Merge branch 'main' of https://github.com/github/codeql into oscarsj/merge-back-rc-3.20

shared/dataflow/codeql/dataflow/internal/ContentDataFlowImpl.qll

@@ -75,6 +75,9 @@ module MakeImplContentDataFlow<LocationSig Location, InputSig<Location> Lang> {
     /** Gets a limit on the number of reads out of sources and number of stores into sinks. */
     default int accessPathLimit() { result = Lang::accessPathLimit() }
 
+    /** Gets the access path limit used in the internal invocation of the standard data flow library. */
+    default int accessPathLimitInternal() { result = Lang::accessPathLimit() }
+
     /** Holds if `c` is relevant for reads out of sources or stores into sinks. */
     default predicate isRelevantContent(ContentSet c) { any() }
   }
@@ -110,7 +113,7 @@ module MakeImplContentDataFlow<LocationSig Location, InputSig<Location> Lang> {
 
       FlowFeature getAFeature() { result = ContentConfig::getAFeature() }
 
-      predicate accessPathLimit = ContentConfig::accessPathLimit/0;
+      predicate accessPathLimit = ContentConfig::accessPathLimitInternal/0;
 
       // needed to record reads/stores inside summarized callables
       predicate includeHiddenNodes() { any() }
@@ -274,6 +277,16 @@ module MakeImplContentDataFlow<LocationSig Location, InputSig<Location> Lang> {
         )
       }
 
+      /**
+       * Gets the length of this access path.
+       */
+      int length() {
+        this = TAccessPathNil() and
+        result = 0
+        or
+        result = this.getTail().length() + 1
+      }
+
       /**
        * Gets the content set at index `i` in this access path, if any.
        */

shared/dataflow/codeql/dataflow/internal/DataFlowImpl.qll

@@ -207,6 +207,28 @@ module MakeImpl<LocationSig Location, InputSig<Location> Lang> {
       flowLocal(source, sink) and Config::observeOverlayInformedIncrementalMode()
     }
 
+    /**
+     * Holds if data can flow from `source` to some sink.
+     * This is a local predicate that only has results local to the overlay/base database.
+     */
+    predicate flowFromLocal(Node source) = forceLocal(Flow::flowFrom/1)(source)
+
+    /**
+     * Holds if data can flow from `source` to some sink.
+     */
+    predicate flowFrom(Node source) {
+      Flow::flowFrom(source)
+      or
+      // If we are overlay informed (i.e. we are not diff-informed), we
+      // merge in the local results which includes the base database results.
+      flowFromLocal(source) and Config::observeOverlayInformedIncrementalMode()
+    }
+
+    /**
+     * Holds if data can flow from `source` to some sink.
+     */
+    predicate flowFromExpr(Lang::DataFlowExpr source) { flowFrom(exprNode(source)) }
+
     /**
      * Holds if data can flow from some source to `sink`.
      * This is a local predicate that only has results local to the overlay/base database.
@@ -3501,6 +3523,16 @@ module MakeImpl<LocationSig Location, InputSig<Location> Lang> {
       )
     }
 
+    /**
+     * Holds if data can flow from `source` to some sink.
+     */
+    predicate flowFrom(Node source) { exists(PathNode n | n.isSource() and n.getNode() = source) }
+
+    /**
+     * Holds if data can flow from `source` to some sink.
+     */
+    predicate flowFromExpr(Expr source) { flowFrom(exprNode(source)) }
+
     /**
      * Holds if data can flow from some source to `sink`.
      */

shared/dataflow/codeql/dataflow/internal/DataFlowImplConsistency.qll

@@ -74,6 +74,9 @@ signature module InputSig<LocationSig Location, DF::InputSig<Location> DataFlowL
   ) {
     none()
   }
+
+  /** Holds if `(n1, n2)` should be excluded from the consistency test `localFlowIsLocal`. */
+  default predicate localFlowIsLocalExclude(DataFlowLang::Node n1, DataFlowLang::Node n2) { none() }
 }
 
 module MakeConsistency<
@@ -169,6 +172,7 @@ module MakeConsistency<
   query predicate localFlowIsLocal(Node n1, Node n2, string msg) {
     simpleLocalFlowStep(n1, n2, _) and
     nodeGetEnclosingCallable(n1) != nodeGetEnclosingCallable(n2) and
+    not Input::localFlowIsLocalExclude(n1, n2) and
     msg = "Local flow step does not preserve enclosing callable."
   }
 
@@ -240,6 +244,13 @@ module MakeConsistency<
 
   private predicate hasPost(Node n) { exists(PostUpdateNode post | post.getPreUpdateNode() = n) }
 
+  /**
+   * Consider code like `a.b.f = source()`. There is flow from `source()` to
+   * `[post] a.b` (with an appropriate access path), but we also want there to
+   * be flow to `[post] a` (with an appropriate access path). The data flow
+   * library is able to infer this step because there is a read step from `a`
+   * to `a.b`, as long as the post-update node for `a` exists.
+   */
   query predicate reverseRead(Node n, string msg) {
     exists(Node n2 | readStep(n, _, n2) and hasPost(n2) and not hasPost(n)) and
     not Input::reverseReadExclude(n) and

shared/mad/codeql/mad/modelgenerator/internal/ModelGeneratorImpl.qll

@@ -277,6 +277,16 @@ module MakeModelGeneratorFactory<
      */
     predicate isAdditionalContentFlowStep(Lang::Node nodeFrom, Lang::Node nodeTo);
 
+    /**
+     * Gets the access path limit for content flow analysis.
+     */
+    default int contentAccessPathLimit() { result = 2 }
+
+    /**
+     * Gets the internal access path limit for content flow analysis.
+     */
+    default int contentAccessPathLimitInternal() { result = Lang::accessPathLimit() }
+
     /**
      * Holds if the content set `c` is field like.
      */
@@ -650,7 +660,10 @@ module MakeModelGeneratorFactory<
           exists(Type t | t = n.(NodeExtended).getType() and not isRelevantType(t))
         }
 
-        int accessPathLimit() { result = 2 }
+        predicate accessPathLimit = SummaryModelGeneratorInput::contentAccessPathLimit/0;
+
+        predicate accessPathLimitInternal =
+          SummaryModelGeneratorInput::contentAccessPathLimitInternal/0;
 
         predicate isRelevantContent(DataFlow::ContentSet s) { isRelevantContent0(s) }
 
@@ -703,14 +716,17 @@ module MakeModelGeneratorFactory<
       }
 
       /**
-       * Holds if the access path `ap` is not a parameter or returnvalue of a callback
-       * stored in a field.
+       * Holds if `ap` is valid for generating summary models.
        *
-       * That is, we currently don't include summaries that rely on parameters or return values
-       * of callbacks stored in fields.
+       * We currently don't include summaries that rely on parameters or return values
+       * of callbacks stored in fields, as those are not supported by the data flow
+       * library.
+       *
+       * We also exclude access paths with contents not supported by `printContent`.
        */
       private predicate validateAccessPath(PropagateContentFlow::AccessPath ap) {
-        not (mentionsField(ap) and mentionsCallback(ap))
+        not (mentionsField(ap) and mentionsCallback(ap)) and
+        forall(int i | i in [0 .. ap.length() - 1] | exists(getContent(ap, i)))
       }
 
       private predicate apiFlow(
@@ -720,7 +736,9 @@ module MakeModelGeneratorFactory<
       ) {
         PropagateContentFlow::flow(p, reads, returnNodeExt, stores, preservesValue) and
         getEnclosingCallable(returnNodeExt) = api and
-        getEnclosingCallable(p) = api
+        getEnclosingCallable(p) = api and
+        validateAccessPath(reads) and
+        validateAccessPath(stores)
       }
 
       /**
@@ -763,9 +781,7 @@ module MakeModelGeneratorFactory<
         PropagateContentFlow::AccessPath reads, ReturnNodeExt returnNodeExt,
         PropagateContentFlow::AccessPath stores, boolean preservesValue
       ) {
-        PropagateContentFlow::flow(p, reads, returnNodeExt, stores, preservesValue) and
-        getEnclosingCallable(returnNodeExt) = api and
-        getEnclosingCallable(p) = api and
+        apiFlow(api, p, reads, returnNodeExt, stores, preservesValue) and
         p = api.getARelevantParameterNode()
       }
 
@@ -956,8 +972,6 @@ module MakeModelGeneratorFactory<
           input = parameterNodeAsExactInput(p) + printReadAccessPath(reads) and
           output = getExactOutput(returnNodeExt) + printStoreAccessPath(stores) and
           input != output and
-          validateAccessPath(reads) and
-          validateAccessPath(stores) and
           (
             if mentionsField(reads) or mentionsField(stores)
             then lift = false and api.isRelevant()