Bug 150663 – Crash making a tail call from a getter to a host function

• Component: JavaScriptCore
• Status: RESOLVED FIXED

[Michael Saboff]

This test crashes with a debug build:
——
"use strict";

class Test {
    constructor(a, b) {
        this.a = a;
        this.b = b;
    }

    get sum() {
        return Number(this.a + this.b);
    }
}

var testObj = new Test(40, 2);

var result = testObj.sum;

print(result);
——

Crash trace

ASSERTION FAILED: cell->isObject()
/Volumes/Data/src/webkit/Source/JavaScriptCore/runtime/JSObject.h(1064) : JSC::JSObject *JSC::asObject(JSC::JSCell *)
1   0x108f84880 WTFCrash
2   0x1083db051 JSC::asObject(JSC::JSCell*)
3   0x1083dafd0 JSC::asObject(JSC::JSValue)
4   0x1083dafa2 JSC::Register::object() const
5   0x1083daf0c JSC::ExecState::callee() const
6   0x1083d65f9 JSC::ExecState::vm() const
7   0x108b1d73e getHostCallReturnValueWithExecState
8   0x108ccf26e vmEntryToJavaScript
9   0x108b08275 JSC::JITCode::execute(JSC::VM*, JSC::ProtoCallFrame*)
10  0x108ada4f2 JSC::Interpreter::executeCall(JSC::ExecState*, JSC::JSObject*, JSC::CallType, JSC::CallData const&, JSC::JSValue, JSC::ArgList const&)
11  0x10847f07e JSC::call(JSC::ExecState*, JSC::JSValue, JSC::CallType, JSC::CallData const&, JSC::JSValue, JSC::ArgList const&)
12  0x108a0360d JSC::callGetter(JSC::ExecState*, JSC::JSValue, JSC::JSValue)
13  0x108e55d3d JSC::PropertySlot::functionGetter(JSC::ExecState*) const
14  0x1083e9aad JSC::PropertySlot::getValue(JSC::ExecState*, JSC::PropertyName) const
15  0x1083e98eb JSC::JSValue::get(JSC::ExecState*, JSC::PropertyName, JSC::PropertySlot&) const
16  0x108cc59c1 llint_slow_path_get_by_id
17  0x108cd2330 llint_entry
18  0x108ccf26e vmEntryToJavaScript
19  0x108b08275 JSC::JITCode::execute(JSC::VM*, JSC::ProtoCallFrame*)
20  0x108ad9e92 JSC::Interpreter::execute(JSC::ProgramExecutable*, JSC::ExecState*, JSC::JSObject*)
21  0x10851e011 JSC::evaluate(JSC::ExecState*, JSC::SourceCode const&, JSC::JSValue, WTF::NakedPtr<JSC::Exception>&)
22  0x108340f8f runWithScripts(GlobalObject*, WTF::Vector<Script, 0ul, WTF::CrashOnOverflow, 16ul> const&, bool, bool)
23  0x10833ff53 jscmain(int, char**)
24  0x10833fa3b main
25  0x7fff940545ad start
Segmentation fault: 11


The problem is that the getHostCallReturnValue() function passes its frame pointer as an argument to getHostCallReturnValueWithExecState(), which wants to use the frame to get the current VM.  The passed frame is the caller of a host function, which is a C++ function in the test above.  A fix is to change getHostCallReturnValue() to pass what will be the callee's frame address to getHostCallReturnValueWithExecState().

<rdar://problem/23287664>

[Michael Saboff]

Created attachment 264300 [details]
Proposed patch for all platforms except Windows X86

Need to work hands on to get Windows X86 working.

[Build Bot]

Comment on attachment 264300 [details]
Proposed patch for all platforms except Windows X86

Attachment 264300 [details] did not pass mac-ews (mac):
Output: http://webkit-queues.webkit.org/results/351761

New failing tests:
js/regress-150663.html

[Build Bot]

Created attachment 264301 [details]
Archive of layout-test-results from ews103 for mac-mavericks

The attached test failures were seen while running run-webkit-tests on the mac-ews.
Bot: ews103  Port: mac-mavericks  Platform: Mac OS X 10.9.5

[Build Bot]

Comment on attachment 264300 [details]
Proposed patch for all platforms except Windows X86

Attachment 264300 [details] did not pass mac-wk2-ews (mac-wk2):
Output: http://webkit-queues.webkit.org/results/351770

New failing tests:
js/regress-150663.html

[Build Bot]

Created attachment 264302 [details]
Archive of layout-test-results from ews105 for mac-mavericks-wk2

The attached test failures were seen while running run-webkit-tests on the mac-wk2-ews.
Bot: ews105  Port: mac-mavericks-wk2  Platform: Mac OS X 10.9.5

[Build Bot]

Comment on attachment 264300 [details]
Proposed patch for all platforms except Windows X86

Attachment 264300 [details] did not pass mac-debug-ews (mac):
Output: http://webkit-queues.webkit.org/results/351760

New failing tests:
js/regress-150663.html

[Build Bot]

Created attachment 264303 [details]
Archive of layout-test-results from ews115 for mac-yosemite

The attached test failures were seen while running run-webkit-tests on the mac-debug-ews.
Bot: ews115  Port: mac-yosemite  Platform: Mac OS X 10.10.5

[Geoffrey Garen]

Comment on attachment 264300 [details]
Proposed patch for all platforms except Windows X86

EWS is angry.

[Michael Saboff]

Created attachment 264314 [details]
Patch with updated expectations for new test.

Forgot to update expected results after changing description() in prior patch.

Still need to implement Windows X86.

[Geoffrey Garen]

> Still need to implement Windows X86.

Will this patch start crashing the windows testers if it lands in its current state?

[Michael Saboff]

(In reply to comment #10)
> > Still need to implement Windows X86.
> 
> Will this patch start crashing the windows testers if it lands in its
> current state?

The new test should crash on Windows in the current state.

I'm testing the Windows-86 specific part of the change now.  I hope to post a new patch soon.

[Geoffrey Garen]

Comment on attachment 264314 [details]
Patch with updated expectations for new test.

OK, let's not land this yet, since it would turn the windows testers red.

[Michael Saboff]

Created attachment 264355 [details]
Patch

Tested on Mac 32 & 64, iOS 32 & 64 and Win 32.

[Geoffrey Garen]

Comment on attachment 264355 [details]
Patch

r=me

[Michael Saboff]

Committed r191765: <http://trac.webkit.org/changeset/191765>

[Michael Saboff]

The Windows X86-64 change was inadvertently missed when fixing this bug.  That change is tracked in https://bugs.webkit.org/show_bug.cgi?id=150737