drm/doc: updates for new framebuffer lifetime rules

Now that framebuffer are reference-counted for all use-sites, update
the documentation accordingly to stress the new rules for
initialization and teardown.

Also add a short paragraph about the implications for drivers of the
new locking rules.

Reviewed-by: Rob Clark <rob@ti.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Documentation/DocBook/drm.tmpl

@@ -978,10 +978,25 @@ int max_width, max_height;</synopsis>
         If the parameters are deemed valid, drivers then create, initialize and
         return an instance of struct <structname>drm_framebuffer</structname>.
         If desired the instance can be embedded in a larger driver-specific
-        structure. The new instance is initialized with a call to
-        <function>drm_framebuffer_init</function> which takes a pointer to DRM
-        frame buffer operations (struct
-        <structname>drm_framebuffer_funcs</structname>). Frame buffer operations are
+	structure. Drivers must fill its <structfield>width</structfield>,
+	<structfield>height</structfield>, <structfield>pitches</structfield>,
+        <structfield>offsets</structfield>, <structfield>depth</structfield>,
+        <structfield>bits_per_pixel</structfield> and
+        <structfield>pixel_format</structfield> fields from the values passed
+        through the <parameter>drm_mode_fb_cmd2</parameter> argument. They
+        should call the <function>drm_helper_mode_fill_fb_struct</function>
+        helper function to do so.
+      </para>
+
+      <para>
+	The initailization of the new framebuffer instance is finalized with a
+	call to <function>drm_framebuffer_init</function> which takes a pointer
+	to DRM frame buffer operations (struct
+	<structname>drm_framebuffer_funcs</structname>). Note that this function
+	publishes the framebuffer and so from this point on it can be accessed
+	concurrently from other threads. Hence it must be the last step in the
+	driver's framebuffer initialization sequence. Frame buffer operations
+	are
         <itemizedlist>
           <listitem>
             <synopsis>int (*create_handle)(struct drm_framebuffer *fb,
@@ -1022,16 +1037,16 @@ int max_width, max_height;</synopsis>
         </itemizedlist>
       </para>
       <para>
-        After initializing the <structname>drm_framebuffer</structname>
-        instance drivers must fill its <structfield>width</structfield>,
-        <structfield>height</structfield>, <structfield>pitches</structfield>,
-        <structfield>offsets</structfield>, <structfield>depth</structfield>,
-        <structfield>bits_per_pixel</structfield> and
-        <structfield>pixel_format</structfield> fields from the values passed
-        through the <parameter>drm_mode_fb_cmd2</parameter> argument. They
-        should call the <function>drm_helper_mode_fill_fb_struct</function>
-        helper function to do so.
-      </para>
+	The lifetime of a drm framebuffer is controlled with a reference count,
+	drivers can grab additional references with
+	<function>drm_framebuffer_reference</function> </para> and drop them
+	again with <function>drm_framebuffer_unreference</function>. For
+	driver-private framebuffers for which the last reference is never
+	dropped (e.g. for the fbdev framebuffer when the struct
+	<structname>drm_framebuffer</structname> is embedded into the fbdev
+	helper struct) drivers can manually clean up a framebuffer at module
+	unload time with
+	<function>drm_framebuffer_unregister_private</function>.
     </sect2>
     <sect2>
       <title>Output Polling</title>
@@ -1043,6 +1058,22 @@ int max_width, max_height;</synopsis>
         operation.
       </para>
     </sect2>
+    <sect2>
+      <title>Locking</title>
+      <para>
+        Beside some lookup structures with their own locking (which is hidden
+	behind the interface functions) most of the modeset state is protected
+	by the <code>dev-&lt;mode_config.lock</code> mutex and additionally
+	per-crtc locks to allow cursor updates, pageflips and similar operations
+	to occur concurrently with background tasks like output detection.
+	Operations which cross domains like a full modeset always grab all
+	locks. Drivers there need to protect resources shared between crtcs with
+	additional locking. They also need to be careful to always grab the
+	relevant crtc locks if a modset functions touches crtc state, e.g. for
+	load detection (which does only grab the <code>mode_config.lock</code>
+	to allow concurrent screen updates on live crtcs).
+      </para>
+    </sect2>
   </sect1>
 
   <!-- Internals: kms initialization and cleanup -->