sim: Refactor the serialization base class

Objects that are can be serialized are supposed to inherit from the
Serializable class. This class is meant to provide a unified API for
such objects. However, so far it has mainly been used by SimObjects
due to some fundamental design limitations. This changeset redesigns
to the serialization interface to make it more generic and hide the
underlying checkpoint storage. Specifically:

  * Add a set of APIs to serialize into a subsection of the current
    object. Previously, objects that needed this functionality would
    use ad-hoc solutions using nameOut() and section name
    generation. In the new world, an object that implements the
    interface has the methods serializeSection() and
    unserializeSection() that serialize into a named /subsection/ of
    the current object. Calling serialize() serializes an object into
    the current section.

  * Move the name() method from Serializable to SimObject as it is no
    longer needed for serialization. The fully qualified section name
    is generated by the main serialization code on the fly as objects
    serialize sub-objects.

  * Add a scoped ScopedCheckpointSection helper class. Some objects
    need to serialize data structures, that are not deriving from
    Serializable, into subsections. Previously, this was done using
    nameOut() and manual section name generation. To simplify this,
    this changeset introduces a ScopedCheckpointSection() helper
    class. When this class is instantiated, it adds a new /subsection/
    and subsequent serialization calls during the lifetime of this
    helper class happen inside this section (or a subsection in case
    of nested sections).

  * The serialize() call is now const which prevents accidental state
    manipulation during serialization. Objects that rely on modifying
    state can use the serializeOld() call instead. The default
    implementation simply calls serialize(). Note: The old-style calls
    need to be explicitly called using the
    serializeOld()/serializeSectionOld() style APIs. These are used by
    default when serializing SimObjects.

  * Both the input and output checkpoints now use their own named
    types. This hides underlying checkpoint implementation from
    objects that need checkpointing and makes it easier to change the
    underlying checkpoint storage code.

1 files changed, 205 insertions, 52 deletions

diff --git a/src/sim/serialize.hh b/src/sim/serialize.hh
index 2fb7cddd0..644ef4005 100644
--- a/src/sim/serialize.hh
+++ b/src/sim/serialize.hh
@@ -1,4 +1,16 @@
 /*
+ * Copyright (c) 2015 ARM Limited
+ * All rights reserved
+ *
+ * The license below extends only to copyright in the software and shall
+ * not be construed as granting a license to any other intellectual
+ * property including but not limited to intellectual property relating
+ * to a hardware implementation of the functionality of the software
+ * licensed hereunder.  You may use the software subject to the license
+ * terms below provided that you ensure that this notice is replicated
+ * unmodified and in its entirety in all distributions of the software,
+ * modified or unmodified, in source code or in binary form.
+ *
  * Copyright (c) 2002-2005 The Regents of The University of Michigan
  * All rights reserved.
  *
@@ -28,6 +40,7 @@
  * Authors: Nathan Binkert
  *          Erik Hallnor
  *          Steve Reinhardt
+ *          Andreas Sandberg
  */
 
 /* @file
@@ -41,6 +54,7 @@
 #include <iostream>
 #include <list>
 #include <map>
+#include <stack>
 #include <vector>
 
 #include "base/bitunion.hh"
@@ -48,10 +62,13 @@
 
 class IniFile;
 class Serializable;
-class Checkpoint;
+class CheckpointIn;
 class SimObject;
 class EventQueue;
 
+typedef std::ostream CheckpointOut;
+
+
 /** The current version of the checkpoint format.
  * This should be incremented by 1 and only 1 for every new version, where a new
  * version is defined as a checkpoint created before this version won't work on
@@ -62,66 +79,61 @@ class EventQueue;
 static const uint64_t gem5CheckpointVersion = 0x000000000000000e;
 
 template <class T>
-void paramOut(std::ostream &os, const std::string &name, const T &param);
+void paramOut(CheckpointOut &cp, const std::string &name, const T &param);
 
 template <typename DataType, typename BitUnion>
-void paramOut(std::ostream &os, const std::string &name,
+void paramOut(CheckpointOut &cp, const std::string &name,
               const BitfieldBackend::BitUnionOperators<DataType, BitUnion> &p)
 {
-    paramOut(os, name, p.__data);
+    paramOut(cp, name, p.__data);
 }
 
 template <class T>
-void paramIn(Checkpoint *cp, const std::string &section,
-             const std::string &name, T &param);
+void paramIn(CheckpointIn &cp, const std::string &name, T &param);
 
 template <typename DataType, typename BitUnion>
-void paramIn(Checkpoint *cp, const std::string &section,
-             const std::string &name,
+void paramIn(CheckpointIn &cp, const std::string &name,
              BitfieldBackend::BitUnionOperators<DataType, BitUnion> &p)
 {
-    paramIn(cp, section, name, p.__data);
+    paramIn(cp, name, p.__data);
 }
 
 template <class T>
-bool optParamIn(Checkpoint *cp, const std::string &section,
-             const std::string &name, T &param);
+bool optParamIn(CheckpointIn &cp, const std::string &name, T &param);
 
 template <typename DataType, typename BitUnion>
-bool optParamIn(Checkpoint *cp, const std::string &section,
-                const std::string &name,
+bool optParamIn(CheckpointIn &cp, const std::string &name,
                 BitfieldBackend::BitUnionOperators<DataType, BitUnion> &p)
 {
-    return optParamIn(cp, section, name, p.__data);
+    return optParamIn(cp, name, p.__data);
 }
 
 template <class T>
-void arrayParamOut(std::ostream &os, const std::string &name,
+void arrayParamOut(CheckpointOut &cp, const std::string &name,
                    const T *param, unsigned size);
 
 template <class T>
-void arrayParamOut(std::ostream &os, const std::string &name,
+void arrayParamOut(CheckpointOut &cp, const std::string &name,
                    const std::vector<T> &param);
 
 template <class T>
-void arrayParamOut(std::ostream &os, const std::string &name,
+void arrayParamOut(CheckpointOut &cp, const std::string &name,
                    const std::list<T> &param);
 
 template <class T>
-void arrayParamIn(Checkpoint *cp, const std::string &section,
-                  const std::string &name, T *param, unsigned size);
+void arrayParamIn(CheckpointIn &cp, const std::string &name,
+                  T *param, unsigned size);
 
 template <class T>
-void arrayParamIn(Checkpoint *cp, const std::string &section,
-                  const std::string &name, std::vector<T> &param);
+void arrayParamIn(CheckpointIn &cp, const std::string &name,
+                  std::vector<T> &param);
 
 template <class T>
-void arrayParamIn(Checkpoint *cp, const std::string &section,
-                  const std::string &name, std::list<T> &param);
+void arrayParamIn(CheckpointIn &cp, const std::string &name,
+                  std::list<T> &param);
 
 void
-objParamIn(Checkpoint *cp, const std::string &section,
-           const std::string &name, SimObject * &param);
+objParamIn(CheckpointIn &cp, const std::string &name, SimObject * &param);
 
 template <typename T>
 void fromInt(T &t, int i)
@@ -139,45 +151,66 @@ void fromSimObject(T &t, SimObject *s)
 // These macros are streamlined to use in serialize/unserialize
 // functions.  It's assumed that serialize() has a parameter 'os' for
 // the ostream, and unserialize() has parameters 'cp' and 'section'.
-#define SERIALIZE_SCALAR(scalar)        paramOut(os, #scalar, scalar)
+#define SERIALIZE_SCALAR(scalar)        paramOut(cp, #scalar, scalar)
 
-#define UNSERIALIZE_SCALAR(scalar)      paramIn(cp, section, #scalar, scalar)
-#define UNSERIALIZE_OPT_SCALAR(scalar)      optParamIn(cp, section, #scalar, scalar)
+#define UNSERIALIZE_SCALAR(scalar)      paramIn(cp, #scalar, scalar)
+#define UNSERIALIZE_OPT_SCALAR(scalar)      optParamIn(cp, #scalar, scalar)
 
 // ENUMs are like SCALARs, but we cast them to ints on the way out
-#define SERIALIZE_ENUM(scalar)          paramOut(os, #scalar, (int)scalar)
+#define SERIALIZE_ENUM(scalar)          paramOut(cp, #scalar, (int)scalar)
 
 #define UNSERIALIZE_ENUM(scalar)                \
  do {                                           \
     int tmp;                                    \
-    paramIn(cp, section, #scalar, tmp);         \
-    fromInt(scalar, tmp);                    \
+    paramIn(cp, #scalar, tmp);                  \
+    fromInt(scalar, tmp);                       \
   } while (0)
 
 #define SERIALIZE_ARRAY(member, size)           \
-        arrayParamOut(os, #member, member, size)
+        arrayParamOut(cp, #member, member, size)
 
 #define UNSERIALIZE_ARRAY(member, size)         \
-        arrayParamIn(cp, section, #member, member, size)
+        arrayParamIn(cp, #member, member, size)
 
 #define SERIALIZE_CONTAINER(member)             \
-        arrayParamOut(os, #member, member)
+        arrayParamOut(cp, #member, member)
 
 #define UNSERIALIZE_CONTAINER(member)           \
-        arrayParamIn(cp, section, #member, member)
+        arrayParamIn(cp, #member, member)
 
-#define SERIALIZE_OBJPTR(objptr)        paramOut(os, #objptr, (objptr)->name())
+#define SERIALIZE_OBJPTR(objptr)        paramOut(cp, #objptr, (objptr)->name())
 
 #define UNSERIALIZE_OBJPTR(objptr)                      \
   do {                                                  \
     SimObject *sptr;                                    \
-    objParamIn(cp, section, #objptr, sptr);             \
+    objParamIn(cp, #objptr, sptr);                      \
     fromSimObject(objptr, sptr);                        \
   } while (0)
 
 /**
  * Basic support for object serialization.
  *
+ * Objects that support serialization should derive from this
+ * class. Such objects can largely be divided into two categories: 1)
+ * True SimObjects (deriving from SimObject), and 2) child objects
+ * (non-SimObjects).
+ *
+ * SimObjects are serialized automatically into their own sections
+ * automatically by the SimObject base class (see
+ * SimObject::serializeAll().
+ *
+ * SimObjects can contain other serializable objects that are not
+ * SimObjects. Much like normal serialized members are not serialized
+ * automatically, these objects will not be serialized automatically
+ * and it is expected that the objects owning such serializable
+ * objects call the required serialization/unserialization methods on
+ * child objects. The preferred method to serialize a child object is
+ * to call serializeSection() on the child, which serializes the
+ * object into a new subsection in the current section. Another option
+ * is to call serialize() directly, which serializes the object into
+ * the current section. The latter is not recommended as it can lead
+ * to naming clashes between objects.
+ *
  * @note Many objects that support serialization need to be put in a
  * consistent state when serialization takes place. We refer to the
  * action of forcing an object into a consistent state as
@@ -187,26 +220,146 @@ void fromSimObject(T &t, SimObject *s)
 class Serializable
 {
   protected:
-    void nameOut(std::ostream &os);
-    void nameOut(std::ostream &os, const std::string &_name);
+    /**
+     * Scoped checkpoint section helper class
+     *
+     * This helper class creates a section within a checkpoint without
+     * the need for a separate serializeable object. It is mainly used
+     * within the Serializable class when serializing or unserializing
+     * section (see serializeSection() and unserializeSection()). It
+     * can also be used to maintain backwards compatibility in
+     * existing code that serializes structs that are not inheriting
+     * from Serializable into subsections.
+     *
+     * When the class is instantiated, it appends a name to the active
+     * path in a checkpoint. The old path is later restored when the
+     * instance is destroyed. For example, serializeSection() could be
+     * implemented by instantiating a ScopedCheckpointSection and then
+     * calling serialize() on an object.
+     */
+    class ScopedCheckpointSection {
+      public:
+        template<class CP>
+        ScopedCheckpointSection(CP &cp, const char *name) {
+            pushName(name);
+            nameOut(cp);
+        }
+
+        template<class CP>
+        ScopedCheckpointSection(CP &cp, const std::string &name) {
+            pushName(name.c_str());
+            nameOut(cp);
+        }
+
+        ~ScopedCheckpointSection();
+
+        ScopedCheckpointSection() = delete;
+        ScopedCheckpointSection(const ScopedCheckpointSection &) = delete;
+        ScopedCheckpointSection &operator=(
+            const ScopedCheckpointSection &) = delete;
+        ScopedCheckpointSection &operator=(
+            ScopedCheckpointSection &&) = delete;
+
+      private:
+        void pushName(const char *name);
+        void nameOut(CheckpointOut &cp);
+        void nameOut(CheckpointIn &cp) {};
+    };
 
   public:
     Serializable();
     virtual ~Serializable();
 
-    // manditory virtual function, so objects must provide names
-    virtual const std::string name() const = 0;
-
-    virtual void serialize(std::ostream &os);
-    virtual void unserialize(Checkpoint *cp, const std::string &section);
-
-    static Serializable *create(Checkpoint *cp, const std::string &section);
+    /**
+     * Serialize an object
+     *
+     * Output an object's state into the current checkpoint section.
+     *
+     * @param cp Checkpoint state
+     */
+    virtual void serialize(CheckpointOut &cp) const = 0;
+
+    /**
+     * Unserialize an object
+     *
+     * Read an object's state from the current checkpoint section.
+     *
+     * @param cp Checkpoint state
+     */
+    virtual void unserialize(CheckpointIn &cp) = 0;
+
+    /**
+     * Serialize an object into a new section
+     *
+     * This method creates a new section in a checkpoint and calls
+     * serialize() to serialize the current object into that
+     * section. The name of the section is appended to the current
+     * checkpoint path.
+     *
+     * @param cp Checkpoint state
+     * @param name Name to append to the active path
+     */
+    void serializeSection(CheckpointOut &cp, const char *name) const;
+
+    void serializeSection(CheckpointOut &cp, const std::string &name) const {
+        serializeSection(cp, name.c_str());
+    }
+
+    /**
+     * Unserialize an a child object
+     *
+     * This method loads a child object from a checkpoint. The object
+     * name is appended to the active path to form a fully qualified
+     * section name and unserialize() is called.
+     *
+     * @param cp Checkpoint state
+     * @param name Name to append to the active path
+     */
+    void unserializeSection(CheckpointIn &cp, const char *name);
+
+    void unserializeSection(CheckpointIn &cp, const std::string &name) {
+        unserializeSection(cp, name.c_str());
+    }
+
+    /**
+     * @{
+     * @name Legacy interface
+     *
+     * Interface for objects that insist on changing their state when
+     * serializing. Such state change should be done in drain(),
+     * memWriteback(), or memInvalidate() and not in the serialization
+     * method. In general, if state changes occur in serialize, it
+     * complicates testing since it breaks assumptions about draining
+     * and serialization. It potentially also makes components more
+     * fragile since they there are no ordering guarantees when
+     * serializing SimObjects.
+     *
+     * @warn This interface is considered deprecated and should never
+     * be used.
+     */
+
+    virtual void serializeOld(CheckpointOut &cp) {
+        serialize(cp);
+    }
+    void serializeSectionOld(CheckpointOut &cp, const char *name);
+    void serializeSectionOld(CheckpointOut &cp, const std::string &name) {
+        serializeSectionOld(cp, name.c_str());
+    }
+    /** @} */
+
+    /** Get the fully-qualified name of the active section */
+    static const std::string &currentSection();
+
+    static Serializable *create(CheckpointIn &cp, const std::string &section);
 
     static int ckptCount;
     static int ckptMaxCount;
     static int ckptPrevCount;
     static void serializeAll(const std::string &cpt_dir);
-    static void unserializeGlobals(Checkpoint *cp);
+    static void unserializeGlobals(CheckpointIn &cp);
+
+  private:
+    static std::stack<std::string> path;
 };
 
 void debug_serialize(const std::string &cpt_dir);
@@ -258,7 +411,7 @@ class SerializableClass
     // for the object (specified by the second string argument), and
     // an optional config hierarchy node (specified by the third
     // argument).  A pointer to the new SerializableBuilder is returned.
-    typedef Serializable *(*CreateFunc)(Checkpoint *cp,
+    typedef Serializable *(*CreateFunc)(CheckpointIn &cp,
                                         const std::string &section);
 
     static std::map<std::string,CreateFunc> *classMap;
@@ -272,7 +425,7 @@ class SerializableClass
 
     // create Serializable given name of class and pointer to
     // configuration hierarchy node
-    static Serializable *createObject(Checkpoint *cp,
+    static Serializable *createObject(CheckpointIn &cp,
                                       const std::string &section);
 };
 
@@ -297,7 +450,7 @@ class SimObjectResolver
     virtual SimObject *resolveSimObject(const std::string &name) = 0;
 };
 
-class Checkpoint
+class CheckpointIn
 {
   private:
 
@@ -306,8 +459,8 @@ class Checkpoint
     SimObjectResolver &objNameResolver;
 
   public:
-    Checkpoint(const std::string &cpt_dir, SimObjectResolver &resolver);
-    ~Checkpoint();
+    CheckpointIn(const std::string &cpt_dir, SimObjectResolver &resolver);
+    ~CheckpointIn();
 
     const std::string cptDir;