1263203 - [wasm] Assertion failure: from->type() == to->type() (Def replacement has different type), at js/src/jit/ValueNumbering.cpp:145

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Christian Holler (:decoder)]

The attached binary WebAssembly testcase crashes on mozilla-inbound revision 49784ed1b9e7+ (build with --enable-posix-nspr-emulation --enable-valgrind --enable-gczeal --disable-tests --enable-address-sanitizer --disable-jemalloc --enable-optimize=-O2 --without-intl-api --enable-debug). To reproduce, you can run the following code in the JS shell:

var data = os.file.readFile(file, 'binary');
Wasm.instantiateModule(new Uint8Array(data.buffer));



Backtrace:

==4264==ERROR: AddressSanitizer: SEGV on unknown address 0x000000000000 (pc 0x00000127e78a bp 0x7ffd67543350 sp 0x7ffd67543350 T0)
==4264==The signal is caused by a WRITE memory access.
==4264==Hint: address points to the zero page.
    #0 0x127e789 in ReplaceAllUsesWith(js::jit::MDefinition*, js::jit::MDefinition*) js/src/jit/ValueNumbering.cpp:144:5
    #1 0x127dd2d in js::jit::ValueNumberer::visitDefinition(js::jit::MDefinition*) js/src/jit/ValueNumbering.cpp:767:9
    #2 0x12805aa in js::jit::ValueNumberer::visitBlock(js::jit::MBasicBlock*, js::jit::MBasicBlock const*) js/src/jit/ValueNumbering.cpp:961:14
    #3 0x1280c1b in js::jit::ValueNumberer::visitDominatorTree(js::jit::MBasicBlock*) js/src/jit/ValueNumbering.cpp:1006:18
    #4 0x128146b in js::jit::ValueNumberer::visitGraph() js/src/jit/ValueNumbering.cpp:1044:18
    #5 0x1282cc5 in js::jit::ValueNumberer::run(js::jit::ValueNumberer::UpdateAliasAnalysisFlag) js/src/jit/ValueNumbering.cpp:1213:14
    #6 0xd43b26 in js::jit::OptimizeMIR(js::jit::MIRGenerator*) js/src/jit/Ion.cpp:1628:14
    #7 0x6cc9aa in js::wasm::IonCompileFunction(js::wasm::IonCompileTask*) js/src/asmjs/WasmIonCompile.cpp:3202:14
    #8 0x693a18 in js::wasm::ModuleGenerator::finishFuncDef(unsigned int, unsigned int, js::wasm::FunctionGenerator*) js/src/asmjs/WasmGenerator.cpp:815:14
    #9 0x6295d0 in DecodeFunctionBody(JSContext*, js::wasm::Decoder&, js::wasm::ModuleGenerator&, unsigned int) js/src/asmjs/Wasm.cpp:1349:12
    #10 0x6295d0 in DecodeFunctionBodies(JSContext*, js::wasm::Decoder&, js::wasm::ModuleGenerator&) js/src/asmjs/Wasm.cpp:1377
    #11 0x6295d0 in DecodeModule(JSContext*, mozilla::UniquePtr<char [], JS::FreePolicy>, unsigned char const*, unsigned int, mozilla::Vector<ImportName, 0ul, js::SystemAllocPolicy>*, mozilla::UniquePtr<js::wasm::ExportMap, JS::DeletePolicy<js::wasm::ExportMap> >*, JS::MutableHandle<js::ArrayBufferObject*>, JS::MutableHandle<js::WasmModuleObject*>) js/src/asmjs/Wasm.cpp:1476
    #12 0x61cd00 in js::wasm::Eval(JSContext*, JS::Handle<js::TypedArrayObject*>, JS::Handle<JSObject*>, JS::MutableHandle<JSObject*>) js/src/asmjs/Wasm.cpp:1633:10
    #13 0x5943b5 in WasmLoop(JSContext*, unsigned int, JS::Value*) js/src/shell/js.cpp:5212:14
[...]
    #26 0x460f01 in _start (/home/ubuntu/build/build/js+0x460f01)

AddressSanitizer can not provide additional info.
SUMMARY: AddressSanitizer: SEGV js/src/jit/ValueNumbering.cpp:144:5 in ReplaceAllUsesWith(js::jit::MDefinition*, js::jit::MDefinition*)
==4264==ABORTING

Comment 1

[Christian Holler (:decoder)]

Attachment: Testcase

Comment 2

[Benjamin Bouvier [:bbouvier] (inactive)]

Attachment: phi.patch

Hah, that's something I've wanted to investigate, in case it could cause bugs: we can end up with MDefinitions that have MIRType_Value type, because of the if/else rules of decoding: (if (cond) (f32.const 0) (i32.const 0)), for instance.

We should make sure we never encounter them during wasm compilations. I need to think a bit more before asking for review; maybe we can avoid pushing if we know there was already a different pushed type, but I am not sure.

Comment 3

[Benjamin Bouvier [:bbouvier] (inactive)]

Attachment: 1263203.patch

A better patch: instead of popping the definition and not returning it if it has MIRType_Value (meaning we've already done a bad thing), just prevent to push definitions with different types (so that no phi with MIRType_Value is created in the first place).

Comment 4

[Luke Wagner [:luke]]

Comment on attachment 8739970 [details] [diff] [review]
1263203.patch

Review of attachment 8739970 [details] [diff] [review]:
-----------------------------------------------------------------

Yes, this seems like the right fix.

::: js/src/asmjs/WasmIonCompile.cpp
@@ +973,5 @@
>          if (def && def->type() != MIRType_None)
>              curBlock_->push(def);
>      }
>  
> +    MDefinition* popDef()

How about popDefIfPushed()?

Comment 5

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/6828e2e05e16

Comment 6

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/6828e2e05e16