Merge pull request #15599 from aschackmull/dataflow/fieldflowbranchlimit-v2

Dataflow: update fieldFlowBranchLimit semantics

Author: aschackmull

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

@@ -290,6 +290,8 @@ predicate knownSourceModel(Node source, string model) { none() }
 
 predicate knownSinkModel(Node sink, string model) { none() }
 
+class DataFlowSecondLevelScope = Unit;
+
 /**
  * Holds if flow is allowed to pass from parameter `p` and back to itself as a
  * side-effect, resulting in a summary from `p` to itself.

cpp/ql/lib/semmle/code/cpp/ir/dataflow/internal/DataFlowImplSpecific.qll

@@ -22,6 +22,8 @@ module CppDataFlow implements InputSig<Location> {
 
   predicate getAdditionalFlowIntoCallNodeTerm = Private::getAdditionalFlowIntoCallNodeTerm/2;
 
+  predicate getSecondLevelScope = Private::getSecondLevelScope/1;
+
   predicate validParameterAliasStep = Private::validParameterAliasStep/2;
 
   predicate mayBenefitFromCallContext = Private::mayBenefitFromCallContext/1;

cpp/ql/lib/semmle/code/cpp/ir/dataflow/internal/DataFlowPrivate.qll

@@ -1583,3 +1583,74 @@ predicate validParameterAliasStep(Node node1, Node node2) {
     )
   )
 }
+
+private predicate isTopLevel(Cpp::Stmt s) { any(Function f).getBlock().getAStmt() = s }
+
+private Cpp::Stmt getAChainedBranch(Cpp::IfStmt s) {
+  result = s.getThen()
+  or
+  exists(Cpp::Stmt elseBranch | s.getElse() = elseBranch |
+    result = getAChainedBranch(elseBranch)
+    or
+    result = elseBranch and not elseBranch instanceof Cpp::IfStmt
+  )
+}
+
+private Instruction getInstruction(Node n) {
+  result = n.asInstruction() or
+  result = n.asOperand().getUse() or
+  result = n.(SsaPhiNode).getPhiNode().getBasicBlock().getFirstInstruction() or
+  n.(IndirectInstruction).hasInstructionAndIndirectionIndex(result, _) or
+  result = getInstruction(n.(PostUpdateNode).getPreUpdateNode())
+}
+
+private newtype TDataFlowSecondLevelScope =
+  TTopLevelIfBranch(Cpp::Stmt s) {
+    exists(Cpp::IfStmt ifstmt | s = getAChainedBranch(ifstmt) and isTopLevel(ifstmt))
+  } or
+  TTopLevelSwitchCase(Cpp::SwitchCase s) {
+    exists(Cpp::SwitchStmt switchstmt | s = switchstmt.getASwitchCase() and isTopLevel(switchstmt))
+  }
+
+/**
+ * A second-level control-flow scope in a `switch` or a chained `if` statement.
+ *
+ * This is a `switch` case or a branch of a chained `if` statement, given that
+ * the `switch` or `if` statement is top level, that is, it is not nested inside
+ * other CFG constructs.
+ */
+class DataFlowSecondLevelScope extends TDataFlowSecondLevelScope {
+  /** Gets a textual representation of this element. */
+  string toString() {
+    exists(Cpp::Stmt s | this = TTopLevelIfBranch(s) | result = s.toString())
+    or
+    exists(Cpp::SwitchCase s | this = TTopLevelSwitchCase(s) | result = s.toString())
+  }
+
+  /** Gets the primary location of this element. */
+  Cpp::Location getLocation() {
+    exists(Cpp::Stmt s | this = TTopLevelIfBranch(s) | result = s.getLocation())
+    or
+    exists(Cpp::SwitchCase s | this = TTopLevelSwitchCase(s) | result = s.getLocation())
+  }
+
+  /**
+   * Gets a statement directly contained in this scope. For an `if` branch, this
+   * is the branch itself, and for a `switch case`, this is one the statements
+   * of that case branch.
+   */
+  private Cpp::Stmt getAStmt() {
+    exists(Cpp::Stmt s | this = TTopLevelIfBranch(s) | result = s)
+    or
+    exists(Cpp::SwitchCase s | this = TTopLevelSwitchCase(s) | result = s.getAStmt())
+  }
+
+  /** Gets a data-flow node nested within this scope. */
+  Node getANode() {
+    getInstruction(result).getAst().(Cpp::ControlFlowNode).getEnclosingStmt().getParentStmt*() =
+      this.getAStmt()
+  }
+}
+
+/** Gets the second-level scope containing the node `n`, if any. */
+DataFlowSecondLevelScope getSecondLevelScope(Node n) { result.getANode() = n }

cpp/ql/test/library-tests/dataflow/taint-tests/swap2.cpp

@@ -84,7 +84,7 @@ void test_copy_assignment_operator()
 
     swap(z1, z2);
 
-    sink(z2.data1); // $ ir MISSING: ast
+    sink(z2.data1); // $ ir ast
     sink(z1.data1); // $ SPURIOUS: ir ast=81:27 ast=82:16
 }
 

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

@@ -2882,6 +2882,8 @@ predicate knownSourceModel(Node source, string model) { sourceNode(source, _, mo
 
 predicate knownSinkModel(Node sink, string model) { sinkNode(sink, _, model) }
 
+class DataFlowSecondLevelScope = Unit;
+
 /**
  * Holds if flow is allowed to pass from parameter `p` and back to itself as a
  * side-effect, resulting in a summary from `p` to itself.

csharp/ql/test/library-tests/dataflow/collections/CollectionFlow.ql

@@ -14,8 +14,6 @@ module ArrayFlowConfig implements DataFlow::ConfigSig {
       mc.getAnArgument() = sink.asExpr()
     )
   }
-
-  int fieldFlowBranchLimit() { result = 100 }
 }
 
 module ArrayFlow = DataFlow::Global<ArrayFlowConfig>;

csharp/ql/test/library-tests/dataflow/types/Types.ql

@@ -18,8 +18,6 @@ module TypesConfig implements DataFlow::ConfigSig {
       mc.getAnArgument() = sink.asExpr()
     )
   }
-
-  int fieldFlowBranchLimit() { result = 1000 }
 }
 
 import ValueFlowTest<TypesConfig>

go/ql/lib/semmle/go/dataflow/internal/DataFlowPrivate.qll

@@ -415,6 +415,8 @@ predicate knownSourceModel(Node source, string model) { sourceNode(source, _, mo
 
 predicate knownSinkModel(Node sink, string model) { sinkNode(sink, _, model) }
 
+class DataFlowSecondLevelScope = Unit;
+
 /**
  * Holds if flow is allowed to pass from parameter `p` and back to itself as a
  * side-effect, resulting in a summary from `p` to itself.

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

@@ -20,6 +20,8 @@ module JavaDataFlow implements InputSig<Location> {
 
   Node exprNode(DataFlowExpr e) { result = Public::exprNode(e) }
 
+  predicate getSecondLevelScope = Private::getSecondLevelScope/1;
+
   predicate mayBenefitFromCallContext = Private::mayBenefitFromCallContext/1;
 
   predicate viableImplInCallContext = Private::viableImplInCallContext/2;

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

@@ -581,6 +581,81 @@ predicate knownSourceModel(Node source, string model) { sourceNode(source, _, mo
 
 predicate knownSinkModel(Node sink, string model) { sinkNode(sink, _, model) }
 
+private predicate isTopLevel(Stmt s) {
+  any(Callable c).getBody() = s
+  or
+  exists(BlockStmt b | s = b.getAStmt() and isTopLevel(b))
+}
+
+private Stmt getAChainedBranch(IfStmt s) {
+  result = s.getThen()
+  or
+  exists(Stmt elseBranch | s.getElse() = elseBranch |
+    result = getAChainedBranch(elseBranch)
+    or
+    result = elseBranch and not elseBranch instanceof IfStmt
+  )
+}
+
+private newtype TDataFlowSecondLevelScope =
+  TTopLevelIfBranch(Stmt s) {
+    exists(IfStmt ifstmt | s = getAChainedBranch(ifstmt) and isTopLevel(ifstmt))
+  } or
+  TTopLevelSwitchCase(SwitchCase s) {
+    exists(SwitchStmt switchstmt | s = switchstmt.getACase() and isTopLevel(switchstmt))
+  }
+
+private SwitchCase getPrecedingCase(Stmt s) {
+  result = s
+  or
+  exists(SwitchStmt switch, int i |
+    s = switch.getStmt(i) and
+    not s instanceof SwitchCase and
+    result = getPrecedingCase(switch.getStmt(i - 1))
+  )
+}
+
+/**
+ * A second-level control-flow scope in a `switch` or a chained `if` statement.
+ *
+ * This is a `switch` case or a branch of a chained `if` statement, given that
+ * the `switch` or `if` statement is top level, that is, it is not nested inside
+ * other CFG constructs.
+ */
+class DataFlowSecondLevelScope extends TDataFlowSecondLevelScope {
+  /** Gets a textual representation of this element. */
+  string toString() {
+    exists(Stmt s | this = TTopLevelIfBranch(s) | result = s.toString())
+    or
+    exists(SwitchCase s | this = TTopLevelSwitchCase(s) | result = s.toString())
+  }
+
+  /**
+   * Gets a statement directly contained in this scope. For an `if` branch, this
+   * is the branch itself, and for a `switch case`, this is one the statements
+   * of that case branch.
+   */
+  private Stmt getAStmt() {
+    exists(Stmt s | this = TTopLevelIfBranch(s) | result = s)
+    or
+    exists(SwitchCase s | this = TTopLevelSwitchCase(s) |
+      result = s.getRuleStatement() or
+      s = getPrecedingCase(result)
+    )
+  }
+
+  /** Gets a data-flow node nested within this scope. */
+  Node getANode() { getRelatedExpr(result).getAnEnclosingStmt() = this.getAStmt() }
+}
+
+private Expr getRelatedExpr(Node n) {
+  n.asExpr() = result or
+  n.(PostUpdateNode).getPreUpdateNode().asExpr() = result
+}
+
+/** Gets the second-level scope containing the node `n`, if any. */
+DataFlowSecondLevelScope getSecondLevelScope(Node n) { result.getANode() = n }
+
 /**
  * Holds if flow is allowed to pass from parameter `p` and back to itself as a
  * side-effect, resulting in a summary from `p` to itself.