Construct PDG exclusively on provenance information

analysis/runtime/Cargo.toml

@@ -19,3 +19,4 @@ enum_dispatch = "0.3"
 fs-err = "2"
 crossbeam-queue = "0.3"
 crossbeam-utils = "0.8"
+indexmap = { version = "1.9", features = ["serde"] }

analysis/runtime/src/events.rs

@@ -30,10 +30,12 @@ pub enum EventKind {
 
     CopyRef,
 
-    /// Field projection. Used for operations like `_2 = &(*_1).0`. Nested field
-    /// accesses like `_4 = &(*_1).x.y.z` are broken into multiple `Node`s, each
-    /// covering one level.
-    Field(Pointer, u32),
+    /// Projection. Used for operations like `_2 = &(*_1).0`.
+    /// The third value is a "projection index" that points to an element
+    /// of the projections data structure in the metadata. It is used to
+    /// disambiguate between different projections with the same pointer,
+    /// e.g., `(*p).x` and `(*p).x.a` where `a` is at offset 0.
+    Project(Pointer, Pointer, usize),
 
     Alloc {
         size: usize,
@@ -90,7 +92,9 @@ impl Debug for EventKind {
         use EventKind::*;
         match *self {
             CopyPtr(ptr) => write!(f, "copy(0x{:x})", ptr),
-            Field(ptr, id) => write!(f, "field(0x{:x}, {})", ptr, id),
+            Project(ptr, new_ptr, idx) => {
+                write!(f, "project(0x{:x}, 0x{:x}, [{}])", ptr, new_ptr, idx)
+            }
             Alloc { size, ptr } => {
                 write!(f, "malloc({}) -> 0x{:x}", size, ptr)
             }

analysis/runtime/src/handlers.rs

@@ -110,10 +110,10 @@ pub const HOOK_FUNCTIONS: &[&str] = &[
     hook_fn!(offset),
 ];
 
-pub fn ptr_field(mir_loc: MirLocId, ptr: usize, field_id: u32) {
+pub fn ptr_project(mir_loc: MirLocId, ptr: usize, new_ptr: usize, proj_idx: usize) {
     RUNTIME.send_event(Event {
         mir_loc,
-        kind: EventKind::Field(ptr, field_id),
+        kind: EventKind::Project(ptr, new_ptr, proj_idx),
     });
 }
 

analysis/runtime/src/metadata.rs

@@ -1,3 +1,4 @@
+use indexmap::IndexSet;
 use std::{
     collections::HashMap,
     fmt::{self, Debug, Formatter},
@@ -13,6 +14,7 @@ use crate::mir_loc::{Func, FuncId, MirLoc, MirLocId};
 pub struct Metadata {
     pub locs: Vec<MirLoc>,
     pub functions: HashMap<FuncId, String>,
+    pub projections: IndexSet<Vec<usize>>,
 }
 
 impl Metadata {
@@ -46,11 +48,17 @@ impl FromIterator<Metadata> for Metadata {
     fn from_iter<I: IntoIterator<Item = Metadata>>(iter: I) -> Self {
         let mut locs = Vec::new();
         let mut functions = HashMap::new();
+        let mut projections = IndexSet::new();
         for metadata in iter {
             locs.extend(metadata.locs);
             functions.extend(metadata.functions);
+            projections.extend(metadata.projections);
+        }
+        Self {
+            locs,
+            functions,
+            projections,
         }
-        Self { locs, functions }
     }
 }