[PATCH v4] rust: compiler_builtins: make stubs non-global

[PATCH v4] rust: compiler_builtins: make stubs non-global

[Gary Guo]

Currently we define a number of stubs for compiler-builtin intrinsics
that compiled libcore generates. The defined stubs are weak so they will
not conflict with genuine implementation of these intrinsics, but their
effect is global and will cause non-libcore code that accidently
generate these intrinsics calls compile and bug on runtime.

Instead of defining a stub that can affect all code, this patch uses
objcopy's `--redefine-sym` flag to redirect these calls (from libcore
only) to a prefixed version (e.g. redirect `__multi3` to `__rust_multi3`),
so we can define panciking stubs that are only visible to libcore.

This patch was previously discussed on GitHub [1]. This approach was also
independently proposed by Nick Desaulniers in [2].

Link: https://github.com/Rust-for-Linux/linux/pull/779 [1]
Link: https://lore.kernel.org/lkml/CAKwvOdkc0Qhwu=gfe1+H23TnAa6jnO6A3ZCO687dH6mSrATmDA@mail.gmail.com/
Suggested-by: Nick Desaulniers <ndesaulniers@google.com>
Acked-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Gary Guo <gary@garyguo.net>
---
 rust/Makefile             | 14 ++++++++++++++
 rust/compiler_builtins.rs |  5 ++++-
 2 files changed, 18 insertions(+), 1 deletion(-)

diff --git a/rust/Makefile b/rust/Makefile
index ff70c4c916f8..aed6f7eca36f 100644
--- a/rust/Makefile
+++ b/rust/Makefile
@@ -359,8 +359,22 @@ rust-analyzer:
 	$(Q)$(srctree)/scripts/generate_rust_analyzer.py $(srctree) $(objtree) \
 		$(RUST_LIB_SRC) > $(objtree)/rust-project.json
 
+redirect-intrinsics = \
+	__eqsf2 __gesf2 __lesf2 __nesf2 __unordsf2 \
+	__unorddf2 \
+	__muloti4 __multi3 \
+	__udivmodti4 __udivti3 __umodti3
+
+ifneq ($(or $(CONFIG_ARM64),$(and $(CONFIG_RISCV),$(CONFIG_64BIT))),)
+	# These intrinsics are defined for ARM64 and RISCV64
+	redirect-intrinsics += \
+		__ashrti3 \
+		__ashlti3 __lshrti3
+endif
+
 $(obj)/core.o: private skip_clippy = 1
 $(obj)/core.o: private skip_flags = -Dunreachable_pub
+$(obj)/core.o: private rustc_objcopy = $(foreach sym,$(redirect-intrinsics),--redefine-sym $(sym)=__rust$(sym))
 $(obj)/core.o: private rustc_target_flags = $(core-cfgs)
 $(obj)/core.o: $(RUST_LIB_SRC)/core/src/lib.rs $(obj)/target.json FORCE
 	$(call if_changed_dep,rustc_library)
diff --git a/rust/compiler_builtins.rs b/rust/compiler_builtins.rs
index f8f39a3e6855..43378357ece9 100644
--- a/rust/compiler_builtins.rs
+++ b/rust/compiler_builtins.rs
@@ -28,7 +28,7 @@ macro_rules! define_panicking_intrinsics(
     ($reason: tt, { $($ident: ident, )* }) => {
         $(
             #[doc(hidden)]
-            #[no_mangle]
+            #[export_name = concat!("__rust", stringify!($ident))]
             pub extern "C" fn $ident() {
                 panic!($reason);
             }
@@ -61,3 +61,6 @@ define_panicking_intrinsics!("`u128` should not be used", {
     __udivti3,
     __umodti3,
 });
+
+// NOTE: if you are adding a new intrinsic here, you should also add it to
+// `redirect-intrinsics` in `rust/Makefile`.
-- 
2.34.1

Re: [PATCH v4] rust: compiler_builtins: make stubs non-global

[Wei Liu]

On Mon, Dec 05, 2022 at 09:50:00PM +0000, Gary Guo wrote:
> Currently we define a number of stubs for compiler-builtin intrinsics
> that compiled libcore generates. The defined stubs are weak so they will
> not conflict with genuine implementation of these intrinsics, but their
> effect is global and will cause non-libcore code that accidently
> generate these intrinsics calls compile and bug on runtime.
> 
> Instead of defining a stub that can affect all code, this patch uses
> objcopy's `--redefine-sym` flag to redirect these calls (from libcore
> only) to a prefixed version (e.g. redirect `__multi3` to `__rust_multi3`),
> so we can define panciking stubs that are only visible to libcore.
> 
> This patch was previously discussed on GitHub [1]. This approach was also
> independently proposed by Nick Desaulniers in [2].
> 
> Link: https://github.com/Rust-for-Linux/linux/pull/779 [1]
> Link: https://lore.kernel.org/lkml/CAKwvOdkc0Qhwu=gfe1+H23TnAa6jnO6A3ZCO687dH6mSrATmDA@mail.gmail.com/
> Suggested-by: Nick Desaulniers <ndesaulniers@google.com>
> Acked-by: Nick Desaulniers <ndesaulniers@google.com>
> Signed-off-by: Gary Guo <gary@garyguo.net>

Reviewed-by: Wei Liu <wei.liu@kernel.org>

It seems that the only difference between v3 and v4 is Nick's ack.

You can put a changelog by adding a `---` section -- it will be stripped
by git-am automatically.

Thanks,
Wei.