Reimplement CWE 476 check in Rust (#111)

caller/src/main.rs

 use cwe_checker_rs::intermediate_representation::Project;
 use cwe_checker_rs::utils::log::print_all_messages;
 use cwe_checker_rs::utils::{get_ghidra_plugin_path, read_config_file};
+use cwe_checker_rs::AnalysisResults;
 use std::collections::HashSet;
 use std::path::Path;
 use std::process::Command;
@@ -138,6 +139,17 @@ fn run_with_ghidra(args: CmdlineArgs) {
     let mut project = get_project_from_ghidra(&Path::new(&args.binary.unwrap()));
     // Normalize the project and gather log messages generated from it.
     let mut all_logs = project.normalize();
+    let mut analysis_results = AnalysisResults::new(&project);
+
+    let pointer_inference_results = if modules
+        .iter()
+        .any(|module| module.name == "CWE476" || module.name == "Memory")
+    {
+        Some(analysis_results.compute_pointer_inference(&config["Memory"]))
+    } else {
+        None
+    };
+    analysis_results = analysis_results.set_pointer_inference(pointer_inference_results.as_ref());
 
     // Print debug and then return.
     // Right now there is only one debug printing function.
@@ -154,7 +166,7 @@ fn run_with_ghidra(args: CmdlineArgs) {
     // Execute the modules and collect their logs and CWE-warnings.
     let mut all_cwes = Vec::new();
     for module in modules {
-        let (mut logs, mut cwes) = (module.run)(&project, &config[&module.name]);
+        let (mut logs, mut cwes) = (module.run)(&analysis_results, &config[&module.name]);
         all_logs.append(&mut logs);
         all_cwes.append(&mut cwes);
     }

cwe_checker_rs/src/abstract_domain/bitvector.rs

@@ -203,17 +203,7 @@ impl RegisterDomain for BitvectorDomain {
                     BitvectorDomain::new_top(self.bytesize())
                 }
             },
-            _ => match op {
-                Piece => BitvectorDomain::new_top(self.bytesize() + rhs.bytesize()),
-                IntAdd | IntSub | IntMult | IntDiv | IntSDiv | IntRem | IntSRem | IntLeft
-                | IntRight | IntSRight | IntAnd | IntOr | IntXOr | FloatAdd | FloatSub
-                | FloatMult | FloatDiv => BitvectorDomain::new_top(self.bytesize()),
-                IntEqual | IntNotEqual | IntLess | IntLessEqual | IntSLess | IntSLessEqual
-                | IntCarry | IntSCarry | IntSBorrow | BoolAnd | BoolOr | BoolXOr | FloatEqual
-                | FloatNotEqual | FloatLess | FloatLessEqual => {
-                    BitvectorDomain::new_top(ByteSize::new(1))
-                }
-            },
+            _ => BitvectorDomain::new_top(self.bin_op_bytesize(op, rhs)),
         }
     }
 

cwe_checker_rs/src/abstract_domain/mem_region.rs

@@ -114,6 +114,12 @@ impl<T: AbstractDomain + HasByteSize + RegisterDomain + std::fmt::Debug> MemRegi
     pub fn add(&mut self, value: T, position: Bitvector) {
         assert_eq!(ByteSize::from(position.width()), self.address_bytesize);
         let position = Int::from(position).try_to_i64().unwrap();
+        self.insert_at_byte_index(value, position);
+    }
+
+    /// Insert a value into the memory region at the given position.
+    /// The position is the index (in bytes) in the memory region.
+    pub fn insert_at_byte_index(&mut self, value: T, position: i64) {
         let size_in_bytes = u64::from(value.bytesize()) as i64;
         assert!(size_in_bytes > 0);
 

cwe_checker_rs/src/abstract_domain/mod.rs

@@ -69,4 +69,19 @@ pub trait RegisterDomain: AbstractDomain + HasByteSize + HasTop {
 
     /// Perform a typecast to extend a bitvector or to cast between integer and floating point types.
     fn cast(&self, kind: CastOpType, width: ByteSize) -> Self;
+
+    /// Return the bytesize of the result of the given binary operation.
+    /// Has a generic implementation that should not be overwritten!
+    fn bin_op_bytesize(&self, op: BinOpType, rhs: &Self) -> ByteSize {
+        use BinOpType::*;
+        match op {
+            Piece => self.bytesize() + rhs.bytesize(),
+            IntAdd | IntSub | IntMult | IntDiv | IntSDiv | IntRem | IntSRem | IntLeft
+            | IntRight | IntSRight | IntAnd | IntOr | IntXOr | FloatAdd | FloatSub | FloatMult
+            | FloatDiv => self.bytesize(),
+            IntEqual | IntNotEqual | IntLess | IntLessEqual | IntSLess | IntSLessEqual
+            | IntCarry | IntSCarry | IntSBorrow | BoolAnd | BoolOr | BoolXOr | FloatEqual
+            | FloatNotEqual | FloatLess | FloatLessEqual => ByteSize::new(1),
+        }
+    }
 }

cwe_checker_rs/src/analysis/pointer_inference/context/mod.rs

@@ -23,14 +23,12 @@ pub struct Context<'a> {
     pub project: &'a Project,
     /// Maps the TIDs of functions that shall be treated as extern symbols to the `ExternSymbol` object representing it.
     pub extern_symbol_map: BTreeMap<Tid, &'a ExternSymbol>,
-    /// A channel where found CWE warnings should be sent to.
+    /// A channel where found CWE warnings and log messages should be sent to.
     /// The receiver may filter or modify the warnings before presenting them to the user.
     /// For example, the same CWE warning will be found several times
     /// if the fixpoint computation does not instantly stabilize at the corresponding code point.
     /// These duplicates need to be filtered out.
-    pub cwe_collector: crossbeam_channel::Sender<CweWarning>,
-    /// A channel where log messages should be sent to.
-    pub log_collector: crossbeam_channel::Sender<LogMessage>,
+    pub log_collector: crossbeam_channel::Sender<LogThreadMsg>,
     /// Names of `malloc`-like extern functions.
     pub allocation_symbols: Vec<String>,
     /// Names of `free`-like extern functions.
@@ -43,8 +41,7 @@ impl<'a> Context<'a> {
     pub fn new(
         project: &Project,
         config: Config,
-        cwe_collector: crossbeam_channel::Sender<CweWarning>,
-        log_collector: crossbeam_channel::Sender<LogMessage>,
+        log_collector: crossbeam_channel::Sender<LogThreadMsg>,
     ) -> Context {
         let mut extern_symbol_map = BTreeMap::new();
         for symbol in project.program.term.extern_symbols.iter() {
@@ -63,7 +60,6 @@ impl<'a> Context<'a> {
             graph,
             project,
             extern_symbol_map,
-            cwe_collector,
             log_collector,
             allocation_symbols: config.allocation_symbols,
             deallocation_symbols: config.deallocation_symbols,
@@ -78,7 +74,7 @@ impl<'a> Context<'a> {
             if let Some(loc) = location {
                 log_message = log_message.location(loc.clone());
             };
-            self.log_collector.send(log_message).unwrap();
+            let _ = self.log_collector.send(LogThreadMsg::Log(log_message));
         }
     }
 
@@ -196,7 +192,7 @@ impl<'a> Context<'a> {
                                         call.tid.address
                                     ),
                                 };
-                                self.cwe_collector.send(warning).unwrap();
+                                let _ = self.log_collector.send(LogThreadMsg::Cwe(warning));
                             }
                         } else {
                             self.log_debug(
@@ -244,7 +240,7 @@ impl<'a> Context<'a> {
                                 extern_symbol.name, call.tid.address
                             ),
                         };
-                        self.cwe_collector.send(warning).unwrap();
+                        let _ = self.log_collector.send(LogThreadMsg::Cwe(warning));
                     }
                 }
                 Err(err) => self.log_debug(
@@ -349,12 +345,7 @@ impl<'a> Context<'a> {
     /// We also assume that the function does not use any parameters saved on the stack,
     /// which may greatly reduce correctness of the analysis for the x86_32 architecture.
     fn handle_call_to_generic_unknown_function(&self, state_before_call: &State) -> Option<State> {
-        if let Some(calling_conv) = self
-            .project
-            .calling_conventions
-            .iter()
-            .find(|cconv| cconv.name == "__stdcall")
-        {
+        if let Some(calling_conv) = self.project.get_standard_calling_convention() {
             let mut new_state = state_before_call.clone();
             new_state.clear_non_callee_saved_register(&calling_conv.callee_saved_register[..]);
             // Adjust stack register value (for x86 architecture).

cwe_checker_rs/src/analysis/pointer_inference/context/tests.rs

@@ -108,9 +108,8 @@ fn context_problem_implementation() {
     use Expression::*;
 
     let (project, config) = mock_project();
-    let (cwe_sender, _cwe_receiver) = crossbeam_channel::unbounded();
     let (log_sender, _log_receiver) = crossbeam_channel::unbounded();
-    let context = Context::new(&project, config, cwe_sender, log_sender);
+    let context = Context::new(&project, config, log_sender);
     let mut state = State::new(&register("RSP"), Tid::new("main"));
 
     let def = Term {
@@ -271,9 +270,8 @@ fn update_return() {
     use crate::analysis::pointer_inference::object::ObjectType;
     use crate::analysis::pointer_inference::Data;
     let (project, config) = mock_project();
-    let (cwe_sender, _cwe_receiver) = crossbeam_channel::unbounded();
     let (log_sender, _log_receiver) = crossbeam_channel::unbounded();
-    let context = Context::new(&project, config, cwe_sender, log_sender);
+    let context = Context::new(&project, config, log_sender);
     let state_before_return = State::new(&register("RSP"), Tid::new("callee"));
     let mut state_before_return = context
         .update_def(

cwe_checker_rs/src/analysis/pointer_inference/context/trait_impls.rs

@@ -29,7 +29,7 @@ impl<'a> crate::analysis::interprocedural_fixpoint::Context<'a> for Context<'a>
                     def.tid.address
                 ),
             };
-            self.cwe_collector.send(warning).unwrap();
+            let _ = self.log_collector.send(LogThreadMsg::Cwe(warning));
         }
 
         match &def.term {

cwe_checker_rs/src/checkers.rs

@@ -2,6 +2,7 @@ pub mod cwe_190;
 pub mod cwe_332;
 pub mod cwe_426;
 pub mod cwe_467;
+pub mod cwe_476;
 pub mod cwe_560;
 pub mod cwe_676;
 pub mod cwe_782;

cwe_checker_rs/src/checkers/cwe_190.rs

@@ -100,9 +100,10 @@ fn generate_cwe_warning(callsite: &Tid, called_symbol: &ExternSymbol) -> CweWarn
 /// For each call to one of the symbols configured in config.json
 /// we check whether the block containing the call also contains a multiplication instruction.
 pub fn check_cwe(
-    project: &Project,
+    analysis_results: &AnalysisResults,
     cwe_params: &serde_json::Value,
 ) -> (Vec<LogMessage>, Vec<CweWarning>) {
+    let project = analysis_results.project;
     let config: Config = serde_json::from_value(cwe_params.clone()).unwrap();
     let mut cwe_warnings = Vec::new();
     let symbol_map = get_symbol_map(project, &config.symbols);

cwe_checker_rs/src/checkers/cwe_332.rs

@@ -20,7 +20,6 @@
 //!
 //! - It is not checked whether the seeding function gets called before the random number generator function.
 
-use crate::intermediate_representation::*;
 use crate::prelude::*;
 use crate::utils::log::{CweWarning, LogMessage};
 use crate::utils::symbol_utils::find_symbol;
@@ -56,9 +55,10 @@ fn generate_cwe_warning(secure_initializer_func: &str, rand_func: &str) -> CweWa
 
 /// Run the CWE check. See the module-level description for more information.
 pub fn check_cwe(
-    project: &Project,
+    analysis_results: &AnalysisResults,
     cwe_params: &serde_json::Value,
 ) -> (Vec<LogMessage>, Vec<CweWarning>) {
+    let project = analysis_results.project;
     let config: Config = serde_json::from_value(cwe_params.clone()).unwrap();
     let mut cwe_warnings = Vec::new();
 

cwe_checker_rs/src/checkers/cwe_426.rs

@@ -70,9 +70,10 @@ fn generate_cwe_warning(sub: &Term<Sub>) -> CweWarning {
 /// We check whether a function calls both `system(..)` and a privilege changing function.
 /// For each such function a CWE warning is generated.
 pub fn check_cwe(
-    project: &Project,
+    analysis_results: &AnalysisResults,
     cwe_params: &serde_json::Value,
 ) -> (Vec<LogMessage>, Vec<CweWarning>) {
+    let project = analysis_results.project;
     let config: Config = serde_json::from_value(cwe_params.clone()).unwrap();
     let mut cwe_warnings = Vec::new();
     let mut privilege_changing_symbols = HashMap::new();

cwe_checker_rs/src/checkers/cwe_467.rs

@@ -103,9 +103,10 @@ fn generate_cwe_warning(jmp: &Term<Jmp>, extern_symbol: &ExternSymbol) -> CweWar
 /// we check whether a parameter has value `sizeof(void*)`,
 /// which may indicate an instance of CWE 467.
 pub fn check_cwe(
-    project: &Project,
+    analysis_results: &AnalysisResults,
     cwe_params: &serde_json::Value,
 ) -> (Vec<LogMessage>, Vec<CweWarning>) {
+    let project = analysis_results.project;
     let config: Config = serde_json::from_value(cwe_params.clone()).unwrap();
     let mut cwe_warnings = Vec::new();
 

cwe_checker_rs/src/checkers/cwe_476.rs

+//! This module implements a check for CWE-476: NULL Pointer Dereference.
+//!
+//! Functions like `malloc()` may return NULL values instead of pointers to indicate
+//! failed calls. If one tries to access memory through this return value without
+//! checking it for being NULL first, this can crash the program.
+//!
+//! See <https://cwe.mitre.org/data/definitions/476.html> for a detailed description.
+//!
+//! ## How the check works
+//!
+//! Using dataflow analysis we search for an execution path where a memory access using the return value of
+//! a symbol happens before the return value is checked through a conditional jump instruction.
+//!
+//! ### Symbols configurable in config.json
+//!
+//! The symbols are the functions whose return values are assumed to be potential
+//! NULL pointers.
+//!
+//! ## False Positives
+//!
+//! - If a possible NULL pointer is temporarily saved in a memory location
+//! that the [Pointer Inference analysis](crate::analysis::pointer_inference) could not track,
+//! the analysis may miss a correct NULL pointer check and thus generate false positives.
+//! - The analysis is intraprocedural.
+//! If a parameter to a function is a potential NULL pointer,
+//! this gets flagged as a CWE hit even if the function may expect NULL pointers in its parameters.
+//! If a function returns a potential NULL pointer this gets flagged as a CWE hit,
+//! although the function may be supposed to return potential NULL pointers.
+//!
+//! ## False Negatives
+//!
+//! - We do not check whether an access to a potential NULL pointer happens regardless
+//! of a prior check.
+//! - We do not check whether the conditional jump instruction checks specifically
+//! for the return value being NULL or something else
+//! - For functions with more than one return value we do not distinguish between
+//! the return values.
+
+use crate::analysis::graph::{Edge, Node};
+use crate::analysis::interprocedural_fixpoint::Computation;
+use crate::analysis::interprocedural_fixpoint::Context as _;
+use crate::analysis::interprocedural_fixpoint::NodeValue;
+use crate::intermediate_representation::*;
+use crate::prelude::*;
+use crate::utils::log::{CweWarning, LogMessage};
+use crate::CweModule;
+use petgraph::visit::EdgeRef;
+use std::collections::HashMap;
+
+mod state;
+use state::*;
+
+mod taint;
+use taint::*;
+
+mod context;
+use context::*;
+
+pub static CWE_MODULE: CweModule = CweModule {
+    name: "CWE476",
+    version: "0.3",
+    run: check_cwe,
+};
+
+/// The configuration struct
+#[derive(Serialize, Deserialize, Debug, PartialEq, Eq, Hash, Clone)]
+pub struct Config {
+    /// The names of symbols for which the analysis should check
+    /// whether the return values are checked for being a Null pointer by the analysed binary.
+    symbols: Vec<String>,
+}
+
+/// Run the CWE check.
+/// We check whether the return values of symbols configurable in the config file are being checked for Null pointers
+/// before any memory access (and thus potential Null pointer dereferences) through these values happen.
+pub fn check_cwe(
+    analysis_results: &AnalysisResults,
+    cwe_params: &serde_json::Value,
+) -> (Vec<LogMessage>, Vec<CweWarning>) {
+    let project = analysis_results.project;
+    let pointer_inference_results = analysis_results.pointer_inference.unwrap();
+
+    let (cwe_sender, cwe_receiver) = crossbeam_channel::unbounded();
+
+    let config: Config = serde_json::from_value(cwe_params.clone()).unwrap();
+    let symbol_map = crate::utils::symbol_utils::get_symbol_map(project, &config.symbols[..]);
+    let general_context = Context::new(project, &pointer_inference_results, cwe_sender);
+
+    for edge in general_context.get_graph().edge_references() {
+        if let Edge::ExternCallStub(jmp) = edge.weight() {
+            if let Jmp::Call { target, .. } = &jmp.term {
+                if let Some(symbol) = symbol_map.get(target) {
+                    let node = edge.target();
+                    let current_sub = match general_context.get_graph()[node] {
+                        Node::BlkStart(_blk, sub) => sub,
+                        _ => panic!(),
+                    };
+                    let mut context = general_context.clone();
+                    context.set_taint_source(jmp, current_sub);
+                    let pi_state_at_taint_source =
+                        match pointer_inference_results.get_node_value(node) {
+                            Some(NodeValue::Value(val)) => Some(val.clone()),
+                            _ => None,
+                        };
+                    let mut computation = Computation::new(context, None);
+                    computation.set_node_value(
+                        node,
+                        NodeValue::Value(State::new(
+                            symbol,
+                            &project.stack_pointer_register,
+                            pi_state_at_taint_source.as_ref(),
+                        )),
+                    );
+                    computation.compute_with_max_steps(100);
+                }
+            }
+        }
+    }
+
+    let mut cwe_warnings = HashMap::new();
+    for cwe in cwe_receiver.try_iter() {
+        match &cwe.addresses[..] {
+            [taint_source_address, ..] => cwe_warnings.insert(taint_source_address.clone(), cwe),
+            _ => panic!(),
+        };
+    }
+    let cwe_warnings = cwe_warnings.into_iter().map(|(_, cwe)| cwe).collect();
+
+    (Vec::new(), cwe_warnings)
+}

cwe_checker_rs/src/checkers/cwe_476/taint.rs

+use crate::abstract_domain::{AbstractDomain, HasByteSize, HasTop, RegisterDomain};
+use crate::intermediate_representation::*;
+use crate::prelude::*;
+use std::fmt::Display;
+
+/// An abstract domain representing a value that is either tainted or not.
+///
+/// Note that the [merge](Taint::merge)-function does not respect the partial order
+/// that is implied by the naming scheme of the variants!
+/// In fact the whole analysis does not enforce any partial order for this domain.
+/// This means that in theory the fixpoint computation may not actually converge to a fixpoint,
+/// but in practice the analysis can make more precise decisions
+/// whether a value should be tainted or not.
+#[derive(Serialize, Deserialize, Debug, PartialEq, Eq, Hash, Clone, Copy)]
+pub enum Taint {
+    /// A tainted value of a particular bytesize.
+    Tainted(ByteSize),
+    /// An untainted value of a particular bytesize
+    Top(ByteSize),
+}
+
+impl Display for Taint {
+    /// Print the value of a `Taint` object.
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self {
+            Self::Tainted(size) => write!(f, "Tainted:{}", size),
+            Self::Top(size) => write!(f, "Top:{}", size),
+        }
+    }
+}
+
+impl AbstractDomain for Taint {
+    /// The result of merging two `Taint` values is tainted if at least one input was tainted.
+    fn merge(&self, other: &Self) -> Self {
+        use Taint::*;
+        match (self, other) {
+            (Tainted(size), _) | (_, Tainted(size)) => Tainted(*size),
+            _ => Top(self.bytesize()),
+        }
+    }
+
+    /// Checks whether the value is an untainted `Top`-value.
+    fn is_top(&self) -> bool {
+        matches!(self, Taint::Top(_))
+    }
+}
+
+impl HasByteSize for Taint {
+    /// The size in bytes of the `Taint` value.
+    fn bytesize(&self) -> ByteSize {
+        match self {
+            Self::Tainted(size) | Self::Top(size) => *size,
+        }
+    }
+}
+
+impl HasTop for Taint {
+    /// Get a new `Top`-value with the same bytesize as `self`.
+    fn top(&self) -> Self {
+        Self::Top(self.bytesize())
+    }
+}
+
+impl RegisterDomain for Taint {
+    /// Get a new `Top`-value with the given bytesize.
+    fn new_top(bytesize: ByteSize) -> Self {
+        Self::Top(bytesize)
+    }
+
+    /// The result of a binary operation is tainted if at least one input value was tainted.
+    fn bin_op(&self, op: BinOpType, rhs: &Self) -> Self {
+        match (self, rhs) {
+            (Self::Tainted(_), _) | (_, Self::Tainted(_)) => {
+                Self::Tainted(self.bin_op_bytesize(op, rhs))
+            }
+            _ => Self::Top(self.bin_op_bytesize(op, rhs)),
+        }
+    }
+
+    /// The result of a unary operation is tainted if the input was tainted.
+    fn un_op(&self, _op: UnOpType) -> Self {
+        *self
+    }
+
+    /// A subpiece of a tainted value is again tainted.
+    fn subpiece(&self, _low_byte: ByteSize, size: ByteSize) -> Self {
+        if let Self::Tainted(_) = self {
+            Self::Tainted(size)
+        } else {
+            Self::Top(size)
+        }
+    }
+
+    /// The result of a cast operation is tainted if the input was tainted.
+    fn cast(&self, _kind: CastOpType, width: ByteSize) -> Self {
+        if let Self::Tainted(_) = self {
+            Self::Tainted(width)
+        } else {
+            Self::Top(width)
+        }
+    }
+}
+
+impl Taint {
+    /// Checks whether the given value is in fact tainted.
+    pub fn is_tainted(&self) -> bool {
+        matches!(self, Taint::Tainted(_))
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn abstract_domain() {
+        let taint = Taint::Tainted(ByteSize::new(4));
+        let top = Taint::Top(ByteSize::new(4));
+        assert_eq!(taint.merge(&top), taint);
+        assert_eq!(top.merge(&top), top);
+        assert_eq!(taint.is_top(), false);
+    }
+
+    #[test]
+    fn register_domain() {
+        use crate::intermediate_representation::*;
+        let taint = Taint::Tainted(ByteSize::new(4));
+        let top = Taint::Top(ByteSize::new(4));
+        assert_eq!(taint.bin_op(BinOpType::IntAdd, &top), taint);
+        assert_eq!(top.bin_op(BinOpType::IntMult, &top), top);
+        assert_eq!(taint.un_op(UnOpType::FloatFloor), taint);
+        assert_eq!(taint.subpiece(ByteSize::new(0), ByteSize::new(4)), taint);
+        assert_eq!(top.cast(CastOpType::IntZExt, ByteSize::new(4)), top);
+    }
+}

cwe_checker_rs/src/checkers/cwe_560.rs

@@ -98,9 +98,10 @@ fn generate_cwe_warning(sub: &Term<Sub>, jmp: &Term<Jmp>, permission_const: u64)
 ///
 /// Only the basic block right before the umask call is evaluated when trying to determine the parameter value of umask.
 pub fn check_cwe(
-    project: &Project,
+    analysis_results: &AnalysisResults,
     _cwe_params: &serde_json::Value,
 ) -> (Vec<LogMessage>, Vec<CweWarning>) {
+    let project = analysis_results.project;
     let mut cwes = Vec::new();
     let mut log_messages = Vec::new();
     let umask_symbol_map = get_symbol_map(project, &["umask".to_string()]);

cwe_checker_rs/src/checkers/cwe_676.rs

@@ -18,10 +18,11 @@ False Negatives
 
 * None known
 */
+use crate::prelude::*;
 use std::collections::HashMap;
 
 use crate::{
-    intermediate_representation::{ExternSymbol, Program, Project, Sub, Term, Tid},
+    intermediate_representation::{ExternSymbol, Program, Sub, Term, Tid},
     utils::{
         log::{CweWarning, LogMessage},
         symbol_utils::get_calls_to_symbols,
@@ -106,9 +107,10 @@ pub fn resolve_symbols<'a>(
 }
 
 pub fn check_cwe(
-    project: &Project,
+    analysis_results: &AnalysisResults,
     cwe_params: &serde_json::Value,
 ) -> (Vec<LogMessage>, Vec<CweWarning>) {
+    let project = analysis_results.project;
     let config: Config = serde_json::from_value(cwe_params.clone()).unwrap();
     let prog: &Term<Program> = &project.program;
     let subfunctions: &Vec<Term<Sub>> = &prog.term.subs;

cwe_checker_rs/src/checkers/cwe_782.rs

@@ -14,10 +14,11 @@ False Negatives:
 
 * There are other ways to expose I/O control without access control.
 */
+use crate::prelude::*;
 use std::collections::HashMap;
 
 use crate::{
-    intermediate_representation::{Program, Project, Sub, Term, Tid},
+    intermediate_representation::{Program, Sub, Term, Tid},
     utils::{
         log::{CweWarning, LogMessage},
         symbol_utils::{find_symbol, get_calls_to_symbols},
@@ -66,9 +67,10 @@ pub fn generate_cwe_warning(calls: &[(&str, &Tid, &str)]) -> Vec<CweWarning> {
 }
 
 pub fn check_cwe(
-    project: &Project,
+    analysis_results: &AnalysisResults,
     _cwe_params: &serde_json::Value,
 ) -> (Vec<LogMessage>, Vec<CweWarning>) {
+    let project = analysis_results.project;
     let prog: &Term<Program> = &project.program;
     let mut warnings: Vec<CweWarning> = Vec::new();
     if let Some((tid, name)) = find_symbol(prog, "ioctl") {

cwe_checker_rs/src/ffi/analysis.rs

@@ -17,7 +17,8 @@ fn run_pointer_inference(program_jsonbuilder_val: ocaml::Value) -> (Vec<CweWarni
     let config: crate::analysis::pointer_inference::Config =
         serde_json::from_value(crate::utils::read_config_file("config.json")["Memory"].clone())
             .unwrap();
-    let (mut logs, cwes) = crate::analysis::pointer_inference::run(&project, config, false);
+    let pi_analysis = crate::analysis::pointer_inference::run(&project, config, false);
+    let (mut logs, cwes) = pi_analysis.collected_logs;
     all_logs.append(&mut logs);
     (
         cwes,

cwe_checker_rs/src/intermediate_representation/term.rs

@@ -331,6 +331,13 @@ impl Project {
     pub fn get_pointer_bytesize(&self) -> ByteSize {
         self.stack_pointer_register.size
     }
+
+    /// Try to guess a standard calling convention from the list of calling conventions in the project.
+    pub fn get_standard_calling_convention(&self) -> Option<&CallingConvention> {
+        self.calling_conventions
+            .iter()
+            .find(|cconv| cconv.name == "__stdcall")
+    }
 }
 
 impl Project {
@@ -437,6 +444,54 @@ impl Project {
 mod tests {
     use super::*;
 
+    impl Blk {
+        pub fn mock() -> Term<Blk> {
+            Term {
+                tid: Tid::new("block"),
+                term: Blk {
+                    defs: Vec::new(),
+                    jmps: Vec::new(),
+                },
+            }
+        }
+    }
+
+    impl Sub {
+        pub fn mock(name: impl ToString) -> Term<Sub> {
+            Term {
+                tid: Tid::new(name.to_string()),
+                term: Sub {
+                    name: name.to_string(),
+                    blocks: Vec::new(),
+                },
+            }
+        }
+    }
+
+    impl Program {
+        pub fn mock_empty() -> Program {
+            Program {
+                subs: Vec::new(),
+                extern_symbols: Vec::new(),
+                entry_points: Vec::new(),
+            }
+        }
+    }
+
+    impl Project {
+        pub fn mock_empty() -> Project {
+            Project {
+                program: Term {
+                    tid: Tid::new("program_tid"),
+                    term: Program::mock_empty(),
+                },
+                cpu_architecture: "x86_64".to_string(),
+                stack_pointer_register: Variable::mock("RSP", 8u64),
+                calling_conventions: Vec::new(),
+            }
+        }
+    }
+
     #[test]
     fn retarget_nonexisting_jumps() {
         let mut jmp_term = Term {

cwe_checker_rs/src/intermediate_representation/variable.rs

@@ -15,3 +15,18 @@ pub struct Variable {
     pub size: ByteSize,
     pub is_temp: bool,
 }
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    impl Variable {
+        pub fn mock(name: impl ToString, size_in_bytes: impl Into<ByteSize>) -> Variable {
+            Variable {
+                name: name.to_string(),
+                size: size_in_bytes.into(),
+                is_temp: false,
+            }
+        }
+    }
+}

cwe_checker_rs/src/lib.rs

@@ -7,6 +7,7 @@ Parts of the cwe_checker that are written in Rust.
 #[macro_use]
 extern crate ocaml;
 
+use crate::analysis::pointer_inference::PointerInference;
 use crate::intermediate_representation::Project;
 use crate::utils::log::{CweWarning, LogMessage};
 
@@ -27,11 +28,13 @@ mod prelude {
     pub use crate::bil::{BitSize, Bitvector};
     pub use crate::intermediate_representation::ByteSize;
     pub use crate::intermediate_representation::{Term, Tid};
+    pub use crate::AnalysisResults;
     pub use anyhow::{anyhow, Error};
 }
 
 /// The generic function signature for the main function of a CWE module
-pub type CweModuleFn = fn(&Project, &serde_json::Value) -> (Vec<LogMessage>, Vec<CweWarning>);
+pub type CweModuleFn =
+    fn(&AnalysisResults, &serde_json::Value) -> (Vec<LogMessage>, Vec<CweWarning>);
 
 /// A structure containing general information about a CWE analysis module,
 /// including the function to be called to run the analysis.
@@ -55,9 +58,51 @@ pub fn get_modules() -> Vec<&'static CweModule> {
         &crate::checkers::cwe_332::CWE_MODULE,
         &crate::checkers::cwe_426::CWE_MODULE,
         &crate::checkers::cwe_467::CWE_MODULE,
+        &crate::checkers::cwe_476::CWE_MODULE,
         &crate::checkers::cwe_560::CWE_MODULE,
-        &crate::checkers::cwe_782::CWE_MODULE,
         &crate::checkers::cwe_676::CWE_MODULE,
+        &crate::checkers::cwe_782::CWE_MODULE,
         &crate::analysis::pointer_inference::CWE_MODULE,
     ]
 }
+
+/// A struct containing pointers to all known analysis results
+/// that may be needed as input for other analyses and CWE checks.
+#[derive(Clone, Copy)]
+pub struct AnalysisResults<'a> {
+    /// A pointer to the project struct
+    pub project: &'a Project,
+    /// The result of the pointer inference analysis if already computed.
+    pub pointer_inference: Option<&'a PointerInference<'a>>,
+}
+
+impl<'a> AnalysisResults<'a> {
+    /// Create a new `AnalysisResults` struct with only the project itself known.
+    pub fn new(project: &'a Project) -> AnalysisResults<'a> {
+        AnalysisResults {
+            project,
+            pointer_inference: None,
+        }
+    }
+
+    /// Compute the pointer inference analysis.
+    /// The result gets returned, but not saved to the `AnalysisResults` struct itself.
+    pub fn compute_pointer_inference(&self, config: &serde_json::Value) -> PointerInference<'a> {
+        crate::analysis::pointer_inference::run(
+            self.project,
+            serde_json::from_value(config.clone()).unwrap(),
+            false,
+        )
+    }
+
+    /// Create a new `AnalysisResults` struct containing the given pointer inference analysis results.
+    pub fn set_pointer_inference<'b: 'a>(
+        self,
+        pi_results: Option<&'b PointerInference<'a>>,
+    ) -> AnalysisResults<'b> {
+        AnalysisResults {
+            pointer_inference: pi_results,
+            ..self
+        }
+    }
+}

cwe_checker_rs/src/utils/log.rs

 use crate::prelude::*;
+use std::thread::JoinHandle;
 
 /// A CWE warning message.
 #[derive(Serialize, Deserialize, Debug, PartialEq, Eq, Hash, Clone, PartialOrd, Ord, Default)]
@@ -180,3 +181,94 @@ pub fn print_all_messages(
         print!("{}", output);
     }
 }
+
+/// The message types a logging thread can receive.
+/// See the [`LogThread`] type for more information.
+#[derive(Serialize, Deserialize, Debug, PartialEq, Eq, Hash, Clone, PartialOrd, Ord)]
+pub enum LogThreadMsg {
+    /// A normal log message.
+    Log(LogMessage),
+    /// A CWE warning
+    Cwe(CweWarning),
+    /// If the log collector thread receives this signal,
+    /// it should stop receiving new messages
+    /// and instead terminate and return the collected messages prior to receiving the termination signal.
+    Terminate,
+}
+
+/// A type for managing threads for collecting log messages.
+///
+/// With [`LogThread::spawn()`] one can create a new log thread
+/// whose handle is contained in the returned `LogThread` struct.
+/// By calling the [`collect()`](LogThread::collect()) method
+/// one can tell the log thread to shut down
+/// and return the logs collected to this point.
+/// If the `LogThread` object gets dropped before calling `collect()`,
+/// the corresponding logging thread will be stopped
+/// and all collected logs will be discarded.
+///
+/// If one deliberately wants to discard all logging messages,
+/// one can simply create a sender to a disconnected channel
+/// via [`LogThread::create_disconnected_sender()`].
+pub struct LogThread {
+    msg_sender: crossbeam_channel::Sender<LogThreadMsg>,
+    thread_handle: Option<JoinHandle<(Vec<LogMessage>, Vec<CweWarning>)>>,
+}
+
+impl Drop for LogThread {
+    /// If the logging thread still exists,
+    /// send it the `Terminate` signal.
+    /// Then wait until the logging thread stopped.
+    fn drop(&mut self) {
+        // Make sure the logging thread gets terminated when dropping this.
+        let _ = self.msg_sender.send(LogThreadMsg::Terminate);
+        if let Some(handle) = self.thread_handle.take() {
+            let _ = handle.join();
+        }
+    }
+}
+
+impl LogThread {
+    /// Create a new `LogThread` object with a handle to a freshly spawned logging collector thread.
+    ///
+    /// The parameter is the function containing the actual log collection logic.
+    /// I.e. the function should receive messages through the given receiver until the channel disconnects
+    /// or until it receives a [`LogThread::Terminate`] message.
+    /// After that it should return the logs collected up to that point.
+    pub fn spawn<F>(collector_func: F) -> LogThread
+    where
+        F: FnOnce(crossbeam_channel::Receiver<LogThreadMsg>) -> (Vec<LogMessage>, Vec<CweWarning>)
+            + Send
+            + 'static,
+    {
+        let (sender, receiver) = crossbeam_channel::unbounded();
+        let thread_handle = std::thread::spawn(move || collector_func(receiver));
+        LogThread {
+            msg_sender: sender,
+            thread_handle: Some(thread_handle),
+        }
+    }
+
+    /// Just create a disconnected sender to a (non-existing) logging thread.
+    /// Can be used like a sender to a channel that deliberately discards all messages sent to it.
+    pub fn create_disconnected_sender() -> crossbeam_channel::Sender<LogThreadMsg> {
+        let (sender, _) = crossbeam_channel::unbounded();
+        sender
+    }
+
+    /// Get a sender that can be used to send messages to the logging thread corresponding to this `LogThread` instance.
+    pub fn get_msg_sender(&self) -> crossbeam_channel::Sender<LogThreadMsg> {
+        self.msg_sender.clone()
+    }
+
+    /// Stop the logging thread by sending it the `Terminate` signal
+    /// and then return all logs collected until that point.
+    pub fn collect(mut self) -> (Vec<LogMessage>, Vec<CweWarning>) {
+        let _ = self.msg_sender.send(LogThreadMsg::Terminate);
+        if let Some(handle) = self.thread_handle.take() {
+            handle.join().unwrap()
+        } else {
+            (Vec::new(), Vec::new())
+        }
+    }
+}

test/src/lib.rs

@@ -363,6 +363,35 @@ mod tests {
 
     #[test]
     #[ignore]
+    fn cwe_476() {
+        let mut error_log = Vec::new();
+        let mut tests = all_test_cases("cwe_476", "CWE476");
+
+        // TODO: Check reason for failure!
+        mark_skipped(&mut tests, "mips64", "gcc");
+        mark_skipped(&mut tests, "mips64el", "gcc");
+        mark_skipped(&mut tests, "mips", "gcc");
+        mark_skipped(&mut tests, "mipsel", "gcc");
+
+        mark_architecture_skipped(&mut tests, "ppc64"); // Ghidra generates mangled function names here for some reason.
+        mark_architecture_skipped(&mut tests, "ppc64le"); // Ghidra generates mangled function names here for some reason.
+
+        mark_compiler_skipped(&mut tests, "mingw32-gcc"); // TODO: Check reason for failure!
+
+        for test_case in tests {
+            let num_expected_occurences = 1;
+            if let Err(error) = test_case.run_test("[CWE476]", num_expected_occurences) {
+                error_log.push((test_case.get_filepath(), error));
+            }
+        }
+        if !error_log.is_empty() {
+            print_errors(error_log);
+            panic!();
+        }
+    }
+
+    #[test]
+    #[ignore]
     fn cwe_560() {
         let mut error_log = Vec::new();
         let mut tests = linux_test_cases("cwe_560", "CWE560");