Dataflow: Add support for not skipping configuration-specific nodes in big-step

cpp/ql/lib/change-notes/2023-07-19-neverskip-additionalstep.md

@@ -0,0 +1,8 @@
+---
+category: minorAnalysis
+---
+* Data flow configurations can now include a predicate `neverSkip(Node node)`
+  in order to ensure inclusion of certain nodes in the path explanations. The
+  predicate defaults to the end-points of the additional flow steps provided in
+  the configuration, which means that such steps now always are visible by
+  default in path explanations.

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

@@ -46,6 +46,14 @@ signature module ConfigSig {
    */
   default predicate allowImplicitRead(Node node, ContentSet c) { none() }
 
+  /**
+   * Holds if `node` should never be skipped over in the `PathGraph` and in path
+   * explanations.
+   */
+  default predicate neverSkip(Node node) {
+    isAdditionalFlowStep(node, _) or isAdditionalFlowStep(_, node)
+  }
+
   /**
    * Gets the virtual dispatch branching limit when calculating field flow.
    * This can be overridden to a smaller value to improve performance (a
@@ -141,6 +149,17 @@ signature module StateConfigSig {
    */
   default predicate allowImplicitRead(Node node, ContentSet c) { none() }
 
+  /**
+   * Holds if `node` should never be skipped over in the `PathGraph` and in path
+   * explanations.
+   */
+  default predicate neverSkip(Node node) {
+    isAdditionalFlowStep(node, _) or
+    isAdditionalFlowStep(_, node) or
+    isAdditionalFlowStep(node, _, _, _) or
+    isAdditionalFlowStep(_, _, node, _)
+  }
+
   /**
    * Gets the virtual dispatch branching limit when calculating field flow.
    * This can be overridden to a smaller value to improve performance (a

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

@@ -66,6 +66,12 @@ signature module FullStateConfigSig {
    */
   predicate allowImplicitRead(Node node, ContentSet c);
 
+  /**
+   * Holds if `node` should never be skipped over in the `PathGraph` and in path
+   * explanations.
+   */
+  predicate neverSkip(Node node);
+
   /**
    * Gets the virtual dispatch branching limit when calculating field flow.
    * This can be overridden to a smaller value to improve performance (a
@@ -2024,7 +2030,8 @@ module Impl<FullStateConfigSig Config> {
         castNode(this.asNode()) or
         clearsContentCached(this.asNode(), _) or
         expectsContentCached(this.asNode(), _) or
-        neverSkipInPathGraph(this.asNode())
+        neverSkipInPathGraph(this.asNode()) or
+        Config::neverSkip(this.asNode())
       }
     }
 

cpp/ql/lib/semmle/code/cpp/dataflow/internal/DataFlowImpl1.qll

@@ -313,6 +313,8 @@ private module Config implements FullStateConfigSig {
     any(Configuration config).allowImplicitRead(node, c)
   }
 
+  predicate neverSkip(Node node) { none() }
+
   int fieldFlowBranchLimit() { result = min(any(Configuration config).fieldFlowBranchLimit()) }
 
   FlowFeature getAFeature() { result = any(Configuration config).getAFeature() }

cpp/ql/lib/semmle/code/cpp/dataflow/internal/DataFlowImpl2.qll

@@ -313,6 +313,8 @@ private module Config implements FullStateConfigSig {
     any(Configuration config).allowImplicitRead(node, c)
   }
 
+  predicate neverSkip(Node node) { none() }
+
   int fieldFlowBranchLimit() { result = min(any(Configuration config).fieldFlowBranchLimit()) }
 
   FlowFeature getAFeature() { result = any(Configuration config).getAFeature() }

cpp/ql/lib/semmle/code/cpp/dataflow/internal/DataFlowImpl3.qll

@@ -313,6 +313,8 @@ private module Config implements FullStateConfigSig {
     any(Configuration config).allowImplicitRead(node, c)
   }
 
+  predicate neverSkip(Node node) { none() }
+
   int fieldFlowBranchLimit() { result = min(any(Configuration config).fieldFlowBranchLimit()) }
 
   FlowFeature getAFeature() { result = any(Configuration config).getAFeature() }

cpp/ql/lib/semmle/code/cpp/dataflow/internal/DataFlowImpl4.qll

@@ -313,6 +313,8 @@ private module Config implements FullStateConfigSig {
     any(Configuration config).allowImplicitRead(node, c)
   }
 
+  predicate neverSkip(Node node) { none() }
+
   int fieldFlowBranchLimit() { result = min(any(Configuration config).fieldFlowBranchLimit()) }
 
   FlowFeature getAFeature() { result = any(Configuration config).getAFeature() }

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

@@ -313,6 +313,8 @@ private module Config implements FullStateConfigSig {
     any(Configuration config).allowImplicitRead(node, c)
   }
 
+  predicate neverSkip(Node node) { none() }
+
   int fieldFlowBranchLimit() { result = min(any(Configuration config).fieldFlowBranchLimit()) }
 
   FlowFeature getAFeature() { result = any(Configuration config).getAFeature() }

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

@@ -46,6 +46,14 @@ signature module ConfigSig {
    */
   default predicate allowImplicitRead(Node node, ContentSet c) { none() }
 
+  /**
+   * Holds if `node` should never be skipped over in the `PathGraph` and in path
+   * explanations.
+   */
+  default predicate neverSkip(Node node) {
+    isAdditionalFlowStep(node, _) or isAdditionalFlowStep(_, node)
+  }
+
   /**
    * Gets the virtual dispatch branching limit when calculating field flow.
    * This can be overridden to a smaller value to improve performance (a
@@ -141,6 +149,17 @@ signature module StateConfigSig {
    */
   default predicate allowImplicitRead(Node node, ContentSet c) { none() }
 
+  /**
+   * Holds if `node` should never be skipped over in the `PathGraph` and in path
+   * explanations.
+   */
+  default predicate neverSkip(Node node) {
+    isAdditionalFlowStep(node, _) or
+    isAdditionalFlowStep(_, node) or
+    isAdditionalFlowStep(node, _, _, _) or
+    isAdditionalFlowStep(_, _, node, _)
+  }
+
   /**
    * Gets the virtual dispatch branching limit when calculating field flow.
    * This can be overridden to a smaller value to improve performance (a

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

@@ -66,6 +66,12 @@ signature module FullStateConfigSig {
    */
   predicate allowImplicitRead(Node node, ContentSet c);
 
+  /**
+   * Holds if `node` should never be skipped over in the `PathGraph` and in path
+   * explanations.
+   */
+  predicate neverSkip(Node node);
+
   /**
    * Gets the virtual dispatch branching limit when calculating field flow.
    * This can be overridden to a smaller value to improve performance (a
@@ -2024,7 +2030,8 @@ module Impl<FullStateConfigSig Config> {
         castNode(this.asNode()) or
         clearsContentCached(this.asNode(), _) or
         expectsContentCached(this.asNode(), _) or
-        neverSkipInPathGraph(this.asNode())
+        neverSkipInPathGraph(this.asNode()) or
+        Config::neverSkip(this.asNode())
       }
     }
 

cpp/ql/lib/semmle/code/cpp/ir/dataflow/internal/DataFlowImpl1.qll

@@ -313,6 +313,8 @@ private module Config implements FullStateConfigSig {
     any(Configuration config).allowImplicitRead(node, c)
   }
 
+  predicate neverSkip(Node node) { none() }
+
   int fieldFlowBranchLimit() { result = min(any(Configuration config).fieldFlowBranchLimit()) }
 
   FlowFeature getAFeature() { result = any(Configuration config).getAFeature() }

cpp/ql/lib/semmle/code/cpp/ir/dataflow/internal/DataFlowImpl2.qll

@@ -313,6 +313,8 @@ private module Config implements FullStateConfigSig {
     any(Configuration config).allowImplicitRead(node, c)
   }
 
+  predicate neverSkip(Node node) { none() }
+
   int fieldFlowBranchLimit() { result = min(any(Configuration config).fieldFlowBranchLimit()) }
 
   FlowFeature getAFeature() { result = any(Configuration config).getAFeature() }

cpp/ql/lib/semmle/code/cpp/ir/dataflow/internal/DataFlowImpl3.qll

@@ -313,6 +313,8 @@ private module Config implements FullStateConfigSig {
     any(Configuration config).allowImplicitRead(node, c)
   }
 
+  predicate neverSkip(Node node) { none() }
+
   int fieldFlowBranchLimit() { result = min(any(Configuration config).fieldFlowBranchLimit()) }
 
   FlowFeature getAFeature() { result = any(Configuration config).getAFeature() }

cpp/ql/lib/semmle/code/cpp/ir/dataflow/internal/DataFlowImpl4.qll

@@ -313,6 +313,8 @@ private module Config implements FullStateConfigSig {
     any(Configuration config).allowImplicitRead(node, c)
   }
 
+  predicate neverSkip(Node node) { none() }
+
   int fieldFlowBranchLimit() { result = min(any(Configuration config).fieldFlowBranchLimit()) }
 
   FlowFeature getAFeature() { result = any(Configuration config).getAFeature() }

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

@@ -448,6 +448,8 @@ module TaintedWithPath {
     }
 
     predicate isBarrierIn(DataFlow::Node node) { nodeIsBarrierIn(node) }
+
+    predicate neverSkip(Node node) { none() }
   }
 
   private module AdjustedFlow = TaintTracking::Global<AdjustedConfig>;

cpp/ql/src/Security/CWE/CWE-611/XXE.ql

@@ -43,6 +43,8 @@ module XxeConfig implements DataFlow::StateConfigSig {
     // flowstate value.
     node.asIndirectExpr().(XxeFlowStateTransformer).transform(flowstate) != flowstate
   }
+
+  predicate neverSkip(DataFlow::Node node) { none() }
 }
 
 module XxeFlow = DataFlow::GlobalWithState<XxeConfig>;