DialogSurfaceManager implementation.

media/base/android/java/src/org/chromium/media/IDialogSurfaceActivityMapper.aidl

Index: media/base/android/java/src/org/chromium/media/IDialogSurfaceActivityMapper.aidl
diff --git a/media/base/android/java/src/org/chromium/media/IDialogSurfaceActivityMapper.aidl b/media/base/android/java/src/org/chromium/media/IDialogSurfaceActivityMapper.aidl
new file mode 100644
index 0000000000000000000000000000000000000000..99f6dbd1bce8bf0cd69cfbd73aada595b7c7926b
--- /dev/null
+++ b/media/base/android/java/src/org/chromium/media/IDialogSurfaceActivityMapper.aidl
@@ -0,0 +1,25 @@
+// Copyright 2016 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+package org.chromium.media;
+
+import android.os.IBinder;
+
+/**
+ * Class to facilitate mapping (pid, render_frame_id) to an Activity.  Should
+ * also handle registering for changes if the render frame is re-parented to
+ * another activity.
+ *
+ * Right now, all of this works on (pid, renderFrameId) pairs.  For security,
+ * we might want to switch to some sort of token that's acquired by the renderer
+ * and sent to the DialogSurface.  Otherwise, the GPU process can guess at
+ * renderFrameIds for any renderer pid.

[watk, 2016/07/26 21:34:31]

Using pids also theoretically means there is a race condition

[liberato (no reviews please), 2016/07/26 22:49:16]

good point, i added a note.

i doubt that it causes problems.  presumably, the gpu process would also know
that the renderer died and would clean up as part of that.  in the case of rogue
gpu processes, it's no worse than guessing pids.

+ */
+interface IDialogSurfaceActivityMapper {
+    /**
+     * Return the window token for the activity that hosts (renderPid,

[watk, 2016/07/26 21:34:30]

rendererPid

[liberato (no reviews please), 2016/07/26 22:49:16]

Done.

+     * renderFrameId).  This may return null.
+     */
+    IBinder getWindowToken(int rendererPid, int renderFrameId);
+}