Auto merge of #122580 - saethlin:compiler-builtins-can-panic, r=pnkfelix

"Handle" calls to upstream monomorphizations in compiler_builtins

This is pretty cooked, but I think it works.

compiler-builtins has a long-standing problem that at link time, its rlib cannot contain any calls to `core`. And yet, in codegen we _love_ inserting calls to symbols in `core`, generally from various panic entrypoints.

I intend this PR to attack that problem as completely as possible. When we generate a function call, we now check if we are generating a function call from `compiler_builtins` and whether the callee is a function which was not lowered in the current crate, meaning we will have to link to it.

If those conditions are met, actually generating the call is asking for a linker error. So we don't. If the callee diverges, we lower to an abort with the same behavior as `core::intrinsics::abort`. If the callee does not diverge, we produce an error. This means that compiler-builtins can contain panics, but they'll SIGILL instead of panicking. I made non-diverging calls a compile error because I'm guessing that they'd mostly get into compiler-builtins by someone making a mistake while working on the crate, and compile errors are better than linker errors. We could turn such calls into aborts as well if that's preferred.

2 files changed, 23 insertions, 1 deletions

diff --git a/compiler/rustc_monomorphize/src/collector.rs b/compiler/rustc_monomorphize/src/collector.rs
index 1a18a84b358..673cd968e28 100644
--- a/compiler/rustc_monomorphize/src/collector.rs
+++ b/compiler/rustc_monomorphize/src/collector.rs
@@ -1107,7 +1107,7 @@ fn visit_instance_use<'tcx>(
 
 /// Returns `true` if we should codegen an instance in the local crate, or returns `false` if we
 /// can just link to the upstream crate and therefore don't need a mono item.
-fn should_codegen_locally<'tcx>(tcx: TyCtxt<'tcx>, instance: &Instance<'tcx>) -> bool {
+pub(crate) fn should_codegen_locally<'tcx>(tcx: TyCtxt<'tcx>, instance: &Instance<'tcx>) -> bool {
     let Some(def_id) = instance.def.def_id_if_not_guaranteed_local_codegen() else {
         return true;
     };
diff --git a/compiler/rustc_monomorphize/src/lib.rs b/compiler/rustc_monomorphize/src/lib.rs
index 7f36ae91f1a..4ec842e8f85 100644
--- a/compiler/rustc_monomorphize/src/lib.rs
+++ b/compiler/rustc_monomorphize/src/lib.rs
@@ -11,7 +11,10 @@ use rustc_hir::lang_items::LangItem;
 use rustc_middle::query::{Providers, TyCtxtAt};
 use rustc_middle::traits;
 use rustc_middle::ty::adjustment::CustomCoerceUnsized;
+use rustc_middle::ty::Instance;
+use rustc_middle::ty::TyCtxt;
 use rustc_middle::ty::{self, Ty};
+use rustc_span::def_id::LOCAL_CRATE;
 use rustc_span::ErrorGuaranteed;
 
 mod collector;
@@ -20,6 +23,8 @@ mod partitioning;
 mod polymorphize;
 mod util;
 
+use collector::should_codegen_locally;
+
 rustc_fluent_macro::fluent_messages! { "../messages.ftl" }
 
 fn custom_coerce_unsize_info<'tcx>(
@@ -45,6 +50,23 @@ fn custom_coerce_unsize_info<'tcx>(
     }
 }
 
+/// Returns whether a call from the current crate to the [`Instance`] would produce a call
+/// from `compiler_builtins` to a symbol the linker must resolve.
+///
+/// Such calls from `compiler_bultins` are effectively impossible for the linker to handle. Some
+/// linkers will optimize such that dead calls to unresolved symbols are not an error, but this is
+/// not guaranteed. So we used this function in codegen backends to ensure we do not generate any
+/// unlinkable calls.
+pub fn is_call_from_compiler_builtins_to_upstream_monomorphization<'tcx>(
+    tcx: TyCtxt<'tcx>,
+    instance: Instance<'tcx>,
+) -> bool {
+    !instance.def_id().is_local()
+        && tcx.is_compiler_builtins(LOCAL_CRATE)
+        && tcx.codegen_fn_attrs(instance.def_id()).link_name.is_none()
+        && !should_codegen_locally(tcx, &instance)
+}
+
 pub fn provide(providers: &mut Providers) {
     partitioning::provide(providers);
     polymorphize::provide(providers);