[ubsan] Detect invalid unsigned pointer index expression (compiler-rt)

[vsk]

Summary

Compiler-rt part of: https://reviews.llvm.org/D33910

[regehr]

test/ubsan/TestCases/Pointer/unsigned-index-expression.cpp

6 ↗

(On Diff #101448)

I don't know that it matters too much in practice but it may be better to make p a valid pointer instead of casting 7 into a pointer, which is at best implementation defined behavior.

[vsk]

• Avoid implementation-defined behavior in the test.
• Fix the signed-ness check: to check that the sign bit is off, we need 'x >= 0'. Caught by running check-ubsan.

compiler-rt/trunk/lib/ubsan/ubsan_handlers.cc

static void handlePointerOverflowImpl(PointerOverflowData *Data,
   SourceLocation Loc = Data->Loc.acquire();
   ErrorType ET = ErrorType::PointerOverflow;
 
   if (ignoreReport(Loc, Opts, ET))
     return;
 
   ScopedReport R(Opts, Loc, ET);
 
-  Diag(Loc, DL_Error, "pointer index expression with base %0 overflowed to %1")
-    << (void *)Base << (void*)Result;
+  if ((sptr(Base) >= 0) == (sptr(Result) >= 0))
+    Diag(Loc, DL_Error, "unsigned pointer index expression result is %0, "
+                        "preceding its base %1")
+        << (void *)Result << (void *)Base;
+  else
+    Diag(Loc, DL_Error,
+         "pointer index expression with base %0 overflowed to %1")
+        << (void *)Base << (void *)Result;
 }
 
 void __ubsan::__ubsan_handle_pointer_overflow(PointerOverflowData *Data,
                                               ValueHandle Base,
                                               ValueHandle Result) {
   GET_REPORT_OPTIONS(false);
   handlePointerOverflowImpl(Data, Base, Result, Opts);
 }

compiler-rt/trunk/test/ubsan/TestCases/Pointer/unsigned-index-expression.cpp

+// RUN: %clangxx -fsanitize=pointer-overflow %s -o %t
+// RUN: %t 2>&1 | FileCheck %s
+
+int main(int argc, char *argv[]) {
+  char c;
+  char *p = &c;
+  unsigned long long offset = -1;
+
+  // CHECK: unsigned-index-expression.cpp:[[@LINE+1]]:15: runtime error: unsigned pointer index expression result is 0x{{.*}}, preceding its base 0x{{.*}}
+  char *q = p + offset;
+
+  return 0;
+}