Auto merge of rust-lang#128362 - folkertdev:naked-function-symbol-visibility, r=<try>

add test for symbol visibility of `#[naked]` functions

tracking issue: rust-lang#90957

This test is extracted from rust-lang#128004

That PR attempts to generated naked functions as an extern function declaration, combined with a global asm block that provides the implementation for that declaration.

In order to link declaration and definition together, some flavor of external linking must be used: LLVM will error for other linkage types. Specifically the allowed options are `#[linkage = "external"]` and `#[linkage = "extern_weak"]`. That is kind of an implementation detail though: to the user, a naked function should just behave like a normal function.

Hence it should be visible to the linker under the same circumstances as a normal, vanilla function and have the same attributes (Weak, External). Getting this behavior right will require some care, so I think it's a good idea to lock it in now, before making any changes, to make sure we don't regress.

Are there any interesting cases that I missed here? E.g. is checking on different architectures worth it? I don't think the other binary types (rlib etc) are relevant here, but may be missing something.

r? `@bjorn3`

try-job: x86_64-msvc
try-job: i686-msvc
try-job: i686-mingw
try-job: x86_64-mingw

Author: bors

tests/run-make/naked-symbol-visibility/a_rust_dylib.rs

@@ -0,0 +1,89 @@
+#![feature(naked_functions, asm_const, linkage)]
+#![crate_type = "dylib"]
+
+use std::arch::asm;
+
+pub trait TraitWithConst {
+    const COUNT: u32;
+}
+
+struct Test;
+
+impl TraitWithConst for Test {
+    const COUNT: u32 = 1;
+}
+
+#[no_mangle]
+fn entry() {
+    private_vanilla();
+    private_naked();
+
+    public_vanilla_generic::<Test>();
+    public_naked_generic::<Test>();
+}
+
+extern "C" fn private_vanilla() -> u32 {
+    42
+}
+
+#[naked]
+extern "C" fn private_naked() -> u32 {
+    unsafe { asm!("mov rax, 42", "ret", options(noreturn)) }
+}
+
+#[no_mangle]
+pub extern "C" fn public_vanilla() -> u32 {
+    42
+}
+
+#[naked]
+#[no_mangle]
+pub extern "C" fn public_naked() -> u32 {
+    unsafe { asm!("mov rax, 42", "ret", options(noreturn)) }
+}
+
+pub extern "C" fn public_vanilla_generic<T: TraitWithConst>() -> u32 {
+    T::COUNT
+}
+
+#[naked]
+pub extern "C" fn public_naked_generic<T: TraitWithConst>() -> u32 {
+    unsafe { asm!("mov rax, {}", "ret", const T::COUNT, options(noreturn)) }
+}
+
+#[linkage = "external"]
+extern "C" fn vanilla_external_linkage() -> u32 {
+    42
+}
+
+#[naked]
+#[linkage = "external"]
+extern "C" fn naked_external_linkage() -> u32 {
+    unsafe { asm!("mov rax, 42", "ret", options(noreturn)) }
+}
+
+#[cfg(not(windows))]
+#[linkage = "weak"]
+extern "C" fn vanilla_weak_linkage() -> u32 {
+    42
+}
+
+#[naked]
+#[cfg(not(windows))]
+#[linkage = "weak"]
+extern "C" fn naked_weak_linkage() -> u32 {
+    unsafe { asm!("mov rax, 42", "ret", options(noreturn)) }
+}
+
+// functions that are declared in an `extern "C"` block are currently not exported
+// this maybe should change in the future, this is just tracking the current behavior
+// reported in https://github.com/rust-lang/rust/issues/128071
+std::arch::global_asm! {
+    ".globl function_defined_in_global_asm",
+    "function_defined_in_global_asm:",
+    "ret",
+}
+
+extern "C" {
+    pub fn function_defined_in_global_asm();
+}

tests/run-make/naked-symbol-visibility/rmake.rs

@@ -0,0 +1,93 @@
+//@ only-x86_64
+use run_make_support::object::read::{File, Object, Symbol};
+use run_make_support::object::ObjectSymbol;
+use run_make_support::targets::is_windows;
+use run_make_support::{dynamic_lib_name, env_var, rfs, rustc};
+
+fn main() {
+    let rdylib_name = dynamic_lib_name("a_rust_dylib");
+    rustc().arg("-Zshare-generics=no").input("a_rust_dylib.rs").run();
+
+    let binary_data = rfs::read(&rdylib_name);
+    let rdylib = File::parse(&*binary_data).unwrap();
+
+    // naked should mirror vanilla
+    not_exported(&rdylib, "private_vanilla");
+    not_exported(&rdylib, "private_naked");
+
+    global_function(&rdylib, "public_vanilla");
+    global_function(&rdylib, "public_naked");
+
+    not_exported(&rdylib, "public_vanilla_generic");
+    not_exported(&rdylib, "public_naked_generic");
+
+    global_function(&rdylib, "vanilla_external_linkage");
+    global_function(&rdylib, "naked_external_linkage");
+
+    // #[linkage = "weak"] does not work well on windows, we get
+    //
+    // lib.def : error LNK2001: unresolved external symbol naked_weak_linkage␍
+    // lib.def : error LNK2001: unresolved external symbol vanilla_weak_linkage
+    //
+    // so just skip weak symbols on windows (for now)
+    if !is_windows() {
+        weak_function(&rdylib, "vanilla_weak_linkage");
+        weak_function(&rdylib, "naked_weak_linkage");
+    }
+
+    // functions that are declared in an `extern "C"` block are currently not exported
+    // this maybe should change in the future, this is just tracking the current behavior
+    // reported in https://github.com/rust-lang/rust/issues/128071
+    not_exported(&rdylib, "function_defined_in_global_asm");
+
+    // share generics should expose the generic functions
+    rustc().arg("-Zshare-generics=yes").input("a_rust_dylib.rs").run();
+    let binary_data = rfs::read(&rdylib_name);
+    let rdylib = File::parse(&*binary_data).unwrap();
+
+    global_function(&rdylib, "public_vanilla_generic");
+    global_function(&rdylib, "public_naked_generic");
+}
+
+#[track_caller]
+fn global_function(file: &File, symbol_name: &str) {
+    let symbols = find_dynamic_symbol(file, symbol_name);
+    let [symbol] = symbols.as_slice() else {
+        panic!("symbol {symbol_name} occurs {} times", symbols.len())
+    };
+
+    assert!(symbol.is_definition(), "`{symbol_name}` is not a function");
+    assert!(symbol.is_global(), "`{symbol_name}` is not marked as global");
+}
+
+#[track_caller]
+fn weak_function(file: &File, symbol_name: &str) {
+    let symbols = find_dynamic_symbol(file, symbol_name);
+    let [symbol] = symbols.as_slice() else {
+        panic!("symbol {symbol_name} occurs {} times", symbols.len())
+    };
+
+    assert!(symbol.is_definition(), "`{symbol_name}` is not a function");
+    assert!(symbol.is_weak(), "`{symbol_name}` is not marked as weak");
+}
+
+#[track_caller]
+fn not_exported(file: &File, symbol_name: &str) {
+    assert_eq!(find_dynamic_symbol(file, symbol_name).len(), 0)
+}
+
+fn find_subsequence(haystack: &[u8], needle: &[u8]) -> bool {
+    haystack.windows(needle.len()).any(|window| window == needle)
+}
+
+fn find_dynamic_symbol<'file, 'data>(
+    file: &'file File<'data>,
+    expected: &str,
+) -> Vec<Symbol<'data, 'file>> {
+    file.exports()
+        .unwrap()
+        .into_iter()
+        .filter(|e| find_subsequence(e.name(), expected.as_bytes()))
+        .filter_map(|e| file.symbol_by_name_bytes(e.name()))
+        .collect()
+}