security-context-v1: new protocol

This is a variation of the unveil protocol I suggested in the Weston
issue about security contexts. This lets sandbox engines such as Flatpak
attach a security context to sandboxed clients. The compositor can then
restrict which features are made available to that client.

The protocol is designed around the assumption that the sandbox engine
uses this protocol when setting up the sandboxed application. After this
inital setup, the sandbox engine isn't necessarily running anymore.
For this reason, a special "close FD" is used to indicate when to stop
the security context listener: the sandbox engine can leak the FD into
the sandboxed app's process, and the OS will automatically close the FD
when the sandboxed app exits.

Signed-off-by: Simon Ser <contact@emersion.fr>
References: https://gitlab.freedesktop.org/wayland/weston/-/issues/206

4 files changed, 190 insertions, 0 deletions

diff --git a/meson.build b/meson.build
index 3b97f85..17fe21f 100644
--- a/meson.build
+++ b/meson.build
@@ -43,6 +43,7 @@ staging_protocols = {
 	'ext-idle-notify': ['v1'],
 	'ext-session-lock': ['v1'],
 	'fractional-scale': ['v1'],
+	'security-context': ['v1'],
 	'single-pixel-buffer': ['v1'],
 	'tearing-control': ['v1'],
 	'xdg-activation': ['v1'],
diff --git a/staging/security-context/README b/staging/security-context/README
new file mode 100644
index 0000000..d12007d
--- /dev/null
+++ b/staging/security-context/README
@@ -0,0 +1,4 @@
+security_context protocol
+
+Maintainers:
+Simon Ser <contact@emersion.fr>
diff --git a/staging/security-context/engines.md b/staging/security-context/engines.md
new file mode 100644
index 0000000..dc5452a
--- /dev/null
+++ b/staging/security-context/engines.md
@@ -0,0 +1,14 @@
+# security-context-v1 engines
+
+This document describes how some specific engine implementations populate the
+metadata in security-context-v1.
+
+## [Flatpak]
+
+* `sandbox_engine` is always set to `flatpak`.
+* `app_id` is the Flatpak application ID (in reverse-DNS style). It is always
+  set.
+* `instance_id` is the Flatpak instance ID of the running sandbox. It is always
+  set.
+
+[Flatpak]: https://flatpak.org/
diff --git a/staging/security-context/security-context-v1.xml b/staging/security-context/security-context-v1.xml
new file mode 100644
index 0000000..c5de6cb
--- /dev/null
+++ b/staging/security-context/security-context-v1.xml
@@ -0,0 +1,171 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<protocol name="security_context_v1">
+  <copyright>
+    Copyright © 2021 Simon Ser
+
+    Permission is hereby granted, free of charge, to any person obtaining a
+    copy of this software and associated documentation files (the "Software"),
+    to deal in the Software without restriction, including without limitation
+    the rights to use, copy, modify, merge, publish, distribute, sublicense,
+    and/or sell copies of the Software, and to permit persons to whom the
+    Software is furnished to do so, subject to the following conditions:
+
+    The above copyright notice and this permission notice (including the next
+    paragraph) shall be included in all copies or substantial portions of the
+    Software.
+
+    THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+    IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+    FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+    THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+    LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+    FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+    DEALINGS IN THE SOFTWARE.
+  </copyright>
+
+  <interface name="wp_security_context_manager_v1" version="1">
+    <description summary="client security context manager">
+      This interface allows a client to register a new Wayland connection to
+      the compositor and attach a security context to it.
+
+      This is intended to be used by sandboxes. Sandbox engines attach a
+      security context to all connections coming from inside the sandbox. The
+      compositor can then restrict the features that the sandboxed connections
+      can use.
+
+      Compositors should forbid nesting multiple security contexts by not
+      exposing wp_security_context_manager_v1 global to clients with a security
+      context attached, or by sending the nested protocol error. Nested
+      security contexts are dangerous because they can potentially allow
+      privilege escalation of a sandboxed client.
+
+      Warning! The protocol described in this file is currently in the testing
+      phase. Backward compatible changes may be added together with the
+      corresponding interface version bump. Backward incompatible changes can
+      only be done by creating a new major version of the extension.
+    </description>
+
+    <enum name="error">
+      <entry name="invalid_listen_fd" value="1"
+        summary="listening socket FD is invalid"/>
+      <entry name="nested" value="2"
+        summary="nested security contexts are forbidden"/>
+    </enum>
+
+    <request name="destroy" type="destructor">
+      <description summary="destroy the manager object">
+        Destroy the manager. This doesn't destroy objects created with the
+        manager.
+      </description>
+    </request>
+
+    <request name="create_listener">
+      <description summary="create a new security context">
+        Creates a new security context with a socket listening FD.
+
+        The compositor will accept new client connections on listen_fd.
+        listen_fd must be ready to accept new connections when this request is
+        sent by the client. In other words, the client must call bind(2) and
+        listen(2) before sending the FD.
+
+        close_fd is a FD closed by the client when the compositor should stop
+        accepting new connections on listen_fd.
+
+        The compositor must continue to accept connections on listen_fd when
+        the Wayland client which created the security context disconnects.
+      </description>
+      <arg name="id" type="new_id" interface="wp_security_context_v1"/>
+      <arg name="listen_fd" type="fd" summary="listening socket FD"/>
+      <arg name="close_fd" type="fd" summary="FD closed when done"/>
+    </request>
+  </interface>
+
+  <interface name="wp_security_context_v1" version="1">
+    <description summary="client security context">
+      The security context allows a client to register a new client and attach
+      security context metadata to the connections.
+
+      When both are set, the combination of the application ID and the sandbox
+      engine must uniquely identify an application. The same application ID
+      will be used across instances (e.g. if the application is restarted, or
+      if the application is started multiple times).
+
+      When both are set, the combination of the instance ID and the sandbox
+      engine must uniquely identify a running instance of an application.
+    </description>
+
+    <enum name="error">
+      <entry name="already_used" value="1"
+        summary="security context has already been committed"/>
+      <entry name="already_set" value="2"
+        summary="metadata has already been set"/>
+      <entry name="invalid_metadata" value="3"
+        summary="metadata is invalid"/>
+    </enum>
+
+    <request name="destroy" type="destructor">
+      <description summary="destroy the security context object">
+        Destroy the security context object.
+      </description>
+    </request>
+
+    <request name="set_sandbox_engine">
+      <description summary="set the sandbox engine">
+        Attach a unique sandbox engine name to the security context.
+
+        A list of well-known engines is maintained at:
+        https://gitlab.freedesktop.org/wayland/wayland-protocols/-/blob/main/staging/security-context/engines.md
+
+        It is a protocol error to call this request twice. The already_set
+        error is sent in this case.
+      </description>
+      <arg name="name" type="string" summary="the sandbox engine name"/>
+    </request>
+
+    <request name="set_app_id">
+      <description summary="set the application ID">
+        Attach an application ID to the security context.
+
+        The application ID is an opaque, sandbox-specific identifier for an
+        application. See the well-known engines document for more details:
+        https://gitlab.freedesktop.org/wayland/wayland-protocols/-/blob/main/staging/security-context/engines.md
+
+        The compositor may use the application ID to group clients belonging to
+        the same security context application.
+
+        Whether this request is optional or not depends on the sandbox engine used.
+
+        It is a protocol error to call this request twice. The already_set
+        error is sent in this case.
+      </description>
+      <arg name="app_id" type="string" summary="the application ID"/>
+    </request>
+
+    <request name="set_instance_id">
+      <description summary="set the instance ID">
+        Attach an instance ID to the security context.
+
+        The instance ID is an opaque, sandbox-specific identifier for a running
+        instance of an application. See the well-known engines document for
+        more details:
+        https://gitlab.freedesktop.org/wayland/wayland-protocols/-/blob/main/staging/security-context/engines.md
+
+        Whether this request is optional or not depends on the sandbox engine used.
+
+        It is a protocol error to call this request twice. The already_set
+        error is sent in this case.
+      </description>
+      <arg name="instance_id" type="string" summary="the instance ID"/>
+    </request>
+
+    <request name="commit">
+      <description summary="register the security context">
+        Atomically register the new client and attach the security context
+        metadata.
+
+        It's a protocol error to send any request other than "destroy" after
+        this request. In this case, the already_used error is sent.
+      </description>
+    </request>
+  </interface>
+</protocol>