreCreate are BackstagePass
// and Components, both of which unconditionally request their canonical
// scope. Since SpiderMonkey only invokes the prewrap callback in
// situations where the object is nominally cross-compartment, we should
// always get a different scope here.
MOZ_RELEASE_ASSERT(JS::GetCompartment(scope) !=
JS::GetCompartment(wrapScope));
retObj.set(waive ? WaiveXray(cx, obj) : obj);
return;
}
// This public WrapNativeToJSVal API enters the compartment of 'wrapScope'