diff options
| author | Mazdak Farrokhzad <twingoow@gmail.com> | 2019-12-20 17:22:22 +0100 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2019-12-20 17:22:22 +0100 |
| commit | 43d1532cd7c2d03e8af3e7179edccd89e989897f (patch) | |
| tree | 4e6f0da977dffdce7073fdf1a681986d7b38650b /src/librustc_codegen_utils | |
| parent | 5a8083c6652f55b5a902dc4c60070ec87a632c22 (diff) | |
| parent | aa0ef5a01ff329daa822e7443ca7d6ae2bfc8476 (diff) | |
| download | rust-43d1532cd7c2d03e8af3e7179edccd89e989897f.tar.gz rust-43d1532cd7c2d03e8af3e7179edccd89e989897f.zip | |
Rollup merge of #67363 - alexcrichton:wasm-import-modules, r=eddyb
Fix handling of wasm import modules and names The WebAssembly targets of rustc have weird issues around name mangling and import the same name from different modules. This all largely stems from the fact that we're using literal symbol names in LLVM IR to represent what a function is called when it's imported, and we're not using the wasm-specific `wasm-import-name` attribute. This in turn leads to two issues: * If, in the same codegen unit, the same FFI symbol is referenced twice then rustc, when translating to LLVM IR, will only reference one symbol from the first wasm module referenced. * There's also a bug in LLD [1] where even if two codegen units reference different modules, having the same symbol names means that LLD coalesces the symbols and only refers to one wasm module. Put another way, all our imported wasm symbols from the environment are keyed off their LLVM IR symbol name, which has lots of collisions today. This commit fixes the issue by implementing two changes: 1. All wasm symbols with `#[link(wasm_import_module = "...")]` are mangled by default in LLVM IR. This means they're all given unique names. 2. Symbols then use the `wasm-import-name` attribute to ensure that the WebAssembly file uses the correct import name. When put together this should ensure we don't trip over the LLD bug [1] and we also codegen IR correctly always referencing the right symbols with the right import module/name pairs. Closes #50021 Closes #56309 Closes #63562 [1]: https://bugs.llvm.org/show_bug.cgi?id=44316
Diffstat (limited to 'src/librustc_codegen_utils')
| -rw-r--r-- | src/librustc_codegen_utils/symbol_names.rs | 28 |
1 files changed, 24 insertions, 4 deletions
diff --git a/src/librustc_codegen_utils/symbol_names.rs b/src/librustc_codegen_utils/symbol_names.rs index c52c6cfa83c..922964ee45f 100644 --- a/src/librustc_codegen_utils/symbol_names.rs +++ b/src/librustc_codegen_utils/symbol_names.rs @@ -142,12 +142,32 @@ fn symbol_name(tcx: TyCtxt<'tcx>, instance: Instance<'tcx>) -> Symbol { }; let attrs = tcx.codegen_fn_attrs(def_id); + + // Foreign items by default use no mangling for their symbol name. There's a + // few exceptions to this rule though: + // + // * This can be overridden with the `#[link_name]` attribute + // + // * On the wasm32 targets there is a bug (or feature) in LLD [1] where the + // same-named symbol when imported from different wasm modules will get + // hooked up incorectly. As a result foreign symbols, on the wasm target, + // with a wasm import module, get mangled. Additionally our codegen will + // deduplicate symbols based purely on the symbol name, but for wasm this + // isn't quite right because the same-named symbol on wasm can come from + // different modules. For these reasons if `#[link(wasm_import_module)]` + // is present we mangle everything on wasm because the demangled form will + // show up in the `wasm-import-name` custom attribute in LLVM IR. + // + // [1]: https://bugs.llvm.org/show_bug.cgi?id=44316 if is_foreign { - if let Some(name) = attrs.link_name { - return name; + if tcx.sess.target.target.arch != "wasm32" || + !tcx.wasm_import_module_map(def_id.krate).contains_key(&def_id) + { + if let Some(name) = attrs.link_name { + return name; + } + return tcx.item_name(def_id); } - // Don't mangle foreign items. - return tcx.item_name(def_id); } if let Some(name) = attrs.export_name { |
