Merge remote-tracking branch 'origin/GT-3309_CodeDatatype' into patch

Ghidra/Features/Decompiler/src/decompile/cpp/database.cc

@@ -398,34 +398,44 @@ void Symbol::restoreXml(const Element *el)
   restoreXmlBody(list.begin());
 }
 
-void FunctionSymbol::buildType(int4 size)
+/// Get the number of bytes consumed by a SymbolEntry representing \b this Symbol.
+/// By default, this is the number of bytes consumed by the Symbol's data-type.
+/// This gives the amount of leeway a search has when the address queried does not match
+/// the exact address of the Symbol. With functions, the bytes consumed by a SymbolEntry
+/// may not match the data-type size.
+/// \return the number of bytes in a default SymbolEntry
+int4 Symbol::getBytesConsumed(void) const
+
+{
+  return type->getSize();
+}
+
+void FunctionSymbol::buildType(void)
 
 {
   TypeFactory *types = scope->getArch()->types;
   type = types->getTypeCode();
-  // Entries for functions have small size starting at the entry address
-  // of the function in order to deal with non-contiguous functions
-  // The size used to always be 1, but now we need sizes (slightly) larger than 1
-  // to accomodate pointer constants that encode extra information in the lower bit(s)
-  // of an otherwise aligned pointer.   If the encoding is not aprior detected, it is interpreted
-  // initially as a straight address that comes up 1 (or more) bytes off of the start of the function
-  // In order to detect this, we need to lay down a slightly larger size than 1
-  if (size > 1)
-    type = types->getTypeArray(size,type);
-
   flags |= Varnode::namelock | Varnode::typelock;
 }
 
 /// Build a function \e shell, made up of just the name of the function and
 /// a placeholder data-type, without the underlying Funcdata object.
+/// A SymbolEntry for a function has a small size starting at the entry address,
+/// in order to deal with non-contiguous functions.
+/// We need a size (slightly) larger than 1 to accommodate pointer constants that encode
+/// extra information in the lower bit(s) of an otherwise aligned pointer.
+/// If the encoding is not initially detected, it is interpreted
+/// as a straight address that comes up 1 (or more) bytes off of the start of the function
+/// In order to detect this, we need to lay down a slightly larger size than 1
 /// \param sc is the Scope that will contain the new Symbol
 /// \param nm is the name of the new Symbol
-/// \param size is the number of bytes the Symbol should consume
+/// \param size is the number of bytes a SymbolEntry should consume
 FunctionSymbol::FunctionSymbol(Scope *sc,const string &nm,int4 size)
   : Symbol(sc)
 {
   fd = (Funcdata *)0;
-  buildType(size);
+  consumeSize = size;
+  buildType();
   name = nm;
 }
 
@@ -433,7 +443,8 @@ FunctionSymbol::FunctionSymbol(Scope *sc,int4 size)
   : Symbol(sc)
 {
   fd = (Funcdata *)0;
-  buildType(size);
+  consumeSize = size;
+  buildType();
 }
 
 FunctionSymbol::~FunctionSymbol(void) {
@@ -469,9 +480,9 @@ void FunctionSymbol::restoreXml(const Element *el)
     fd = new Funcdata("",scope,Address());
     fd->restoreXml(el);
     name = fd->getName();
-    if (type->getSize() < fd->getSize()) {
+    if (consumeSize < fd->getSize()) {
       if ((fd->getSize()>1)&&(fd->getSize() <= 8))
-	buildType(fd->getSize());
+	consumeSize = fd->getSize();
     }
   }
   else {			// functionshell
@@ -934,8 +945,9 @@ SymbolEntry *Scope::addMap(const SymbolEntry &entry)
     entry.symbol->flags |= Varnode::persist;
 
   SymbolEntry *res;
+  int4 consumeSize = entry.symbol->getBytesConsumed();
   if (entry.addr.isInvalid())
-    res = addDynamicMapInternal(entry.symbol,Varnode::mapped,entry.hash,0,entry.symbol->getType()->getSize(),entry.uselimit);
+    res = addDynamicMapInternal(entry.symbol,Varnode::mapped,entry.hash,0,consumeSize,entry.uselimit);
   else {
     if (entry.uselimit.empty()) {
       entry.symbol->flags |= Varnode::addrtied;
@@ -943,7 +955,7 @@ SymbolEntry *Scope::addMap(const SymbolEntry &entry)
       // can only happen if use is not limited
       entry.symbol->flags |= glb->symboltab->getProperty(entry.addr);
     }
-    res = addMapInternal(entry.symbol,Varnode::mapped,entry.addr,0,entry.symbol->getType()->getSize(),entry.uselimit);
+    res = addMapInternal(entry.symbol,Varnode::mapped,entry.addr,0,consumeSize,entry.uselimit);
     if (entry.addr.isJoin()) {
       // The address is a join,  we add extra SymbolEntry maps for each of the pieces
       JoinRecord *rec = glb->findJoin(entry.addr.getOffset());

Ghidra/Features/Decompiler/src/decompile/cpp/database.hh

@@ -208,6 +208,7 @@ public:
   void restoreXmlBody(List::const_iterator iter);		///< Restore details of the Symbol from XML
   virtual void saveXml(ostream &s) const;			///< Save \b this Symbol to an XML stream
   virtual void restoreXml(const Element *el);			///< Restore \b this Symbol from an XML stream
+  virtual int4 getBytesConsumed(void) const;			///< Get number of bytes consumed within the address->symbol map
 };
 
 /// Force a specific display format for constant symbols
@@ -235,14 +236,16 @@ inline bool SymbolEntry::isAddrTied(void) const {
 /// Symbol is thus associated with all the meta-data about the function.
 class FunctionSymbol : public Symbol {
   Funcdata *fd;				///< The underlying meta-data object for the function
+  int4 consumeSize;			///< Minimum number of bytes to consume with the start address
   virtual ~FunctionSymbol(void);
-  void buildType(int4 size);		///< Build the data-type associated with \b this Symbol
+  void buildType(void);			///< Build the data-type associated with \b this Symbol
 public:
   FunctionSymbol(Scope *sc,const string &nm,int4 size);	///< Construct given the name
   FunctionSymbol(Scope *sc,int4 size);			///< Constructor for use with restoreXml
   Funcdata *getFunction(void);				///< Get the underlying Funcdata object
   virtual void saveXml(ostream &s) const;
   virtual void restoreXml(const Element *el);
+  virtual int4 getBytesConsumed(void) const { return consumeSize; }
 };
 
 /// \brief A Symbol that holds \b equate information for a constant

Ghidra/Features/Decompiler/src/decompile/cpp/type.cc

@@ -1893,6 +1893,24 @@ Datatype *TypeFactory::downChain(Datatype *ptrtype,uintb &off)
   return getTypePointer(ptype->size,pt,ptype->getWordSize());
 }
 
+/// The data-type propagation system can push around data-types that are \e partial or are
+/// otherwise unrepresentable in the source language.  This method substitutes those data-types
+/// with a concrete data-type that is representable, or returns the same data-type if is already concrete.
+/// Its important that the returned data-type have the same size as the original data-type regardless.
+/// \param ct is the given data-type
+/// \return the concrete data-type
+Datatype *TypeFactory::concretize(Datatype *ct)
+
+{
+  type_metatype metatype = ct->getMetatype();
+  if (metatype == TYPE_CODE) {
+    if (ct->getSize() != 1)
+      throw LowlevelError("Primitive code data-type that is not size 1");
+    ct = getBase(1, TYPE_UNKNOWN);
+  }
+  return ct;
+}
+
 /// Restore a Datatype object from an XML tag description: either \<type>, \<typeref>, or \<void>
 /// \param el is the XML element describing the data-type
 /// \return the restored Datatype object

Ghidra/Features/Decompiler/src/decompile/cpp/type.hh

@@ -435,6 +435,7 @@ public:
 			bool dotdotdot);			///< Create a "function" datatype
   void destroyType(Datatype *ct);				///< Remove a data-type from \b this
   Datatype *downChain(Datatype *ptrtype,uintb &off);		///< Find a sub-type matching a pointer and offset
+  Datatype *concretize(Datatype *ct);				///< Convert given data-type to concrete form
   void dependentOrder(vector<Datatype *> &deporder) const;	///< Place all data-types in dependency order
   void saveXml(ostream &s) const;			///< Save \b this container to stream
   void saveXmlCoreTypes(ostream &s) const;		///< Save core types to stream

Ghidra/Features/Decompiler/src/decompile/cpp/varmap.cc

@@ -513,7 +513,7 @@ void ScopeLocal::createEntry(const RangeHint &a)
 {
   Address addr(space,a.start);
   Address usepoint;
-  Datatype *ct = a.type;
+  Datatype *ct = glb->types->concretize(a.type);
   int4 num = a.size/ct->getSize();
   if (num>1)
     ct = glb->types->getTypeArray(num,ct);