I'm not sure that we will be able to do much without a testcase. Kind of looks like the ownerNode went away but I don't think I understand the lifetime of CSSStyleSheets and their style elements well enough to find the bug. I'll investigate a bit but I'm not overly optimistic :)

I asked him to file this so the v8 guys could have a look, in case the stack trace gives them a clue. If not we'll close it out. However, since it's not confirmed, please don't assign labels that could be misleading.

This looks like a crash in the bindings, not in V8 itself. Adding the bindings guys.

cdn: What makes you think this is related to ownerNode? I'm curious because we rolled back a change that involved ownerNode. See https://bugs.webkit.org/show_bug.cgi?id=80880 for more details

arv, Total guess based on a look at the code around where the crash occurs. A bad ownerNode seemed like the only thing in WebCore::toV8() that could cause a read av.

I am not at all confident in that analysis though :)

Oh also if you look at the stack where it was allocated it is creating a style element. Would the ownerNode for a CssStyleSheet would be a style element?

I'll stop guessing now

Good news - I am able to reproduce crash now, will try to provide reduced testcase tomorrow.

Excellent!

Want to post what you have now and we can run it through our minimizer. I promise it won't affect the bounty :)

Unfortunately can't do it right now, because testcase is too complicated - requires too much stuff to be done manually and cannot be automated. Actually, it's a part of fuzzing framework and has also bunch of dependencies. Tomorrow I will post minimized version, I'm working on it.

No worries. Thanks for continuing to bang on this one.

Attaching testcase and here is a new ASan log:
=================================================================
==4607== ERROR: AddressSanitizer heap-use-after-free on address 0x7fc65ad84c88 at pc 0x7fc6786ffaa9 bp 0x7fffd98b71d0 sp 0x7fffd98b71c8
READ of size 8 at 0x7fc65ad84c88 thread T0
    #0 0x7fc6786ffaa9 in WebCore::toV8(WebCore::CSSStyleSheet*) ???:0
    #1 0x7fc67339b48e in v8::internal::JSObject::GetElementWithInterceptor(v8::internal::Object*, unsigned int) /media/Chromium/chromium/depot_tools/src/v8/src/objects.cc:10097
    #2 0x7fc6734f6004 in v8::internal::Runtime::GetObjectProperty(v8::internal::Isolate*, v8::internal::Handle<v8::internal::Object>, v8::internal::Handle<v8::internal::Object>) /media/Chromium/chromium/depot_tools/src/v8/src/runtime.cc:4182
    #3 0x7fc6737d93fc in v8::internal::KeyedLoadIC::Load(v8::internal::InlineCacheState, v8::internal::Handle<v8::internal::Object>, v8::internal::Handle<v8::internal::Object>, bool) /media/Chromium/chromium/depot_tools/src/v8/src/ic.cc:1195
    #4 0x7fc6737e4cd5 in v8::internal::KeyedLoadIC_Miss(v8::internal::Arguments, v8::internal::Isolate*) /media/Chromium/chromium/depot_tools/src/v8/src/ic.cc:1987
    #5 0x226d1c0614e
    #6 0x226d1c45020
    #7 0x226d1c22647
    #8 0x226d1c11137
    #9 0x7fc6730db278 in v8::internal::Invoke(bool, v8::internal::Handle<v8::internal::JSFunction>, v8::internal::Handle<v8::internal::Object>, int, v8::internal::Handle<v8::internal::Object>*, bool*) /media/Chromium/chromium/depot_tools/src/v8/src/execution.cc:118
    #10 0x7fc673025586 in v8::Script::Run() /media/Chromium/chromium/depot_tools/src/v8/src/api.cc:1588
    #11 0x7fc674d4aa03 in WebCore::V8Proxy::runScript(v8::Handle<v8::Script>) /media/Chromium/chromium/depot_tools/src/third_party/WebKit/Source/WebCore/bindings/v8/V8Proxy.cpp:382
    #12 0x7fc674d49bab in WebCore::V8Proxy::evaluate(WebCore::ScriptSourceCode const&, WebCore::Node*) /media/Chromium/chromium/depot_tools/src/third_party/WebKit/Source/WebCore/bindings/v8/V8Proxy.cpp:353
    #13 0x7fc675576da7 in ~Scope /media/Chromium/chromium/depot_tools/src/v8/include/v8.h:3577
    #14 0x7fc6753b0d68 in WebCore::InspectorInstrumentation::hasFrontends() /media/Chromium/chromium/depot_tools/src/third_party/WebKit/Source/WebCore/inspector/InspectorInstrumentation.h:225
    #15 0x7fc6749c05d8 in WebCore::ThreadTimers::sharedTimerFiredInternal() /media/Chromium/chromium/depot_tools/src/third_party/WebKit/Source/WebCore/platform/ThreadTimers.cpp:118
    #16 0x7fc6725e9256 in base::Callback<void ()()>::Run() const /media/Chromium/chromium/depot_tools/src/./base/callback.h:272
    #17 0x7fc6725e9ab8 in MessageLoop::DeferOrRunPendingTask(base::PendingTask const&) /media/Chromium/chromium/depot_tools/src/base/message_loop.cc:470
    #18 0x7fc6725eada9 in MessageLoop::DoWork() /media/Chromium/chromium/depot_tools/src/base/message_loop.cc:660
    #19 0x7fc6725f5277 in base::MessagePumpDefault::Run(base::MessagePump::Delegate*) /media/Chromium/chromium/depot_tools/src/base/message_pump_default.cc:28
    #20 0x7fc6725e7e1e in MessageLoop::RunInternal() /media/Chromium/chromium/depot_tools/src/base/message_loop.cc:418
    #21 0x7fc6725e600f in ~AutoRunState /media/Chromium/chromium/depot_tools/src/base/message_loop.cc:745
    #22 0x7fc6779c1b5c in RendererMain(content::MainFunctionParams const&) /media/Chromium/chromium/depot_tools/src/content/renderer/renderer_main.cc:241
    #23 0x7fc67253e4b6 in RunZygote /media/Chromium/chromium/depot_tools/src/content/app/content_main_runner.cc:245
    #24 0x7fc67253ca2a in content::ContentMain(int, char const**, content::ContentMainDelegate*) /media/Chromium/chromium/depot_tools/src/content/app/content_main.cc:35
    #25 0x7fc670fa3c97 in ChromeMain /media/Chromium/chromium/depot_tools/src/chrome/app/chrome_main.cc:32
    #26 0x7fc670fa3beb in main /media/Chromium/chromium/depot_tools/src/chrome/app/chrome_exe_main_gtk.cc:18
    #27 0x7fc66a3b3d8e in __libc_start_main /build/buildd/eglibc-2.12.1/csu/libc-start.c:258
0x7fc65ad84c88 is located 8 bytes inside of 144-byte region [0x7fc65ad84c80,0x7fc65ad84d10)
freed by thread T0 here:
    #0 0x7fc678d439d2 in operator delete(void*) ??:0
    #1 0x7fc673136b39 in v8::internal::RuntimeProfiler::IsEnabled() /media/Chromium/chromium/depot_tools/src/v8/src/runtime-profiler.h:50
previously allocated by thread T0 here:
    #0 0x7fc678d43852 in operator new(unsigned long) ??:0
    #1 0x7fc676655fe4 in WebCore::SVGStyleElement::create(WebCore::QualifiedName const&, WebCore::Document*, bool) /media/Chromium/chromium/depot_tools/src/third_party/WebKit/Source/WebCore/svg/SVGStyleElement.cpp:51
    #2 0x7fc6763fcf79 in WTF::PassRefPtr<WebCore::SVGStyleElement>::leakRef() const /media/Chromium/chromium/depot_tools/src/third_party/WebKit/Source/JavaScriptCore/wtf/PassRefPtr.h:161
    #3 0x7fc6763f6728 in WTF::PassRefPtr<WebCore::SVGElement>::operator WebCore::SVGElement* WTF::PassRefPtr<WebCore::SVGElement>::*() const /media/Chromium/chromium/depot_tools/src/third_party/WebKit/Source/JavaScriptCore/wtf/PassRefPtr.h:82
    #4 0x7fc6740eb7c8 in WTF::PassRefPtr<WebCore::SVGElement>::leakRef() const /media/Chromium/chromium/depot_tools/src/third_party/WebKit/Source/JavaScriptCore/wtf/PassRefPtr.h:161
    #5 0x7fc6754cc4fa in WTF::PassRefPtr<WebCore::Element>::leakRef() const /media/Chromium/chromium/depot_tools/src/third_party/WebKit/Source/JavaScriptCore/wtf/PassRefPtr.h:161
    #6 0x7fc6742fd3b6 in xmlParseStartTag2 /media/Chromium/chromium/depot_tools/src/third_party/libxml/src/parser.c:9126
    #7 0x7fc6743061c9 in xmlParseTryOrFinish /media/Chromium/chromium/depot_tools/src/third_party/libxml/src/parser.c:10847
    #8 0x7fc67430395c in xmlParseChunk /media/Chromium/chromium/depot_tools/src/third_party/libxml/src/parser.c:11625
    #9 0x7fc6754cac50 in WebCore::DocumentParser::isStopped() const /media/Chromium/chromium/depot_tools/src/third_party/WebKit/Source/WebCore/dom/DocumentParser.h:69
    #10 0x7fc6754c4965 in ~RefPtr /media/Chromium/chromium/depot_tools/src/third_party/WebKit/Source/JavaScriptCore/wtf/RefPtr.h:58
    #11 0x7fc678358279 in ~Deque /media/Chromium/chromium/depot_tools/src/third_party/WebKit/Source/JavaScriptCore/wtf/Deque.h:370
    #12 0x7fc6752a0e9f in WebCore::DocumentLoader::commitData(char const*, unsigned long) /media/Chromium/chromium/depot_tools/src/third_party/WebKit/Source/WebCore/loader/DocumentLoader.cpp:327
    #13 0x7fc67403b618 in WebKit::FrameLoaderClientImpl::committedLoad(WebCore::DocumentLoader*, char const*, int) /media/Chromium/chromium/depot_tools/src/third_party/WebKit/Source/WebKit/chromium/src/FrameLoaderClientImpl.cpp:1137
    #14 0x7fc6752a0abc in WebCore::DocumentLoader::commitLoad(char const*, int) /media/Chromium/chromium/depot_tools/src/third_party/WebKit/Source/WebCore/loader/DocumentLoader.cpp:313
    #15 0x7fc67533622e in WebCore::ResourceLoader::didReceiveData(char const*, int, long long, bool) /media/Chromium/chromium/depot_tools/src/third_party/WebKit/Source/WebCore/loader/ResourceLoader.cpp:292
    #16 0x7fc675313450 in WebCore::MainResourceLoader::didReceiveData(char const*, int, long long, bool) /media/Chromium/chromium/depot_tools/src/third_party/WebKit/Source/WebCore/loader/MainResourceLoader.cpp:464
    #17 0x7fc67533791b in WebCore::InspectorInstrumentation::hasFrontends() /media/Chromium/chromium/depot_tools/src/third_party/WebKit/Source/WebCore/inspector/InspectorInstrumentation.h:225
    #18 0x7fc673af77f2 in ResourceDispatcher::OnReceivedData(IPC::Message const&, int, base::FileDescriptor, int, int) /media/Chromium/chromium/depot_tools/src/content/common/resource_dispatcher.cc:404
    #19 0x7fc673af5f45 in ResourceDispatcher::DispatchMessage(IPC::Message const&) /media/Chromium/chromium/depot_tools/src/./content/common/resource_messages.h:155
    #20 0x7fc673af4290 in ResourceDispatcher::OnMessageReceived(IPC::Message const&) /media/Chromium/chromium/depot_tools/src/content/common/resource_dispatcher.cc:326
    #21 0x7fc6739ee75f in ChildThread::OnMessageReceived(IPC::Message const&) /media/Chromium/chromium/depot_tools/src/content/common/child_thread.cc:172
    #22 0x7fc672700e13 in IPC::ChannelProxy::Context::OnDispatchMessage(IPC::Message const&) /media/Chromium/chromium/depot_tools/src/ipc/ipc_channel_proxy.cc:268
==4607== ABORTING
Stats: 48M malloced (35M for red zones) by 74186 calls
Stats: 1M realloced by 1905 calls
Stats: 45M freed by 58757 calls
Stats: 0M really freed by 0 calls
Stats: 116M (29716 full pages) mmaped in 29 calls
  mmaps   by size class: 8:65532; 9:16382; 10:8190; 11:2047; 12:1024; 13:1536; 14:256; 15:384; 16:64; 17:64; 18:16; 19:8; 20:8; 22:5;
  mallocs by size class: 8:57634; 9:8577; 10:4433; 11:1317; 12:378; 13:1335; 14:144; 15:254; 16:44; 17:46; 18:12; 19:2; 20:5; 22:5;
  frees   by size class: 8:43495; 9:7992; 10:4138; 11:1065; 12:290; 13:1313; 14:127; 15:247; 16:37; 17:29; 18:12; 19:2; 20:5; 22:5;
  rfrees  by size class:
Stats: malloc large: 70 small slow: 415
Shadow byte and word:
  0x1ff8cb5b0991: fd
  0x1ff8cb5b0990: fd fd fd fd fd fd fd fd
More shadow bytes:
  0x1ff8cb5b0970: fd fd fd fd fd fd fd fd
  0x1ff8cb5b0978: fd fd fd fd fd fd fd fd
  0x1ff8cb5b0980: fa fa fa fa fa fa fa fa
  0x1ff8cb5b0988: fa fa fa fa fa fa fa fa
=>0x1ff8cb5b0990: fd fd fd fd fd fd fd fd
  0x1ff8cb5b0998: fd fd fd fd fd fd fd fd
  0x1ff8cb5b09a0: fd fd fd fd fd fd fd fd
  0x1ff8cb5b09a8: fd fd fd fd fd fd fd fd
  0x1ff8cb5b09b0: fa fa fa fa fa fa fa fa

tc-20-03-12-uaf.zip 637 bytes Download

Unfortunately cluster-fuzz still isn't giving us any information, so your ASAN stack trace is still the most information we have.

I can reproduce this fairly easily in debug on Windows, however. It looks like a CSSStyleSheet object has a stale pointer in m_ownerNode.

Ax330d's ASAN dump says it is freed by the runtime profiler?

arv can you have a peek at this again?

Assigning to arv, since he's been looking at it up to this point and it should really get attention since it's a reproducible crash. If you need any V8 help, please let me know.

This seems to have stalled. It looks like high-severity based on the report.

@arv - Are you working on this, and if not, do you know who should be?

I'll look at it again.

The test case seems to do a lot of thing that should not matter. I'll see if I can repro this in ASAN and scale back the test further.

One more thing. This is not a binding issue. It just a use after free.

It seems like the timer is calling on a svg element that has already been freed?

My gut feeling is that we are not correctly invalidating the StyleSheetList when we removed the nodes.

I managed to reduce this further. It no longer uses a range and it now uses an HTML file (SVGStyleElement has the same removeFromDocument as HTMLStyleElement).

test-case.zip 992 bytes Download

https://bugs.webkit.org/show_bug.cgi?id=83605

https://bugs.webkit.org/show_bug.cgi?id=83605

http://trac.webkit.org/changeset/113887

Does anyone happen to know if this affects M18 stable? Or just an M19 regression?

Thanks Ax330d for finally getting a awesome repro to reproduce this bug. It was really helpful to see the root problem. This qualifies for $1000 Chromium Security Reward.

M19: http://trac.webkit.org/changeset/115611

Updating status to Fixed on security bugs which were fixed when m19 went to stable.

This issue has been closed for some time. No one will pay attention to new comments.
If you are seeing this bug or have new data, please click New Issue to start a new bug.

This bug has been closed for more than 14 weeks. Removing security view restrictions.

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot

This bug has been closed for more than 14 weeks. Removing security view restrictions.

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot