Merge pull request #13982 from aschackmull/dataflow/typeflow-calledge-pruning

Dataflow: Add type-based call-edge pruning.

java/ql/lib/semmle/code/java/Expr.qll

@@ -1676,13 +1676,25 @@ abstract class InstanceAccess extends Expr {
   /** Holds if this instance access is to an enclosing instance of type `t`. */
   predicate isEnclosingInstanceAccess(RefType t) {
     t = this.getQualifier().getType().(RefType).getSourceDeclaration() and
-    t != this.getEnclosingCallable().getDeclaringType()
+    t != this.getEnclosingCallable().getDeclaringType() and
+    not this.isSuperInterfaceAccess()
     or
-    not exists(this.getQualifier()) and
+    (not exists(this.getQualifier()) or this.isSuperInterfaceAccess()) and
     exists(LambdaExpr lam | lam.asMethod() = this.getEnclosingCallable() |
       t = lam.getAnonymousClass().getEnclosingType()
     )
   }
+
+  // A default method on an interface, `I`, may be invoked using `I.super.m()`.
+  // This always refers to the implemented interfaces of `this`. This form of
+  // qualified `super` cannot be combined with accessing an enclosing instance.
+  // JLS 15.11.2. "Accessing Superclass Members using super"
+  // JLS 15.12. "Method Invocation Expressions"
+  // JLS 15.12.1. "Compile-Time Step 1: Determine Type to Search"
+  private predicate isSuperInterfaceAccess() {
+    this instanceof SuperAccess and
+    this.getQualifier().getType().(RefType).getSourceDeclaration() instanceof Interface
+  }
 }
 
 /**

java/ql/lib/semmle/code/java/dataflow/TypeFlow.qll

@@ -440,6 +440,18 @@ predicate arrayInstanceOfGuarded(ArrayAccess aa, RefType t) {
   )
 }
 
+/**
+ * Holds if `t` is the type of the `this` value corresponding to the the
+ * `SuperAccess`. As the `SuperAccess` expression has the type of the supertype,
+ * the type `t` is a stronger type bound.
+ */
+private predicate superAccess(SuperAccess sup, RefType t) {
+  sup.isEnclosingInstanceAccess(t)
+  or
+  sup.isOwnInstanceAccess() and
+  t = sup.getEnclosingCallable().getDeclaringType()
+}
+
 /**
  * Holds if `n` has type `t` and this information is discarded, such that `t`
  * might be a better type bound for nodes where `n` flows to. This might include
@@ -452,7 +464,8 @@ private predicate typeFlowBaseCand(TypeFlowNode n, RefType t) {
     downcastSuccessor(n.asExpr(), srctype) or
     instanceOfGuarded(n.asExpr(), srctype) or
     arrayInstanceOfGuarded(n.asExpr(), srctype) or
-    n.asExpr().(FunctionalExpr).getConstructedType() = srctype
+    n.asExpr().(FunctionalExpr).getConstructedType() = srctype or
+    superAccess(n.asExpr(), srctype)
   |
     t = srctype.(BoundedType).getAnUltimateUpperBoundType()
     or

java/ql/lib/semmle/code/java/dataflow/internal/DataFlowPrivate.qll

@@ -326,13 +326,18 @@ string ppReprType(DataFlowType t) {
   else result = t.toString()
 }
 
+pragma[nomagic]
+private predicate compatibleTypes0(DataFlowType t1, DataFlowType t2) {
+  erasedHaveIntersection(t1, t2)
+}
+
 /**
  * Holds if `t1` and `t2` are compatible, that is, whether data can flow from
  * a node of type `t1` to a node of type `t2`.
  */
 bindingset[t1, t2]
 pragma[inline_late]
-predicate compatibleTypes(DataFlowType t1, DataFlowType t2) { erasedHaveIntersection(t1, t2) }
+predicate compatibleTypes(DataFlowType t1, DataFlowType t2) { compatibleTypes0(t1, t2) }
 
 /** A node that performs a type cast. */
 class CastNode extends ExprNode {

java/ql/lib/semmle/code/java/dataflow/internal/DataFlowUtil.qll

@@ -133,6 +133,39 @@ private module Cached {
   }
 }
 
+/**
+ * Holds if the value of `node2` is given by `node1`.
+ */
+predicate localMustFlowStep(Node node1, Node node2) {
+  exists(Callable c | node1.(InstanceParameterNode).getCallable() = c |
+    exists(InstanceAccess ia |
+      ia = node2.asExpr() and ia.getEnclosingCallable() = c and ia.isOwnInstanceAccess()
+    )
+    or
+    c =
+      node2.(ImplicitInstanceAccess).getInstanceAccess().(OwnInstanceAccess).getEnclosingCallable()
+  )
+  or
+  exists(SsaImplicitInit init |
+    init.isParameterDefinition(node1.asParameter()) and init.getAUse() = node2.asExpr()
+  )
+  or
+  exists(SsaExplicitUpdate upd |
+    upd.getDefiningExpr().(VariableAssign).getSource() = node1.asExpr() and
+    upd.getAUse() = node2.asExpr()
+  )
+  or
+  node2.asExpr().(CastingExpr).getExpr() = node1.asExpr()
+  or
+  node2.asExpr().(AssignExpr).getSource() = node1.asExpr()
+  or
+  node1 =
+    unique(FlowSummaryNode n1 |
+      FlowSummaryImpl::Private::Steps::summaryLocalStep(n1.getSummaryNode(),
+        node2.(FlowSummaryNode).getSummaryNode(), true)
+    )
+}
+
 import Cached
 
 private predicate capturedVariableRead(Node n) {

java/ql/lib/semmle/code/java/frameworks/Thrift.qll

@@ -8,6 +8,7 @@ import java
  * A file detected as generated by the Apache Thrift Compiler.
  */
 class ThriftGeneratedFile extends GeneratedFile {
+  cached
   ThriftGeneratedFile() {
     exists(JavadocElement t | t.getFile() = this |
       exists(string msg | msg = t.getText() | msg.regexpMatch("(?i).*\\bAutogenerated by Thrift.*"))

java/ql/test/library-tests/dataflow/typeflow-dispatch/A.java

@@ -0,0 +1,60 @@
+import java.util.*;
+import java.util.function.*;
+
+public class A {
+  static String source(String tag) { return null; }
+
+  static void sink(Object o) { }
+
+  interface MyConsumer {
+    void run(Object o);
+  }
+
+  void apply(MyConsumer f, Object x) {
+    f.run(x);
+  }
+
+  void apply_wrap(MyConsumer f, Object x) {
+    apply(f, x);
+  }
+
+  void testLambdaDispatch1() {
+    apply_wrap(x -> { sink(x); }, source("A")); // $ hasValueFlow=A
+    apply_wrap(x -> { sink(x); }, null); // no flow
+    apply_wrap(x -> { }, source("B"));
+    apply_wrap(x -> { }, null);
+  }
+
+  void forEach_wrap(List<Object> l, Consumer<Object> f) {
+    l.forEach(f);
+  }
+
+  void testLambdaDispatch2() {
+    List<Object> tainted = new ArrayList<>();
+    tainted.add(source("L"));
+    List<Object> safe = new ArrayList<>();
+    forEach_wrap(safe, x -> { sink(x); }); // no flow
+    forEach_wrap(tainted, x -> { sink(x); }); // $ hasValueFlow=L
+  }
+
+  static class TaintedClass {
+    public String toString() { return source("TaintedClass"); }
+  }
+
+  static class SafeClass {
+    public String toString() { return "safe"; }
+  }
+
+  String convertToString(Object o) {
+    return o.toString();
+  }
+
+  String convertToString_wrap(Object o) {
+    return convertToString(o);
+  }
+
+  void testToString1() {
+    String unused = convertToString_wrap(new TaintedClass());
+    sink(convertToString_wrap(new SafeClass())); // no flow
+  }
+}

java/ql/test/library-tests/dataflow/typeflow-dispatch/test.ql

@@ -0,0 +1,2 @@
+import TestUtilities.InlineFlowTest
+import DefaultFlowTest