[clang] [analyzer] Do not reason about locations passed as inline asm input (PR #103714)

[via cfe-commits]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-clang

Author: Pavel Skripkin (pskrgag)

<details>
<summary>Changes</summary>

If pointer is passed as input operand for inline assembly, it's possible that asm block will change memory behind this pointer. So if pointer is passed inside inline asm block, it's better to not guess and assume memory has unknown state.

Without such change, we observed a lot of FP with hand-written `memcpy` and friends.

---
Full diff: https://github.com/llvm/llvm-project/pull/103714.diff


2 Files Affected:

- (modified) clang/lib/StaticAnalyzer/Core/ExprEngine.cpp (+8) 
- (modified) clang/test/Analysis/asm.cpp (+35-1) 


``````````diff
diff --git a/clang/lib/StaticAnalyzer/Core/ExprEngine.cpp b/clang/lib/StaticAnalyzer/Core/ExprEngine.cpp
index 686310d38ebd58..d64134102d96c7 100644
--- a/clang/lib/StaticAnalyzer/Core/ExprEngine.cpp
+++ b/clang/lib/StaticAnalyzer/Core/ExprEngine.cpp
@@ -3807,6 +3807,14 @@ void ExprEngine::VisitGCCAsmStmt(const GCCAsmStmt *A, ExplodedNode *Pred,
       state = state->bindLoc(*LV, UnknownVal(), Pred->getLocationContext());
   }
 
+  // Do not reason about locations passed inside inline assembly.
+  for (const Expr *O : A->inputs()) {
+    SVal X = state->getSVal(O, Pred->getLocationContext());
+
+    if (std::optional<Loc> LV = X.getAs<Loc>())
+      state = state->bindLoc(*LV, UnknownVal(), Pred->getLocationContext());
+  }
+
   Bldr.generateNode(A, Pred, state);
 }
 
diff --git a/clang/test/Analysis/asm.cpp b/clang/test/Analysis/asm.cpp
index 1180063502168f..5d158b62e7221f 100644
--- a/clang/test/Analysis/asm.cpp
+++ b/clang/test/Analysis/asm.cpp
@@ -1,4 +1,5 @@
-// RUN: %clang_analyze_cc1 -analyzer-checker debug.ExprInspection -fheinous-gnu-extensions -w %s -verify
+// RUN: %clang_analyze_cc1 -triple=x86_64-unknown-unknown \
+// RUN:      -analyzer-checker debug.ExprInspection,core -fheinous-gnu-extensions -w %s -verify
 
 int clang_analyzer_eval(int);
 
@@ -10,3 +11,36 @@ void testRValueOutput() {
   clang_analyzer_eval(global == 1); // expected-warning{{UNKNOWN}}
   clang_analyzer_eval(ref == 1);    // expected-warning{{UNKNOWN}}
 }
+
+void *MyMemcpy(void *d, const void *s, const int n) {
+  asm volatile (
+    "cld\n rep movsb\n"
+    :: "S" (s), "D" (d), "c" (n) : "memory"
+  );
+
+  return d;
+}
+
+void testInlineAsmMemcpy(void)
+{
+    int a, b = 10, c;
+
+    MyMemcpy(&a, &b, sizeof(b));
+    c = a; // no-warning
+}
+
+void testInlineAsmMemcpyArray(void)
+{
+    int a[10], b[10] = {}, c;
+
+    MyMemcpy(&a, &b, sizeof(b));
+    c = a[8]; // no-warning
+}
+
+void testInlineAsmMemcpyUninit(void)
+{
+    int a[10], b[10] = {}, c;
+
+    MyMemcpy(&a[1], &b[1], sizeof(b) - sizeof(b[1]));
+    c = a[0]; // expected-warning{{Assigned value is garbage or undefined}}
+}

``````````

</details>


https://github.com/llvm/llvm-project/pull/103714