Bug 191252 – Null pointer dereference in JSC::WriteBarrierBase()

• Component: JavaScriptCore
• Status: RESOLVED FIXED

[Kamil Frankowicz]

Created attachment 353836 [details]
Crashing test case (jsc)

After some fuzz testing I found a crashing test case.

Git HEAD: 43cf251237d5c47f1d0fd8978fd1254870af6721

To reproduce: jsc jsc_nullptr_WriteBarrierBase

Compiler: Clang 4.0

OS: Ubuntu 18.04 x64

ASAN:

==21789==ERROR: AddressSanitizer: SEGV on unknown address 0x000000000018 (pc 0x7f0ed74c493e bp 0x7ffddeadda70 sp 0x7ffddeadd900 T0)
==21789==The signal is caused by a READ memory access.
==21789==Hint: address points to the zero page.
    #0 0x7f0ed74c493d in JSC::WriteBarrierBase<JSC::JSFunction, WTF::DumbPtrTraits<JSC::JSFunction> >::get() const XYZ/webkit_asan/WebKitBuild/Release/../../Source/JavaScriptCore/runtime/WriteBarrier.h:102:28
    #1 0x7f0ed74c493d in JSC::JSPromiseDeferred::reject(JSC::ExecState*, JSC::JSValue) XYZ/webkit_asan/WebKitBuild/Release/../../Source/JavaScriptCore/runtime/JSPromiseDeferred.cpp:118
    #2 0x7f0ed73d3588 in JSC::JSInternalPromiseDeferred::reject(JSC::ExecState*, JSC::JSValue) XYZ/webkit_asan/WebKitBuild/Release/../../Source/JavaScriptCore/runtime/JSInternalPromiseDeferred.cpp:71:11
    #3 0x526ef1 in GlobalObject::moduleLoaderFetch(JSC::JSGlobalObject*, JSC::ExecState*, JSC::JSModuleLoader*, JSC::JSValue, JSC::JSValue, JSC::JSValue) XYZ/webkit_asan/WebKitBuild/Release/../../Source/JavaScriptCore/jsc.cpp:1000:26
    #4 0x7f0ed73f05d6 in JSC::JSModuleLoader::fetch(JSC::ExecState*, JSC::JSValue, JSC::JSValue, JSC::JSValue) XYZ/webkit_asan/WebKitBuild/Release/../../Source/JavaScriptCore/runtime/JSModuleLoader.cpp:285:16
    #5 0x7f0ed7406103 in JSC::moduleLoaderFetch(JSC::ExecState*) XYZ/webkit_asan/WebKitBuild/Release/../../Source/JavaScriptCore/runtime/JSModuleLoader.cpp:461:36
    #6 0x7f0e8c1fe176  (<unknown module>)

AddressSanitizer can not provide additional info.
SUMMARY: AddressSanitizer: SEGV XYZ/webkit_asan/WebKitBuild/Release/../../Source/JavaScriptCore/runtime/WriteBarrier.h:102:28 in JSC::WriteBarrierBase<JSC::JSFunction, WTF::DumbPtrTraits<JSC::JSFunction> >::get() const
==21789==ABORTING

[Alexey Proskuryakov]

I could reproduce in shipping Safari/WebKit.

[Radar WebKit Bug Importer]

<rdar://problem/45824650>

[Alexey Proskuryakov]

This looks more like a consequence of infinite recursion.

[Yusuke Suzuki]

JSInternalPromiseDeferred::create fails if stack overflow happens. We should

1. rename it to `tryCreate`
2. and put error check after calling it.

[Yusuke Suzuki]

Created attachment 353950 [details]
Patch

[Yusuke Suzuki]

Currently, our module mechanism leverages JS builtin code. And right now, it is not immune to the stack overflow. Let's just `RELEASE_ASSERT` instead of nullptr dereference to avoid incorrect memory loading.

With the example test case, JSC shows expected RELEASE_ASSERT crash instead of nullptr dereference.

[EWS Watchlist]

Attachment 353950 [details] did not pass style-queue:


ERROR: Source/WebCore/ChangeLog:8:  You should remove the 'No new tests' and either add and list tests, or explain why no new tests were possible.  [changelog/nonewtests] [5]
Total errors found: 1 in 17 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Yusuke Suzuki]

Ping?

[Keith Miller]

Comment on attachment 353950 [details]
Patch

r=me.

[Yusuke Suzuki]

Committed r239256: <https://trac.webkit.org/changeset/239256>