add ability to trace location information in concolic tracer (#322)

* add ability to trace location information in concolic tracer

* fix formatting

* introduce location new-type

* fix conolic smoke test

* impl From instead of custom into_inner function

* fmt

* change to use usize instead of NonZeroUsize

in order to no over-constrain the implementation

fuzzers/libfuzzer_stb_image_concolic/runtime/src/lib.rs

@@ -15,7 +15,7 @@ export_runtime!(
     CallStackCoverage::default() => CallStackCoverage; // QSym-style expression pruning
     tracing::TracingRuntime::new(
         StdShMemMessageFileWriter::from_stdshmem_default_env()
-            .expect("unable to construct tracing runtime writer. (missing env?)")
-        )
-        => tracing::TracingRuntime
+            .expect("unable to construct tracing runtime writer. (missing env?)"),
+        false
+    ) => tracing::TracingRuntime
 );

libafl/src/observers/concolic/mod.rs

@@ -1,5 +1,8 @@
 //! # Concolic Tracing
-use core::num::NonZeroUsize;
+use core::{
+    fmt::{Debug, Display, Error, Formatter},
+    num::NonZeroUsize,
+};
 
 #[cfg(feature = "std")]
 use serde::{Deserialize, Serialize};
@@ -12,6 +15,39 @@ use serde::{Deserialize, Serialize};
 /// `SymExprRef`s are not valid across traces.
 pub type SymExprRef = NonZeroUsize;
 
+/// [`Location`]s are code locations encountered during concolic tracing, that are constructed from pointers, but not always in a meaningful way.
+/// Therefore, a location is an opague value that can only be compared against itself.
+///
+/// It is possible to get at the underlying value using [`Into::into`], should this restriction be too inflexible for your usecase.
+#[cfg_attr(feature = "std", derive(Serialize, Deserialize))]
+#[derive(Copy, Clone, PartialEq, Eq, Hash, PartialOrd, Ord)]
+#[repr(transparent)]
+pub struct Location(usize);
+
+impl Debug for Location {
+    fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error> {
+        core::fmt::Debug::fmt(&self.0, f)
+    }
+}
+
+impl Display for Location {
+    fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error> {
+        core::fmt::Display::fmt(&self.0, f)
+    }
+}
+
+impl From<Location> for usize {
+    fn from(l: Location) -> Self {
+        l.0
+    }
+}
+
+impl From<usize> for Location {
+    fn from(v: usize) -> Self {
+        Self(v)
+    }
+}
+
 /// `SymExpr` represents a message in the serialization format.
 /// The messages in the format are a perfect mirror of the methods that are called on the runtime during execution.
 #[cfg(feature = "std")]
@@ -297,13 +333,26 @@ pub enum SymExpr {
     PathConstraint {
         constraint: SymExprRef,
         taken: bool,
-        site_id: usize,
+        location: Location,
     },
 
     /// These expressions won't be referenced again
     ExpressionsUnreachable {
         exprs: Vec<SymExprRef>,
     },
+
+    /// Location information regarding a call. Tracing this information is optional.
+    Call {
+        location: Location,
+    },
+    /// Location information regarding a return. Tracing this information is optional.
+    Return {
+        location: Location,
+    },
+    /// Location information regarding a basic block. Tracing this information is optional.
+    BasicBlock {
+        location: Location,
+    },
 }
 
 #[cfg(feature = "std")]

libafl/src/observers/concolic/serialization_format.rs

@@ -198,6 +198,7 @@ impl<R: Read> MessageFileReader<R> {
                     *expr = self.make_absolute(*expr);
                 }
             }
+            SymExpr::Call { .. } | SymExpr::Return { .. } | SymExpr::BasicBlock { .. } => {}
         }
         SymExprRef::new(ret).unwrap()
     }
@@ -348,6 +349,7 @@ impl<W: Write + Seek> MessageFileWriter<W> {
                     *expr = self.make_relative(*expr);
                 }
             }
+            SymExpr::Call { .. } | SymExpr::Return { .. } | SymExpr::BasicBlock { .. } => {}
         }
         self.serialization_options
             .serialize_into(&mut self.writer, &message)?;

libafl/src/stages/concolic.rs

@@ -280,9 +280,7 @@ fn generate_mutations(iter: impl Iterator<Item = (SymExprRef, SymExpr)>) -> Vec<
         if let Some(expr) = z3_expr {
             translation.insert(id, expr);
         } else if let SymExpr::PathConstraint {
-            constraint,
-            site_id: _,
-            taken,
+            constraint, taken, ..
         } = msg
         {
             let op = translation[&constraint].as_bool().unwrap();

libafl_concolic/symcc_runtime/src/lib.rs

@@ -253,7 +253,7 @@ where
 /// ```no_run
 /// # #[macro_use] extern crate symcc_runtime;
 /// # use symcc_runtime::{tracing::TracingRuntime, Runtime};
-/// export_runtime!(TracingRuntime::new(todo!()) => TracingRuntime);
+/// export_runtime!(TracingRuntime::new(todo!(), todo!()) => TracingRuntime);
 /// ```
 ///
 /// ## Runtime composition using `Filter`s
@@ -267,7 +267,7 @@ where
 /// ```no_run
 /// # #[macro_use] extern crate symcc_runtime;
 /// # use symcc_runtime::{tracing::TracingRuntime, Runtime, filter::NoFloat};
-/// export_runtime!(NoFloat => NoFloat; TracingRuntime::new(todo!()) => TracingRuntime);
+/// export_runtime!(NoFloat => NoFloat; TracingRuntime::new(todo!(), todo!()) => TracingRuntime);
 /// ```
 /// This will construct a runtime that is first filtered by [`filter::NoFloat`] and then traced by the tracing runtime.
 ///

libafl_concolic/symcc_runtime/src/tracing.rs

@@ -1,5 +1,7 @@
 //! Tracing of expressions in a serialized form.
 
+use std::num::NonZeroUsize;
+
 pub use libafl::observers::concolic::serialization_format::StdShMemMessageFileWriter;
 use libafl::observers::concolic::SymExpr;
 
@@ -9,13 +11,19 @@ use crate::{RSymExpr, Runtime};
 /// The format can be read from elsewhere to perform processing of the expressions outside of the runtime.
 pub struct TracingRuntime {
     writer: StdShMemMessageFileWriter,
+    trace_locations: bool,
 }
 
 impl TracingRuntime {
     /// Creates the runtime, tracing using the given writer.
+    /// When `trace_locations` is true, location information for calls, returns and basic blocks will also be part of the trace.
+    /// Tracing location information can drastically increase trace size. It is therefore recommended to not active this if not needed.
     #[must_use]
-    pub fn new(writer: StdShMemMessageFileWriter) -> Self {
-        Self { writer }
+    pub fn new(writer: StdShMemMessageFileWriter, trace_locations: bool) -> Self {
+        Self {
+            writer,
+            trace_locations,
+        }
     }
 
     #[allow(clippy::unnecessary_wraps)]
@@ -143,11 +151,29 @@ impl Runtime for TracingRuntime {
     binary_expression_builder!(concat_helper, Concat);
     expression_builder!(extract_helper(op: RSymExpr, first_bit:usize, last_bit:usize) => Extract);
 
-    fn notify_call(&mut self, _site_id: usize) {}
+    fn notify_call(&mut self, site_id: usize) {
+        if self.trace_locations {
+            self.write_message(SymExpr::Call {
+                location: site_id.into(),
+            });
+        }
+    }
 
-    fn notify_ret(&mut self, _site_id: usize) {}
+    fn notify_ret(&mut self, site_id: usize) {
+        if self.trace_locations {
+            self.write_message(SymExpr::Return {
+                location: site_id.into(),
+            });
+        }
+    }
 
-    fn notify_basic_block(&mut self, _site_id: usize) {}
+    fn notify_basic_block(&mut self, site_id: usize) {
+        if self.trace_locations {
+            self.write_message(SymExpr::BasicBlock {
+                location: site_id.into(),
+            });
+        }
+    }
 
     fn expression_unreachable(&mut self, exprs: &[RSymExpr]) {
         self.write_message(SymExpr::ExpressionsUnreachable {
@@ -159,7 +185,7 @@ impl Runtime for TracingRuntime {
         self.write_message(SymExpr::PathConstraint {
             constraint,
             taken,
-            site_id,
+            location: site_id.into(),
         });
     }
 }

libafl_concolic/test/expected_constraints.txt

@@ -19,7 +19,7 @@
 19	Mul { a: 18, b: 17 }
 20	Integer { value: 7, bits: 32 }
 21	SignedLessThan { a: 19, b: 20 }
-22	PathConstraint { constraint: 21, taken: false, site_id: 11229456 }
+22	PathConstraint { constraint: 21, taken: false, location: 11229456 }
 22	Concat { a: 12, b: 11 }
 23	Concat { a: 13, b: 22 }
 24	Concat { a: 14, b: 23 }
@@ -27,4 +27,4 @@
 26	SignedRem { a: 24, b: 25 }
 27	Integer { value: 0, bits: 32 }
 28	NotEqual { a: 26, b: 27 }
-29	PathConstraint { constraint: 28, taken: true, site_id: 11122032 }
+29	PathConstraint { constraint: 28, taken: true, location: 11122032 }

libafl_concolic/test/runtime_test/src/lib.rs

@@ -13,7 +13,8 @@ export_runtime!(
     NoFloat => NoFloat;
     tracing::TracingRuntime::new(
         StdShMemMessageFileWriter::from_stdshmem_default_env()
-            .expect("unable to construct tracing runtime writer. (missing env?)")
-        )
-        => tracing::TracingRuntime
+            .expect("unable to construct tracing runtime writer. (missing env?)"),
+        false
+    )
+    => tracing::TracingRuntime
 );

libafl_concolic/test/smoke_test.sh

@@ -46,7 +46,7 @@ echo "constraints: "
 cat constraints.txt
 
 # site_id's in the constraints trace will differ for every run. we therefore filter those.
-sed 's/, site_id: .* / /' < constraints.txt > constraints_filtered.txt
-sed 's/, site_id: .* / /' < expected_constraints.txt > expected_constraints_filtered.txt
+sed 's/, location: .* / /' < constraints.txt > constraints_filtered.txt
+sed 's/, location: .* / /' < expected_constraints.txt > expected_constraints_filtered.txt
 
 diff constraints_filtered.txt expected_constraints_filtered.txt