Merge pull request github#6682 from aschackmull/java/callbacks

Java: Add support for callback-based library models.

Author: aschackmull

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

@@ -7,10 +7,10 @@ private import semmle.code.java.dispatch.VirtualDispatch as VirtualDispatch
 
 private module DispatchImpl {
   /** Gets a viable implementation of the target of the given `Call`. */
-  Callable viableCallable(Call c) {
-    result = VirtualDispatch::viableCallable(c)
+  DataFlowCallable viableCallable(DataFlowCall c) {
+    result = VirtualDispatch::viableCallable(c.asCall())
     or
-    result.(SummarizedCallable) = c.getCallee().getSourceDeclaration()
+    result.(SummarizedCallable) = c.asCall().getCallee().getSourceDeclaration()
   }
 
   /**
@@ -45,7 +45,7 @@ private module DispatchImpl {
   private predicate relevantContext(Call ctx, int i) {
     exists(Callable c |
       mayBenefitFromCallContext(_, c, i) and
-      c = viableCallable(ctx)
+      c = VirtualDispatch::viableCallable(ctx)
     )
   }
 
@@ -88,24 +88,25 @@ private module DispatchImpl {
   }
 
   /**
-   * Holds if the set of viable implementations that can be called by `ma`
+   * Holds if the set of viable implementations that can be called by `call`
    * might be improved by knowing the call context. This is the case if the
    * qualifier is a parameter of the enclosing callable `c`.
    */
-  predicate mayBenefitFromCallContext(MethodAccess ma, Callable c) {
-    mayBenefitFromCallContext(ma, c, _)
+  predicate mayBenefitFromCallContext(DataFlowCall call, DataFlowCallable c) {
+    mayBenefitFromCallContext(call.asCall(), c, _)
   }
 
   /**
-   * Gets a viable dispatch target of `ma` in the context `ctx`. This is
-   * restricted to those `ma`s for which a context might make a difference.
+   * Gets a viable dispatch target of `call` in the context `ctx`. This is
+   * restricted to those `call`s for which a context might make a difference.
    */
-  Method viableImplInCallContext(MethodAccess ma, Call ctx) {
-    result = viableCallable(ma) and
-    exists(int i, Callable c, Method def, RefType t, boolean exact |
+  Method viableImplInCallContext(DataFlowCall call, DataFlowCall ctx) {
+    result = viableCallable(call) and
+    exists(int i, Callable c, Method def, RefType t, boolean exact, MethodAccess ma |
+      ma = call.asCall() and
       mayBenefitFromCallContext(ma, c, i) and
       c = viableCallable(ctx) and
-      contextArgHasType(ctx, i, t, exact) and
+      contextArgHasType(ctx.asCall(), i, t, exact) and
       ma.getMethod().getSourceDeclaration() = def
     |
       exact = true and result = VirtualDispatch::exactMethodImpl(def, t.getSourceDeclaration())

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

@@ -326,12 +326,14 @@ module Private {
      * The instance argument is considered to have index `-1`.
      */
     predicate argumentOf(DataFlowCall call, int pos) {
-      exists(Argument arg | this.asExpr() = arg | call = arg.getCall() and pos = arg.getPosition())
+      exists(Argument arg | this.asExpr() = arg |
+        call.asCall() = arg.getCall() and pos = arg.getPosition()
+      )
       or
-      call = this.(ImplicitVarargsArray).getCall() and
-      pos = call.getCallee().getNumberOfParameters() - 1
+      call.asCall() = this.(ImplicitVarargsArray).getCall() and
+      pos = call.asCall().getCallee().getNumberOfParameters() - 1
       or
-      pos = -1 and this = getInstanceArgument(call)
+      pos = -1 and this = getInstanceArgument(call.asCall())
       or
       this.(SummaryNode).isArgumentOf(call, pos)
     }
@@ -361,7 +363,7 @@ module Private {
 
     /** Gets the underlying call. */
     DataFlowCall getCall() {
-      result = this.asExpr()
+      result.asCall() = this.asExpr()
       or
       this.(SummaryNode).isOut(result)
     }

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

@@ -5,6 +5,7 @@ private import DataFlowDispatch
 private import semmle.code.java.controlflow.Guards
 private import semmle.code.java.dataflow.SSA
 private import ContainerFlow
+private import semmle.code.java.dataflow.FlowSummary
 private import FlowSummaryImpl as FlowSummaryImpl
 import DataFlowNodes::Private
 
@@ -24,7 +25,7 @@ class ReturnKind extends TReturnKind {
  * `kind`.
  */
 OutNode getAnOutNode(DataFlowCall call, ReturnKind kind) {
-  result = call.getNode() and
+  result.getCall() = call and
   kind = TNormalReturnKind()
 }
 
@@ -214,9 +215,55 @@ class DataFlowExpr = Expr;
 
 class DataFlowType = RefType;
 
-class DataFlowCall extends Call {
-  /** Gets the data flow node corresponding to this call. */
-  ExprNode getNode() { result.getExpr() = this }
+private newtype TDataFlowCall =
+  TCall(Call c) or
+  TSummaryCall(SummarizedCallable c, Node receiver) {
+    FlowSummaryImpl::Private::summaryCallbackRange(c, receiver)
+  }
+
+/** A call relevant for data flow. Includes both source calls and synthesized calls. */
+class DataFlowCall extends TDataFlowCall {
+  /** Gets the source (non-synthesized) call this corresponds to, if any. */
+  Call asCall() { this = TCall(result) }
+
+  /** Gets the enclosing callable of this call. */
+  abstract DataFlowCallable getEnclosingCallable();
+
+  /** Gets a textual representation of this call. */
+  abstract string toString();
+
+  /** Gets the location of this call. */
+  abstract Location getLocation();
+}
+
+/** A source call, that is, a `Call`. */
+class SrcCall extends DataFlowCall, TCall {
+  Call call;
+
+  SrcCall() { this = TCall(call) }
+
+  override DataFlowCallable getEnclosingCallable() { result = call.getEnclosingCallable() }
+
+  override string toString() { result = call.toString() }
+
+  override Location getLocation() { result = call.getLocation() }
+}
+
+/** A synthesized call inside a `SummarizedCallable`. */
+class SummaryCall extends DataFlowCall, TSummaryCall {
+  private SummarizedCallable c;
+  private Node receiver;
+
+  SummaryCall() { this = TSummaryCall(c, receiver) }
+
+  /** Gets the data flow node that this call targets. */
+  Node getReceiver() { result = receiver }
+
+  override DataFlowCallable getEnclosingCallable() { result = c }
+
+  override string toString() { result = "[summary] call to " + receiver + " in " + c }
+
+  override Location getLocation() { result = c.getLocation() }
 }
 
 /** Holds if `e` is an expression that always has the same Boolean value `val`. */
@@ -275,13 +322,23 @@ predicate isImmutableOrUnobservable(Node n) {
 /** Holds if `n` should be hidden from path explanations. */
 predicate nodeIsHidden(Node n) { n instanceof SummaryNode }
 
-class LambdaCallKind = Unit;
+class LambdaCallKind = Method; // the "apply" method in the functional interface
 
 /** Holds if `creation` is an expression that creates a lambda of kind `kind` for `c`. */
-predicate lambdaCreation(Node creation, LambdaCallKind kind, DataFlowCallable c) { none() }
+predicate lambdaCreation(Node creation, LambdaCallKind kind, DataFlowCallable c) {
+  exists(FunctionalExpr func, FunctionalInterface interface |
+    creation.asExpr() = func and
+    func.asMethod() = c and
+    func.getType().(RefType).getSourceDeclaration() = interface and
+    kind = interface.getRunMethod()
+  )
+}
 
 /** Holds if `call` is a lambda call of kind `kind` where `receiver` is the lambda expression. */
-predicate lambdaCall(DataFlowCall call, LambdaCallKind kind, Node receiver) { none() }
+predicate lambdaCall(DataFlowCall call, LambdaCallKind kind, Node receiver) {
+  receiver = call.(SummaryCall).getReceiver() and
+  getNodeDataFlowType(receiver).getSourceDeclaration().(FunctionalInterface).getRunMethod() = kind
+}
 
 /** Extra data-flow steps needed for lambda flow analysis. */
 predicate additionalLambdaFlowStep(Node nodeFrom, Node nodeTo, boolean preservesValue) { none() }

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

@@ -151,7 +151,7 @@ predicate simpleLocalFlowStep(Node node1, Node node2) {
     ma.getCallee().getSourceDeclaration() = m and m.returnsValue(argNo)
   |
     node2.asExpr() = ma and
-    node1.(ArgumentNode).argumentOf(ma, argNo)
+    node1.(ArgumentNode).argumentOf(any(DataFlowCall c | c.asCall() = ma), argNo)
   )
   or
   FlowSummaryImpl::Private::Steps::summaryLocalStep(node1, node2, true)

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

@@ -20,7 +20,7 @@ predicate parameterPosition(int i) { i in [-1 .. any(Parameter p).getPosition()]
 Node summaryNode(SummarizedCallable c, SummaryNodeState state) { result = getSummaryNode(c, state) }
 
 /** Gets the synthesized data-flow call for `receiver`. */
-DataFlowCall summaryDataFlowCall(Node receiver) { none() }
+SummaryCall summaryDataFlowCall(Node receiver) { result.getReceiver() = receiver }
 
 /** Gets the type of content `c`. */
 DataFlowType getContentType(Content c) { result = c.getType() }
@@ -35,13 +35,18 @@ DataFlowType getReturnType(SummarizedCallable c, ReturnKind rk) {
  * Gets the type of the `i`th parameter in a synthesized call that targets a
  * callback of type `t`.
  */
-DataFlowType getCallbackParameterType(DataFlowType t, int i) { none() }
+DataFlowType getCallbackParameterType(DataFlowType t, int i) {
+  result = getErasedRepr(t.(FunctionalInterface).getRunMethod().getParameterType(i))
+}
 
 /**
  * Gets the return type of kind `rk` in a synthesized call that targets a
  * callback of type `t`.
  */
-DataFlowType getCallbackReturnType(DataFlowType t, ReturnKind rk) { none() }
+DataFlowType getCallbackReturnType(DataFlowType t, ReturnKind rk) {
+  result = getErasedRepr(t.(FunctionalInterface).getRunMethod().getReturnType()) and
+  exists(rk)
+}
 
 /**
  * Holds if an external flow summary exists for `c` with input specification
@@ -106,13 +111,13 @@ class InterpretNode extends TInterpretNode {
   Node asNode() { this = TNode(result) }
 
   /** Gets the call that this node corresponds to, if any. */
-  DataFlowCall asCall() { result = this.asElement() }
+  DataFlowCall asCall() { result.asCall() = this.asElement() }
 
   /** Gets the callable that this node corresponds to, if any. */
   DataFlowCallable asCallable() { result = this.asElement() }
 
   /** Gets the target of this call, if any. */
-  Callable getCallTarget() { result = this.asCall().getCallee().getSourceDeclaration() }
+  Callable getCallTarget() { result = this.asCall().asCall().getCallee().getSourceDeclaration() }
 
   /** Gets a textual representation of this node. */
   string toString() {

java/ql/lib/semmle/code/java/dispatch/DispatchFlow.qll

@@ -142,8 +142,8 @@ private predicate viableParamCand(Call call, int i, ParameterNode p) {
  * Holds if `arg` is a possible argument to `p` taking virtual dispatch into account.
  */
 private predicate viableArgParamCand(ArgumentNode arg, ParameterNode p) {
-  exists(int i, Call call |
-    viableParamCand(call, i, p) and
+  exists(int i, DataFlowCall call |
+    viableParamCand(call.asCall(), i, p) and
     arg.argumentOf(call, i)
   )
 }

java/ql/lib/semmle/code/java/dispatch/ObjFlow.qll

@@ -43,8 +43,8 @@ private predicate viableParam(Call call, int i, ParameterNode p) {
  * Holds if `arg` is a possible argument to `p` taking virtual dispatch into account.
  */
 private predicate viableArgParam(ArgumentNode arg, ParameterNode p) {
-  exists(int i, Call call |
-    viableParam(call, i, p) and
+  exists(int i, DataFlowCall call |
+    viableParam(call.asCall(), i, p) and
     arg.argumentOf(call, i)
   )
 }

java/ql/src/Telemetry/ExternalAPI.qll

@@ -48,15 +48,15 @@ class ExternalAPI extends Callable {
   private DataFlow::Node getAnInput() {
     exists(Call call | call.getCallee().getSourceDeclaration() = this |
       result.asExpr().(Argument).getCall() = call or
-      result.(ArgumentNode).getCall() = call
+      result.(ArgumentNode).getCall().asCall() = call
     )
   }
 
   /** Gets a node that is an output from a call to this API. */
   private DataFlow::Node getAnOutput() {
     exists(Call call | call.getCallee().getSourceDeclaration() = this |
       result.asExpr() = call or
-      result.(DataFlow::PostUpdateNode).getPreUpdateNode().(ArgumentNode).getCall() = call
+      result.(DataFlow::PostUpdateNode).getPreUpdateNode().(ArgumentNode).getCall().asCall() = call
     )
   }
 

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

@@ -0,0 +1,100 @@
+package my.callback.qltest;
+
+public class A {
+  public interface Consumer1 {
+    void eat(Object o);
+  }
+
+  public interface Consumer2 {
+    void eat(Object o);
+  }
+
+  public interface Consumer3<T> {
+    void eat(T o);
+  }
+
+  static void applyConsumer1(Object x, Consumer1 con) {
+    // summary:
+    // con.eat(x);
+  }
+
+  static void applyConsumer2(Object x, Consumer2 con) {
+    // summary:
+    // con.eat(x);
+  }
+
+  static <T> void applyConsumer3(T x, Consumer3<T> con) {
+    // summary:
+    // con.eat(x);
+  }
+
+  public interface Producer1<T> {
+    T make();
+  }
+
+  static <T> T applyProducer1(Producer1<T> prod) {
+    // summary:
+    // return prod.make();
+    return null;
+  }
+
+  public interface Converter1<T1,T2> {
+    T2 conv(T1 x);
+  }
+
+  static <T1,T2> T2 applyConverter1(T1 x, Converter1<T1,T2> con) {
+    // summary:
+    // return con.conv(x);
+    return null;
+  }
+
+  static Object source(int i) { return null; }
+
+  static void sink(Object o) { }
+
+  void foo(boolean b1, boolean b2) {
+    applyConsumer1(source(1), p -> {
+      sink(p); // $ flow=1
+    });
+
+    Object handler;
+    if (b1) {
+      handler = (Consumer1)(p -> { sink(p); }); // $ flow=2
+    } else {
+      handler = (Consumer2)(p -> { sink(p); }); // $ flow=3
+    }
+    if (b2) {
+      applyConsumer1(source(2), (Consumer1)handler);
+    } else {
+      applyConsumer2(source(3), (Consumer2)handler);
+    }
+
+    applyConsumer1(source(4), new Consumer1() {
+      @Override public void eat(Object o) {
+        sink(o); // $ MISSING: flow=4
+      }
+    });
+
+    applyConsumer1(source(5), A::sink); // $ flow=5
+
+    Consumer2 c = new MyConsumer2();
+    applyConsumer2(source(6), c);
+  }
+
+  static class MyConsumer2 implements Consumer2 {
+    @Override public void eat(Object o) {
+      sink(o); // $ MISSING: flow=6
+    }
+  }
+
+  void foo2() {
+    Consumer3<Integer> c = i -> sink(i); // $ flow=7
+    applyConsumer3((Integer)source(7), c);
+
+    sink(applyProducer1(() -> (Integer)source(8))); // $ flow=8
+
+    sink(applyConverter1((Integer)source(9), i -> i)); // $ flow=9
+
+    sink(applyConverter1((Integer)source(10), i -> new int[]{i})[0]); // $ flow=10
+  }
+}