Merge branch 'main' into rdmarsh2/cpp/ir-synthetic-destructors

Author: rdmarsh2

.bazelrc

@@ -1,4 +1,12 @@
 common --enable_platform_specific_config
+common --enable_bzlmod
+# because we use --override_module with `%workspace%`, the lock file is not stable
+common --lockfile_mode=off
+
+# when building from this repository in isolation, the internal repository will not be found at ..
+# where `MODULE.bazel` looks for it. The following will get us past the module loading phase, so
+# that we can build things that do not rely on that
+common --override_module=semmle_code=%workspace%/misc/bazel/semmle_code_stub
 
 build --repo_env=CC=clang --repo_env=CXX=clang++
 

.bazelversion

@@ -1 +1 @@
-6.3.1
+6.5.0

CODEOWNERS

@@ -25,6 +25,7 @@
 
 # Bazel (excluding BUILD.bazel files)
 WORKSPACE.bazel @github/codeql-ci-reviewers
+MODULE.bazel @github/codeql-ci-reviewers
 .bazelversion @github/codeql-ci-reviewers
 .bazelrc @github/codeql-ci-reviewers
 **/*.bzl @github/codeql-ci-reviewers

MODULE.bazel

@@ -0,0 +1,53 @@
+module(
+    name = "codeql",
+    version = "0.0",
+)
+
+# this points to our internal repository when `codeql` is checked out as a submodule thereof
+# when building things from `codeql` independently this is stubbed out in `.bazelrc`
+bazel_dep(name = "semmle_code", version = "0.0")
+local_path_override(
+    module_name = "semmle_code",
+    path = "..",
+)
+
+# see https://registry.bazel.build/ for a list of available packages
+
+bazel_dep(name = "platforms", version = "0.0.8")
+bazel_dep(name = "rules_pkg", version = "0.9.1")
+bazel_dep(name = "rules_nodejs", version = "6.0.3")
+bazel_dep(name = "rules_python", version = "0.31.0")
+bazel_dep(name = "bazel_skylib", version = "1.5.0")
+bazel_dep(name = "abseil-cpp", version = "20240116.0", repo_name = "absl")
+bazel_dep(name = "nlohmann_json", version = "3.11.3", repo_name = "json")
+bazel_dep(name = "fmt", version = "10.0.0")
+
+pip = use_extension("@rules_python//python/extensions:pip.bzl", "pip")
+pip.parse(
+    hub_name = "codegen_deps",
+    python_version = "3.11",
+    requirements_lock = "//misc/codegen:requirements_lock.txt",
+)
+use_repo(pip, "codegen_deps")
+
+swift_deps = use_extension("//swift/third_party:load.bzl", "swift_deps")
+use_repo(
+    swift_deps,
+    "binlog",
+    "picosha2",
+    "swift_prebuilt_darwin_x86_64",
+    "swift_prebuilt_linux",
+    "swift_toolchain_linux",
+    "swift_toolchain_macos",
+)
+
+node = use_extension("@rules_nodejs//nodejs:extensions.bzl", "node")
+node.toolchain(
+    name = "nodejs",
+    node_version = "18.15.0",
+)
+use_repo(node, "nodejs", "nodejs_toolchains")
+
+register_toolchains(
+    "@nodejs_toolchains//:all",
+)

WORKSPACE.bazel

@@ -1,12 +1,2 @@
-# Please notice that any bazel targets and definitions in this repository are currently experimental
-# and for internal use only.
-
-workspace(name = "codeql")
-
-load("//misc/bazel:workspace.bzl", "codeql_workspace")
-
-codeql_workspace()
-
-load("//misc/bazel:workspace_deps.bzl", "codeql_workspace_deps")
-
-codeql_workspace_deps()
+# please use MODULE.bazel to add dependencies
+# this empty file is required by internal repositories, don't remove it

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

@@ -55,29 +55,12 @@ private newtype TIRDataFlowNode =
   TFinalParameterNode(Parameter p, int indirectionIndex) {
     exists(Ssa::FinalParameterUse use |
       use.getParameter() = p and
-      use.getIndirectionIndex() = indirectionIndex and
-      parameterIsRedefined(p)
+      use.getIndirectionIndex() = indirectionIndex
     )
   } or
   TFinalGlobalValue(Ssa::GlobalUse globalUse) or
   TInitialGlobalValue(Ssa::GlobalDef globalUse)
 
-/**
- * Holds if the value of `*p` (or `**p`, `***p`, etc.) is redefined somewhere in the body
- * of the enclosing function of `p`.
- *
- * Only parameters satisfying this predicate will generate a `FinalParameterNode` transferring
- * flow out of the function.
- */
-private predicate parameterIsRedefined(Parameter p) {
-  exists(Ssa::Def def |
-    def.getSourceVariable().getBaseVariable().(Ssa::BaseIRVariable).getIRVariable().getAst() = p and
-    def.getIndirectionIndex() = 0 and
-    def.getIndirection() > 1 and
-    not def.getValue().asInstruction() instanceof InitializeParameterInstruction
-  )
-}
-
 /**
  * An operand that is defined by a `FieldAddressInstruction`.
  */

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

@@ -4,7 +4,11 @@ private import DataFlowUtil
 private import DataFlowImplCommon as DataFlowImplCommon
 private import semmle.code.cpp.models.interfaces.Allocation as Alloc
 private import semmle.code.cpp.models.interfaces.DataFlow as DataFlow
+private import semmle.code.cpp.models.interfaces.Taint as Taint
+private import semmle.code.cpp.models.interfaces.PartialFlow as PartialFlow
+private import semmle.code.cpp.models.interfaces.FunctionInputsAndOutputs as FIO
 private import semmle.code.cpp.ir.internal.IRCppLanguage
+private import semmle.code.cpp.ir.dataflow.internal.ModelUtil
 private import DataFlowPrivate
 private import ssa0.SsaInternals as SsaInternals0
 import SsaInternalsCommon
@@ -138,12 +142,11 @@ private newtype TDefOrUseImpl =
     isIteratorUse(container, iteratorAddress, _, indirectionIndex)
   } or
   TFinalParameterUse(Parameter p, int indirectionIndex) {
-    // Avoid creating parameter nodes if there is no definitions of the variable other than the initializaion.
-    exists(SsaInternals0::Def def |
-      def.getSourceVariable().getBaseVariable().(BaseIRVariable).getIRVariable().getAst() = p and
-      not def.getValue().asInstruction() instanceof InitializeParameterInstruction and
-      underlyingTypeIsModifiableAt(p.getUnderlyingType(), indirectionIndex)
-    )
+    underlyingTypeIsModifiableAt(p.getUnderlyingType(), indirectionIndex) and
+    // Only create an SSA read for the final use of a parameter if there's
+    // actually a body of the enclosing function. If there's no function body
+    // then we'll never need to flow out of the function anyway.
+    p.getFunction().hasDefinition()
   }
 
 private predicate isGlobalUse(
@@ -796,10 +799,58 @@ private Node getAPriorDefinition(SsaDefOrUse defOrUse) {
   )
 }
 
+private predicate inOut(FIO::FunctionInput input, FIO::FunctionOutput output) {
+  exists(int indirectionIndex |
+    input.isQualifierObject(indirectionIndex) and
+    output.isQualifierObject(indirectionIndex)
+    or
+    exists(int i |
+      input.isParameterDeref(i, indirectionIndex) and
+      output.isParameterDeref(i, indirectionIndex)
+    )
+  )
+}
+
+/**
+ * Holds if there should not be use-use flow out of `n`. That is, `n` is
+ * an out-barrier to use-use flow. This includes:
+ *
+ * - an input to a call that would be assumed to have use-use flow to the same
+ *   argument as an output, but this flow should be blocked because the
+ *   function is modeled with another flow to that output (for example the
+ *   first argument of `strcpy`).
+ * - a conversion that flows to such an input.
+ */
+private predicate modeledFlowBarrier(Node n) {
+  exists(
+    FIO::FunctionInput input, FIO::FunctionOutput output, CallInstruction call,
+    PartialFlow::PartialFlowFunction partialFlowFunc
+  |
+    n = callInput(call, input) and
+    inOut(input, output) and
+    exists(callOutput(call, output)) and
+    partialFlowFunc = call.getStaticCallTarget() and
+    not partialFlowFunc.isPartialWrite(output)
+  |
+    call.getStaticCallTarget().(DataFlow::DataFlowFunction).hasDataFlow(_, output)
+    or
+    call.getStaticCallTarget().(Taint::TaintFunction).hasTaintFlow(_, output)
+  )
+  or
+  exists(Operand operand, Instruction instr, Node n0, int indirectionIndex |
+    modeledFlowBarrier(n0) and
+    nodeHasInstruction(n0, instr, indirectionIndex) and
+    conversionFlow(operand, instr, false, _) and
+    nodeHasOperand(n, operand, indirectionIndex)
+  )
+}
+
 /** Holds if there is def-use or use-use flow from `nodeFrom` to `nodeTo`. */
 predicate ssaFlow(Node nodeFrom, Node nodeTo) {
   exists(Node nFrom, boolean uncertain, SsaDefOrUse defOrUse |
-    ssaFlowImpl(defOrUse, nFrom, nodeTo, uncertain) and nodeFrom != nodeTo
+    ssaFlowImpl(defOrUse, nFrom, nodeTo, uncertain) and
+    not modeledFlowBarrier(nFrom) and
+    nodeFrom != nodeTo
   |
     if uncertain = true then nodeFrom = [nFrom, getAPriorDefinition(defOrUse)] else nodeFrom = nFrom
   )

cpp/ql/lib/semmle/code/cpp/models/implementations/GetDelim.qll

@@ -15,6 +15,8 @@ private class GetDelimFunction extends TaintFunction, AliasFunction, SideEffectF
     i.isParameter(3) and o.isParameterDeref(0)
   }
 
+  override predicate isPartialWrite(FunctionOutput o) { o.isParameterDeref(3) }
+
   override predicate parameterNeverEscapes(int index) { index = [0, 1, 3] }
 
   override predicate parameterEscapesOnlyViaReturn(int index) { none() }

cpp/ql/lib/semmle/code/cpp/models/implementations/Gets.qll

@@ -27,6 +27,8 @@ private class FgetsFunction extends DataFlowFunction, TaintFunction, ArrayFuncti
     output.isReturnValue()
   }
 
+  override predicate isPartialWrite(FunctionOutput output) { output.isParameterDeref(2) }
+
   override predicate hasTaintFlow(FunctionInput input, FunctionOutput output) {
     input.isParameter(2) and
     output.isParameterDeref(0)

cpp/ql/lib/semmle/code/cpp/models/implementations/Inet.qll

@@ -20,6 +20,8 @@ private class InetAton extends TaintFunction, ArrayFunction {
     output.isParameterDeref(1)
   }
 
+  override predicate isPartialWrite(FunctionOutput output) { output.isParameterDeref(1) }
+
   override predicate hasArrayInput(int bufParam) { bufParam = 0 }
 
   override predicate hasArrayOutput(int bufParam) { bufParam = 1 }