New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.
Starred by 1 user

Issue metadata

Status: Verified
Last visit > 30 days ago
Closed: Oct 2016
EstimatedDays: ----
NextAction: ----
OS: All
Pri: 1
Type: Bug-Security

Sign in to add a comment

Security: UNKNOWN in v8::internal::GlobalHandles::Node::Release

Reported by, Oct 4 2016

Issue description

Chrome Version: 53.0.2785.116 stable
Operating System: Windows 7

1. Log into Facebook 
2. Navigate to
3. Click on the first result to open a popup window and wait until the page is ready
4. Close that tab >> Crash!

Type of crash: 878613ed00000000

rax=000007fed2a1b79c rbx=00000000045ea7e0 rcx=92085f0200010463
rdx=00000000002e7f90 rsi=0000000015d273e0 rdi=00000000001cde40
rip=000007fed0adff0c rsp=00000000001cddc0 rbp=00000000001ce0a0
 r8=0000000000000036  r9=006d070d01ce0063 r10=0000000078c7e20e
r11=00000000001cde90 r12=0000000000000000 r13=00000000001ce340
r14=000007fed300f970 r15=00000000002caca0
iopl=0         nv up ei pl nz na po nc
cs=0033  ss=0000  ds=0000  es=0000  fs=0053  gs=002b             efl=00010206
*** WARNING: Unable to verify checksum for chrome_child.dll
000007fe`d0adff0c 80610bf0        and     byte ptr [rcx+0Bh],0F0h ds:92085f02`0001046e=??
0:000> k
  *** Stack trace for last set context - .thread/.cxr resets it
Child-SP          RetAddr           Call Site
00000000`001cddc0 000007fe`d2a1b7c2 chrome_child!v8::internal::GlobalHandles::Node::Release+0x14 [c:\b\build\slave\win64-pgo\build\src\v8\src\ @ 118]
00000000`001cddf0 000007fe`d0b2b4cc chrome_child!blink::InspectedContext::weakCallback+0x26 [c:\b\build\slave\win64-pgo\build\src\third_party\webkit\source\platform\v8_inspector\inspectedcontext.cpp @ 20]
00000000`001cde20 000007fe`d0b2b310 chrome_child!v8::internal::GlobalHandles::PendingPhantomCallback::Invoke+0x40 [c:\b\build\slave\win64-pgo\build\src\v8\src\ @ 1097]
00000000`001cde80 000007fe`d0ce6ed7 chrome_child!v8::internal::GlobalHandles::InvokeSecondPassPhantomCallbacks+0x64 [c:\b\build\slave\win64-pgo\build\src\v8\src\ @ 949]
00000000`001cdee0 000007fe`d0b4f9d9 chrome_child!v8::internal::GlobalHandles::PendingPhantomCallbacksSecondPassTask::RunInternal+0x53 [c:\b\build\slave\win64-pgo\build\src\v8\src\ @ 562]
00000000`001cdf70 000007fe`d0cae6bd chrome_child!v8::internal::CancelableTask::Run+0x1d [c:\b\build\slave\win64-pgo\build\src\v8\src\cancelable-task.h @ 132]
00000000`001cdfa0 000007fe`d0caffb1 chrome_child!base::debug::TaskAnnotator::RunTask+0x19d [c:\b\build\slave\win64-pgo\build\src\base\debug\ @ 53]
00000000`001ce120 000007fe`d0cace0e chrome_child!scheduler::TaskQueueManager::ProcessTaskFromWorkQueue+0x149 [c:\b\build\slave\win64-pgo\build\src\components\scheduler\base\ @ 295]
00000000`001ce2b0 000007fe`d116d99a chrome_child!scheduler::TaskQueueManager::DoWork+0xf2 [c:\b\build\slave\win64-pgo\build\src\components\scheduler\base\ @ 203]
00000000`001ce700 000007fe`d0cae6bd chrome_child!base::internal::Invoker<base::internal::BindState<base::internal::RunnableAdapter<void (__cdecl scheduler::TaskQueueManager::*)(base::TimeTicks,bool) __ptr64>,base::WeakPtr<scheduler::TaskQueueManager>,base::TimeTicks,bool>,void __cdecl(void)>::Run+0x2e [c:\b\build\slave\win64-pgo\build\src\base\bind_internal.h @ 346]
00000000`001ce730 000007fe`d0cada9e chrome_child!base::debug::TaskAnnotator::RunTask+0x19d [c:\b\build\slave\win64-pgo\build\src\base\debug\ @ 53]
00000000`001ce8b0 000007fe`d0cade41 chrome_child!base::MessageLoop::RunTask+0x96 [c:\b\build\slave\win64-pgo\build\src\base\message_loop\ @ 494]
00000000`001ce990 000007fe`d0cadb99 chrome_child!base::MessageLoop::DoWork+0x1c9 [c:\b\build\slave\win64-pgo\build\src\base\message_loop\ @ 625]
00000000`001cec40 000007fe`d0f66492 chrome_child!base::MessagePumpDefault::Run+0x1d [c:\b\build\slave\win64-pgo\build\src\base\message_loop\ @ 36]
00000000`001cec90 000007fe`d0f66421 chrome_child!base::MessageLoop::RunHandler+0x3a [c:\b\build\slave\win64-pgo\build\src\base\message_loop\ @ 458]
00000000`001ced20 000007fe`d1061795 chrome_child!base::RunLoop::Run+0x79 [c:\b\build\slave\win64-pgo\build\src\base\ @ 36]
00000000`001ced70 000007fe`d117584f chrome_child!base::MessageLoop::Run+0x1d [c:\b\build\slave\win64-pgo\build\src\base\message_loop\ @ 296]
00000000`001cedd0 000007fe`d106001c chrome_child!content::RendererMain+0x1ef [c:\b\build\slave\win64-pgo\build\src\content\renderer\ @ 198]
00000000`001cef80 000007fe`d0e98d1a chrome_child!content::RunNamedProcessTypeMain+0xb4 [c:\b\build\slave\win64-pgo\build\src\content\app\ @ 435]
00000000`001cf0e0 000007fe`d0e98c49 chrome_child!content::ContentMainRunnerImpl::Run+0xae [c:\b\build\slave\win64-pgo\build\src\content\app\ @ 785]

1.2 MB View Download
Status: Assigned (was: Unconfirmed)
Possibly related to 646810.  Haraken, can you tell if this is the same issue?

jochen@: This looks like an issue of first/second weak callbacks used in InspectedContext. Would you find an owner for this?

(Maybe we can remove the second weak callbacks -- Blink has already removed them entirely.)

Components: Blink>JavaScript
Labels: OS-All
happening during the second callback. I wonder why the if (context->m_context.IsEmpty()) check returns false? We could just split up the callback in two differnt functions
Able to reproduce this issue under ASan-Windows bluid. This is a use-after-free vulnerability.
8.2 KB View Download
Labels: Security_Severity-High M-54 Security_Impact-Stable Pri-1
Khalil, could you symbolize the asan trace for us?
I couldn't repro this crash on Linux.
re #c2: what's the alternative for second pass callbacks used in blink? We would definitely switch if possible.

Comment 10 Deleted

Comment 11 Deleted

I prepared CL [1] that removes second pass callback from inspected context - it was added to provide additionally guarantee that we remove InspectedContext and related objects when context is gone. But we have another mechanism to remove it and as long as inspector client call contextDestroyed for every contextCreated - nothing will be leaked.


Comment 13 Deleted

Project Member

Comment 14 by, Oct 14 2016

The following revision refers to this bug:

commit 2082afcf3c8afe5002689ff3208d546f5775fb9a
Author: kozyatinskiy <>
Date: Fri Oct 14 01:59:36 2016

[inspector] added check that context always survives inspected context

Inspected context is created in V8InspectorImpl::contextCreated method and destroyed in V8InspectorImpl::contextDestroyed.
Both methods takes valid v8::Local<v8::Context> handle to the same context, it means that context is created before InspectedContext constructor and is always destroyed after InspectedContext destructor therefore context weak callback in inspected context should be never called.
It's possible only if inspector client doesn't call contextDestroyed which is considered an error.

Therefore CHECK(false) is added into context weak callback to be sure that v8::Context always survives inspected context.

BUG= chromium:652548

Cr-Commit-Position: refs/heads/master@{#40290}


Verified on 56.0.2891.0 (Windows). The crash is no longer seen.
Labels: -M-54 M-55
Status: Verified (was: Assigned)
Labels: Merge-Triage
Merge-Triage as we'll want to get it merged into M55 once it's had bake time.
Project Member

Comment 19 by, Oct 18 2016

Labels: -Restrict-View-SecurityTeam Restrict-View-SecurityNotify
Labels: reward-topanel
Project Member

Comment 21 by, Oct 21 2016

Labels: Merge-Request-55

Comment 22 by, Oct 21 2016

Labels: -Merge-Request-55 Merge-Approved-55 Hotlist-Merge-Approved
Your change meets the bar and is auto-approved for M55 (branch: 2883)
I'm not sure that we need to merge this change - it just make crash easier to detect by replacing SecondPassCallback with explicit CHECK(false).
But this check allows me to find root of issue:, and I think we need merge fix of it instead.
Labels: -reward-topanel reward-unpaid reward-500
$500 for this bug.  We classified this as a low quality report (see  Would have been great to have had a symbolized memory tool output and a clean repro.  Cheers!
Hmm... this qualified only for 500$ because I didn't provide the asan trace (symbolized). Anyway thanks for that!

Next time I'll pay attention for that.
Labels: -reward-unpaid reward-inprocess
Labels: -Merge-Triage -Hotlist-Merge-Approved -Merge-Approved-55
Looks like 656626 is already in M55, so removing labels from this bug.
Labels: Release-0-M55
Labels: CVE-2016-5213
Project Member

Comment 31 by, Jan 24 2017

Labels: -Restrict-View-SecurityNotify allpublic
This bug has been closed for more than 14 weeks. Removing security view restrictions.

For more details visit - Your friendly Sheriffbot
Labels: CVE_description-submitted

Sign in to add a comment