Java: Adopt shared SSA data-flow integration (WIP)

Author: hvitved

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

@@ -55,7 +55,7 @@ class BaseSsaSourceVariable extends TBaseSsaSourceVariable {
 }
 
 cached
-private module SsaImpl {
+private module BaseSsaImpl {
   /** Gets the destination variable of an update of a tracked variable. */
   cached
   BaseSsaSourceVariable getDestVar(VariableUpdate upd) {
@@ -436,7 +436,7 @@ private module SsaImpl {
   }
 }
 
-private import SsaImpl
+private import BaseSsaImpl
 private import SsaDefReaches
 import SsaPublic
 

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

@@ -10,6 +10,7 @@ private import FlowSummaryImpl as FlowSummaryImpl
 private import DataFlowImplCommon as DataFlowImplCommon
 private import semmle.code.java.controlflow.Guards
 private import semmle.code.java.dataflow.RangeUtils
+private import SsaImpl as SsaImpl
 
 /** Gets a string for approximating the name of a field. */
 string approximateFieldContent(FieldContent fc) { result = fc.getField().getName().prefix(1) }
@@ -21,6 +22,30 @@ private predicate deadcode(Expr e) {
   )
 }
 
+module SsaFlow {
+  module Impl = SsaImpl::DataFlowIntegration;
+
+  Impl::Node asNode(Node n) {
+    n = TSsaNode(result)
+    or
+    result.(Impl::ExprNode).getExpr() = n.asExpr()
+    or
+    result.(Impl::ExprPostUpdateNode).getExpr() = n.(PostUpdateNode).getPreUpdateNode().asExpr()
+    or
+    TExplicitParameterNode(result.(Impl::ParameterNode).getParameter()) = n
+  }
+
+  predicate localFlowStep(
+    SsaImpl::Impl::DefinitionExt def, Node nodeFrom, Node nodeTo, boolean isUseStep
+  ) {
+    Impl::localFlowStep(def, asNode(nodeFrom), asNode(nodeTo), isUseStep)
+  }
+
+  predicate localMustFlowStep(SsaImpl::Impl::DefinitionExt def, Node nodeFrom, Node nodeTo) {
+    Impl::localMustFlowStep(def, asNode(nodeFrom), asNode(nodeTo))
+  }
+}
+
 cached
 private module Cached {
   cached
@@ -31,6 +56,7 @@ private module Cached {
       not e.getType() instanceof VoidType and
       not e.getParent*() instanceof Annotation
     } or
+    TSsaNode(SsaFlow::Impl::SsaNode node) or
     TExplicitParameterNode(Parameter p) { exists(p.getCallable().getBody()) } or
     TImplicitVarargsArray(Call c) {
       c.getCallee().isVarargs() and
@@ -137,6 +163,8 @@ module Public {
       result = this.(FieldValueNode).getField().getType()
       or
       result instanceof TypeObject and this instanceof AdditionalNode
+      or
+      result = this.(SsaNode).getDefinitionExt().getSourceVariable().getType()
     }
 
     /** Gets the callable in which this node occurs. */
@@ -198,6 +226,18 @@ module Public {
     abstract predicate isParameterOf(DataFlowCallable c, int pos);
   }
 
+  class SsaNode extends Node, TSsaNode {
+    private SsaFlow::Impl::SsaNode node;
+
+    SsaNode() { this = TSsaNode(node) }
+
+    SsaImpl::Impl::DefinitionExt getDefinitionExt() { result = node.getDefinitionExt() }
+
+    override Location getLocation() { result = node.getLocation() }
+
+    override string toString() { result = node.toString() }
+  }
+
   /**
    * A parameter, viewed as a node in a data flow graph.
    */
@@ -398,7 +438,8 @@ module Private {
     result.asSummarizedCallable() = n.(FlowSummaryNode).getSummarizedCallable() or
     result.asCallable() = n.(CaptureNode).getSynthesizedCaptureNode().getEnclosingCallable() or
     result.asFieldScope() = n.(FieldValueNode).getField() or
-    result.asCallable() = any(Expr e | n.(AdditionalNode).nodeAt(e, _)).getEnclosingCallable()
+    result.asCallable() = any(Expr e | n.(AdditionalNode).nodeAt(e, _)).getEnclosingCallable() or
+    result.asCallable() = n.(SsaNode).getDefinitionExt().getBasicBlock().getEnclosingCallable()
   }
 
   /** Holds if `p` is a `ParameterNode` of `c` with position `pos`. */

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

@@ -610,7 +610,11 @@ predicate forceHighPrecision(Content c) {
 }
 
 /** Holds if `n` should be hidden from path explanations. */
-predicate nodeIsHidden(Node n) { n instanceof FlowSummaryNode }
+predicate nodeIsHidden(Node n) {
+  n instanceof FlowSummaryNode
+  or
+  n instanceof SsaNode
+}
 
 class LambdaCallKind = Method; // the "apply" method in the functional interface
 

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

@@ -12,6 +12,7 @@ private import semmle.code.java.dataflow.FlowSummary
 private import semmle.code.java.dataflow.InstanceAccess
 private import FlowSummaryImpl as FlowSummaryImpl
 private import TaintTrackingUtil as TaintTrackingUtil
+private import SsaImpl as SsaImpl
 private import DataFlowNodes
 import DataFlowNodes::Public
 
@@ -99,6 +100,10 @@ predicate hasNonlocalValue(FieldRead fr) {
   )
 }
 
+private predicate capturedVariableRead(Node n) {
+  n.asExpr().(VarRead).getVariable() instanceof CapturedVariable
+}
+
 cached
 private module Cached {
   /**
@@ -108,7 +113,7 @@ private module Cached {
   predicate localFlowStep(Node node1, Node node2) {
     simpleLocalFlowStep0(node1, node2, _)
     or
-    adjacentUseUse(node1.asExpr(), node2.asExpr())
+    SsaFlow::localFlowStep(_, node1, node2, _)
     or
     // Simple flow through library code is included in the exposed local
     // step relation, even though flow is technically inter-procedural
@@ -125,13 +130,31 @@ private module Cached {
   predicate simpleLocalFlowStep(Node node1, Node node2, string model) {
     simpleLocalFlowStep0(node1, node2, model)
     or
+    exists(boolean isUseStep |
+      SsaFlow::localFlowStep(_, node1, node2, isUseStep) and
+      not capturedVariableRead(node2) and
+      model = ""
+    |
+      isUseStep = false
+      or
+      not exists(FieldRead fr |
+        hasNonlocalValue(fr) and fr.getField().isStatic() and fr = node1.asExpr()
+      ) and
+      not FlowSummaryImpl::Private::Steps::prohibitsUseUseFlow(node1, _)
+    )
+    or
     any(AdditionalValueStep a).step(node1, node2) and
     pragma[only_bind_out](node1.getEnclosingCallable()) =
       pragma[only_bind_out](node2.getEnclosingCallable()) and
     model = "AdditionalValueStep" and
     // prevent recursive call
     (any(AdditionalValueStep a).step(_, _) implies any())
   }
+
+  cached
+  Node getABarrierNode(Guard guard, SsaVariable def, boolean branch) {
+    SsaFlow::asNode(result) = SsaImpl::DataFlowIntegration::getABarrierNode(guard, def, branch)
+  }
 }
 
 /**
@@ -147,14 +170,7 @@ predicate localMustFlowStep(Node node1, Node node2) {
       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()
-  )
+  SsaFlow::localMustFlowStep(_, node1, node2)
   or
   node2.asExpr().(CastingExpr).getExpr() = node1.asExpr()
   or
@@ -169,10 +185,6 @@ predicate localMustFlowStep(Node node1, Node node2) {
 
 import Cached
 
-private predicate capturedVariableRead(Node n) {
-  n.asExpr().(VarRead).getVariable() instanceof CapturedVariable
-}
-
 /**
  * Holds if there is a data flow step from `e1` to `e2` that only steps from
  * child to parent in the AST.
@@ -214,34 +226,8 @@ predicate simpleAstFlowStep(Expr e1, Expr e2) {
 private predicate simpleLocalFlowStep0(Node node1, Node node2, string model) {
   (
     TaintTrackingUtil::forceCachingInSameStage() and
-    // Variable flow steps through adjacent def-use and use-use pairs.
-    exists(SsaExplicitUpdate upd |
-      upd.getDefiningExpr().(VariableAssign).getSource() = node1.asExpr() or
-      upd.getDefiningExpr().(AssignOp) = node1.asExpr() or
-      upd.getDefiningExpr().(RecordBindingVariableExpr) = node1.asExpr()
-    |
-      node2.asExpr() = upd.getAFirstUse() and
-      not capturedVariableRead(node2)
-    )
-    or
-    exists(SsaImplicitInit init |
-      init.isParameterDefinition(node1.asParameter()) and
-      node2.asExpr() = init.getAFirstUse() and
-      not capturedVariableRead(node2)
-    )
-    or
-    adjacentUseUse(node1.asExpr(), node2.asExpr()) and
-    not exists(FieldRead fr |
-      hasNonlocalValue(fr) and fr.getField().isStatic() and fr = node1.asExpr()
-    ) and
-    not FlowSummaryImpl::Private::Steps::prohibitsUseUseFlow(node1, _) and
-    not capturedVariableRead(node2)
-    or
     ThisFlow::adjacentThisRefs(node1, node2)
     or
-    adjacentUseUse(node1.(PostUpdateNode).getPreUpdateNode().asExpr(), node2.asExpr()) and
-    not capturedVariableRead(node2)
-    or
     ThisFlow::adjacentThisRefs(node1.(PostUpdateNode).getPreUpdateNode(), node2)
     or
     simpleAstFlowStep(node1.asExpr(), node2.asExpr())
@@ -401,13 +387,16 @@ signature predicate guardChecksSig(Guard g, Expr e, boolean branch);
  * in data flow and taint tracking.
  */
 module BarrierGuard<guardChecksSig/3 guardChecks> {
+  pragma[nomagic]
+  private predicate guardChecksSsaDef(Guard g, SsaVariable v, boolean branch) {
+    guardChecks(g, v.getAUse(), branch)
+  }
+
   /** Gets a node that is safely guarded by the given guard check. */
   Node getABarrierNode() {
-    exists(Guard g, SsaVariable v, boolean branch, VarRead use |
-      guardChecks(g, v.getAUse(), branch) and
-      use = v.getAUse() and
-      g.controls(use.getBasicBlock(), branch) and
-      result.asExpr() = use
+    exists(Guard g, SsaVariable v, boolean branch |
+      guardChecksSsaDef(g, v, branch) and
+      result = getABarrierNode(g, v, branch)
     )
   }
 }