[clang][diagnostics] Reject embedded NUL characters in inline asm

[el-ev]

As suggested by @jmorse and @efriedma-quic in #196223.

[llvmorg-github-actions]

@llvm/pr-subscribers-clang

Author: Iris Shi (el-ev)

Changes

As suggested by @jmorse and @efriedma-quic in #196223.

---

Full diff: https://github.com/llvm/llvm-project/pull/196462.diff

5 Files Affected:

• (modified) clang/docs/ReleaseNotes.rst (+3)
• (modified) clang/include/clang/Basic/DiagnosticSemaKinds.td (+3)
• (modified) clang/lib/Sema/SemaStmtAsm.cpp (+18)
• (removed) clang/test/CodeGen/inline-asm-constraint-embedded-null.c (-8)
• (added) clang/test/Sema/inline-asm-constraint-embedded-null.c (+16)

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index cb19b80b7e994..51e98aef4d251 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -479,6 +479,9 @@ Improvements to Clang's diagnostics
 
 - Removed the body of lambdas from some diagnostic messages.
 
+- Clang now rejects inline asm constraints and clobbers that contain an
+  embedded null character, instead of silently truncating them. (#GH173900)
+
 Improvements to Clang's time-trace
 ----------------------------------
 
diff --git a/clang/include/clang/Basic/DiagnosticSemaKinds.td b/clang/include/clang/Basic/DiagnosticSemaKinds.td
index c69b2ce3648f8..c30ddf445ed65 100644
--- a/clang/include/clang/Basic/DiagnosticSemaKinds.td
+++ b/clang/include/clang/Basic/DiagnosticSemaKinds.td
@@ -10074,6 +10074,9 @@ let CategoryName = "Inline Assembly Issue" in {
     "invalid lvalue in asm input for constraint '%0'">;
   def err_asm_invalid_input_constraint : Error<
     "invalid input constraint '%0' in asm">;
+  def err_asm_constraint_embedded_null : Error<
+    "%select{output constraint|input constraint|clobber}0 contains "
+    "embedded null character">;
   def err_asm_tying_incompatible_types : Error<
     "unsupported inline asm: input with type "
     "%diff{$ matching output with type $|}0,1">;
diff --git a/clang/lib/Sema/SemaStmtAsm.cpp b/clang/lib/Sema/SemaStmtAsm.cpp
index f957bdf7156c7..f80c1a5b65f93 100644
--- a/clang/lib/Sema/SemaStmtAsm.cpp
+++ b/clang/lib/Sema/SemaStmtAsm.cpp
@@ -306,6 +306,12 @@ StmtResult Sema::ActOnGCCAsmStmt(SourceLocation AsmLoc, bool IsSimple,
     std::string ConstraintStr =
         GCCAsmStmt::ExtractStringFromGCCAsmStmtComponent(Constraint);
 
+    if (ConstraintStr.find('\0') != std::string::npos) {
+      Diag(Constraint->getBeginLoc(), diag::err_asm_constraint_embedded_null)
+          << /*output*/0;
+      return CreateGCCAsmStmt();
+    }
+
     TargetInfo::ConstraintInfo Info(ConstraintStr, OutputName);
     if (!Context.getTargetInfo().validateOutputConstraint(Info) &&
         !(LangOpts.HIPStdPar && LangOpts.CUDAIsDevice)) {
@@ -396,6 +402,12 @@ StmtResult Sema::ActOnGCCAsmStmt(SourceLocation AsmLoc, bool IsSimple,
     std::string ConstraintStr =
         GCCAsmStmt::ExtractStringFromGCCAsmStmtComponent(Constraint);
 
+    if (ConstraintStr.find('\0') != std::string::npos) {
+      Diag(Constraint->getBeginLoc(), diag::err_asm_constraint_embedded_null)
+          << /*input*/1;
+      return CreateGCCAsmStmt();
+    }
+
     TargetInfo::ConstraintInfo Info(ConstraintStr, InputName);
     if (!Context.getTargetInfo().validateInputConstraint(OutputConstraintInfos,
                                                          Info)) {
@@ -503,6 +515,12 @@ StmtResult Sema::ActOnGCCAsmStmt(SourceLocation AsmLoc, bool IsSimple,
     std::string Clobber =
         GCCAsmStmt::ExtractStringFromGCCAsmStmtComponent(ClobberExpr);
 
+    if (Clobber.find('\0') != std::string::npos) {
+      Diag(ClobberExpr->getBeginLoc(), diag::err_asm_constraint_embedded_null)
+          << /*clobber*/2;
+      return CreateGCCAsmStmt();
+    }
+
     if (!Context.getTargetInfo().isValidClobber(Clobber)) {
       targetDiag(ClobberExpr->getBeginLoc(),
                  diag::err_asm_unknown_register_name)
diff --git a/clang/test/CodeGen/inline-asm-constraint-embedded-null.c b/clang/test/CodeGen/inline-asm-constraint-embedded-null.c
deleted file mode 100644
index c2cd3ace0ddd3..0000000000000
--- a/clang/test/CodeGen/inline-asm-constraint-embedded-null.c
+++ /dev/null
@@ -1,8 +0,0 @@
-// REQUIRES: x86-registered-target
-// RUN: %clang_cc1 -triple x86_64-unknown-unknown -emit-llvm -o - %s | FileCheck %s
-
-// Regression test for issue173900.
-
-// CHECK-LABEL: define {{.*}}void @f(
-// CHECK: call void asm sideeffect "", "f,{{[^"]*}}"(double 0.000000e+00)
-void f(void) { __asm__("" : : "f\0001"(0.0)); }
diff --git a/clang/test/Sema/inline-asm-constraint-embedded-null.c b/clang/test/Sema/inline-asm-constraint-embedded-null.c
new file mode 100644
index 0000000000000..e8587e469ba85
--- /dev/null
+++ b/clang/test/Sema/inline-asm-constraint-embedded-null.c
@@ -0,0 +1,16 @@
+// RUN: %clang_cc1 -triple x86_64-unknown-unknown -fsyntax-only -verify %s
+
+// Regression test for issue173900.
+
+void test_input(void) {
+  __asm__("" : : "f\0001"(0.0)); // expected-error {{input constraint contains embedded null character}}
+}
+
+void test_output(void) {
+  double x;
+  __asm__("" : "=r\0"(x)); // expected-error {{output constraint contains embedded null character}}
+}
+
+void test_clobber(void) {
+  __asm__("" : : : "rax\0"); // expected-error {{clobber contains embedded null character}}
+}

[github-actions]

✅ With the latest revision this PR passed the C/C++ code formatter.

[github-actions]

🐧 Linux x64 Test Results

• 116031 tests passed
• 4688 tests skipped

✅ The build succeeded and all tests passed.

[github-actions]

🪟 Windows x64 Test Results

• 55484 tests passed
• 2526 tests skipped

✅ The build succeeded and all tests passed.

[jmorse]

Thanks for refining this into a diagnostic, I feel it's a much more complete solution. In principle this looks good to me, but it's a lot further into clang than I usually venture, so I'd much prefer someone else tick the approve button.

[cor3ntin]

LGTM, thanks!