Bug 149957 – [Win] Null pointer crash.

• Component: JavaScriptCore
• Status: NEW

I just got a null pointer crash in JSC::speculationFromCell(). The JSCell object looks more or less ok, but the m_structureID member is 0, causing the null pointer crash. 


JavaScriptCore.dll!JSC::speculationFromCell(JSC::JSCell * cell)  Line 363 + 0x20 bytes	C++
JavaScriptCore.dll!JSC::speculationFromValue(JSC::JSValue value)  Line 391 + 0x8 bytes	C++
JavaScriptCore.dll!JSC::ValueProfileBase<1>::computeUpdatedPrediction(const JSC::ConcurrentJITLocker & __formal)  Line 145 + 0x7 bytes	C++
JavaScriptCore.dll!JSC::CodeBlock::updateAllPredictionsAndCountLiveness(unsigned int & numberOfLiveNonArgumentValueProfiles, unsigned int & numberOfSamplesInProfiles)  Line 3770	C++
JavaScriptCore.dll!JSC::CodeBlock::updateAllPredictions()  Line 3815	C++
JavaScriptCore.dll!operationOptimize(JSC::ExecState * exec, int bytecodeIndex)  Line 1142	C++

Created attachment 262776 [details]
Patch

(In reply to comment #1)
> Created attachment 262776 [details]
> Patch

I'm not sure this is the correct way to deal with this crash :)

[Geoffrey Garen]

Comment on attachment 262776 [details]
Patch

While this might fix the crash, I think it's the wrong fix.

A cell with a null structureID has been garbage collected. You're lucky if you find the null structureID -- that happens soon after sweeping. If you're unlucky, you'll just get garbage memory, or a crash.

We need to investigate how cell got garbage collected in the first place.

(In reply to comment #3)
> Comment on attachment 262776 [details]
> Patch
> 
> While this might fix the crash, I think it's the wrong fix.
> 
> A cell with a null structureID has been garbage collected. You're lucky if
> you find the null structureID -- that happens soon after sweeping. If you're
> unlucky, you'll just get garbage memory, or a crash.
> 
> We need to investigate how cell got garbage collected in the first place.

Ok, sounds good :)

This is the state of the JSCell object when the crash occured:

m_structureID   0x00000000 JSC::Structure*
m_indexingType  0          unsigned char
m_type          StringType JSC::JSType
m_flags         224        unsigned char
m_cellState     NewWhite   JSC::CellState

[Dan Zimmerman]

I've found a similar crash on iOS 10.2's version of JavascriptCore.

I have the following backtrace:

#0	0x0000000104d6d2ef in JSC::speculationFromCell(JSC::JSCell*) ()
#1	0x00000001047d2ec3 in JSC::CodeBlock::updateAllPredictionsAndCountLiveness(unsigned int&, unsigned int&) ()
#2	0x00000001047cebc7 in JSC::CodeBlock::stronglyVisitStrongReferences(JSC::SlotVisitor&) ()
#3	0x00000001047ce969 in JSC::CodeBlock::visitChildren(JSC::SlotVisitor&) ()
#4	0x0000000104d699ac in JSC::SlotVisitor::drain() ()
#5	0x0000000104a0869c in JSC::Heap::markRoots(double, void*, void*, int (&) [37]) ()
#6	0x0000000104a0b065 in JSC::Heap::collectImpl(JSC::HeapOperation, void*, void*, int (&) [37]) ()
#7	0x0000000104a0ada1 in JSC::Heap::collect(JSC::HeapOperation) ()
#8	0x0000000104c3c957 in JSC::MarkedAllocator::allocateSlowCase(unsigned long) ()
#9	0x0000000104b72ee0 in JSObjectMake ()

and the state of the JSCell is:

m_structureID: 0
m_indexingType: 0
m_type: UnspecifiedType (0)
m_flags: 0
m_cellState: AnthraciteOrBlack (0)

Is there anyway I can help to find the source of the issue? I have a setup where the bug is pretty reproducible.