Apple Event: May 7th at 7 am PT

Looks like no one’s replied in a while. To start the conversation again, simply ask a new question.

Latest ElCap upgrade hanging 10.11.6 - "Stalling for detach"

The most recent upgrade to ElCap (10.11.6) has resulted in the OS regularly hanging for one or more minutes. It has been reported on 2012 machines, and lately on more recent machines. Sometimes one can still move the mouse pointer during the hangand from time to time multiple actions will take place, and then the hang will resume.


The message in the Console is:

kernel: Stalling for detach from IntelAccelerator

or

kernel: Stalling for detach from nvAccelerator


There has been some discussion that this could be related to VirtualBox (Re: Stalling for detach from IntelAccelerator, nvAccelerator), but it also takes place on systems with no such installation.


It smells like an OS problem. How does one get someone at Apple to address this?

Posted on Dec 30, 2016 12:49 PM

Reply
Question marked as Best reply

Posted on Jan 3, 2017 2:31 PM

As it has been noted above, this was not introduced with 10.11.6, but likely with the updates Apple issue last month.

I have appended a screen shot of the updates I installed in December. Based on the comments, I am guessing that the culprit, as has been suggested, was the Security Update 2016-003, and I would be curious, gbbgu, if those you say are running 10.11.6 and not having the problem, have as yet to install that update, and whether the issue appearsafter they did an update..


As you can see it is not possible for me to differentiate among the 4 OS updates I installed on the 13th. Moreover, the issue did not appear until I was pushing quit a few high graphic windows, and the problem then disappears on rebooting until one gets to that point again....



User uploaded file

111 replies

Jan 4, 2017 1:32 PM in response to Robert Lipe

I was able to get a repeatable hang while using Outlook 365. Safari was also running. After UI became responsive I quit Safari, and the problem disappeared. The console log entries make me think that the WindowServer was somehow corrupted by Safari. I can load the log (from which I've stripped an enormous amount of un-responded-to pings) if there is any interest.

Jan 4, 2017 2:52 PM in response to cgn

Two machines, MacBook Air and Mac Pro, Both having OS X 10.11.6 (15G1212), have this problem.


Error messages from system.log:


Jan 4 17:24:44 MacPro kernel[0]: stalling for detach from AMDTahitiGraphicsAccelerator


Jan 4 14:05:49 MacBookAir kernel[0]: stalling for detach from IntelAccelerator


Switching from Safari to Chrome can provide some relief from the problem, but other apps such as VLC may sometimes trigger it.


Safari is Version 10.0.2 (11602.3.12.0.1)

Jan 4, 2017 5:17 PM in response to net-buoy

I have been having the same issue on my mid-2012 non-Retina MacBook Pro that has both Intel graphics and NVidia GPU, with OS-X 10.11.6. The issue is not CPU load or memory load; I can watch it happen with only a couple of gigabytes of user application and only background tasks engaged in any activity. This last time I happened to be able to start Activity Monitor and noticed the spindump executing which led me to the log files that showed the detach messages for nvAccelerator.


As some others described, I was able to ssh in and the system was responsive.


Earlier today (before I saw the log entries), I got control back by shooting Firefox, which was the largest application at the time. More recently, I got control back by shooting Maple, which uses a Java user interface. In both of those particular cases, Maple was running. I seem to recall that I have also had this difficulty when Maple was not running but when MATLAB was running; MATLAB also uses a Java user interface. It is possible that I am misremembering: it is possible that I did have Maple running at the time as well (but shooting MATLAB was what did the trick then.)


I do not have VMWare running on my system at all. I do have Parallels, but it was not active this afternoon.


So, the commonalities are on the 2012 machine, and with a Java-based program executing at the time of the trouble. Killing one of my user processes gets it un-hung, but it is not necessarily the Java-based process that needs to be killed in my experience.

Jan 4, 2017 7:50 PM in response to sandra-leigh

I got a request from Apple for more info in my bug report today, so will see what response I get. As with others, the failing build is 15G1212.


One other datapoint is that if you look at the kernel source code at opensource.apple.com, the "stalling for detach" message is only in the source code for 10.12.1 (Sierra), not 10.11.6 (El Capitan). However, the 10.11.6 source code is "xnu-3248.60.10" while a system running build 15G1212 seems to be "xnu-3248.60.11.1.2~2".


This seems to confirm that the original 10.11.6 was probably OK, but also implies that whatever changed between xnu-3248.60.10 and xnu-3248.60.11.1.2~2 may have caused the bug. Luckily this is a small version number change, so hopefully this makes it easier to find the problematic change.


It also implies that upgrading to Sierra may not help though 😟

Jan 4, 2017 8:19 PM in response to sandra-leigh

A short time ago, I was beginning to be affected again while using MATLAB, but originally not badly. I seemed to be having graphics problems (graphics not showing up) so I quit MATLAB. At that point I went into the stall. The only user applications I was left running were Firefox and Slack. I ssh'd in, shot firefox, and got back control immediately.


Checking around, I find that firefox is not written in Java at all, so it is not clear that the problem is Java. However, as MATLAB is Java based, it just might be the case that using Java somehow triggers the problem but that then quitting that Java program might not (always) be enough. It could be the case, for example, that what is required is to quit some windowed program and that could hypothetically trigger a state cleanup .

Jan 4, 2017 9:05 PM in response to sandra-leigh

I am not getting stalling for detach from IntelAccelerator where I was not before. And now the only user application I am running is Firefox.


I can also now seem to provoke Firefox to hang by asking to Save Link As of the MCC 6220 User Manual on the page http://www.mccdaq.com/products/6220.htm -- I get a save window that is has a thick black stripe on the left and is otherwise empty and gives the spinning beach ball. My system stability seems to be deteriorating, so I am going to reboot right after posting this.

Jan 5, 2017 8:42 AM in response to net-buoy

If this helps anyone, I've found that I can fix the stalling/freezes by quitting Photoshop (2017.0.1). I'm wondering if this is because Photoshop uses the GPU and is really just the manifestation of a bug on the macOS graphics layer. So, for some it might be another app that is using the GPU for accelerated graphics, and causing the issue. I've seen some can fix the problem by quitting Safari. Your mileage may vary. And it's definitely no a permanent solution, as the problem comes back, but a quicker bandaid to apply than having to restart the entire system.

Jan 5, 2017 11:53 AM in response to net-buoy

So it looks like we are all experiencing the same issue. I installed the update on the 18th. Haven't restarted my machine since then. Today I came back to it after not using it since Xmas day and as soon as I started to work there was the problem. I quit safari, used firefox, same issues. I quit outlook and opened only safari, same issue. I have been scouring the security update document line by line to see if I can find the culprit. Also speaking with apple but they just want me to run disk utility.

Jan 5, 2017 5:40 PM in response to Sean Field

Hey everyone. Quick update. Spoke to apple tech on the phone for a while today trying to narrow this down. We could find nothing. We did notice that no one here mentioned booting into safe mode to see if the issue could be recreated or went away. For me after a reboot (before safe mode) it went away. Of course this doesn't mean that we're out of the woods but just a quick FYI.

Jan 5, 2017 6:33 PM in response to Sean Field

Some observations:
* The bug is NOT dependent on a specific application.
* The bug does seem to be related to apps tat are switching windows with intensive graphics or local acceleration
* Rebooting does appear to cause the bug symptoms to cease. The symptoms come back.
* The symptoms effect the GUI and apparently do not impact non-gui sessions (terminal, ssh, etc)
* The symptoms appear to cascade
* The bug was introduced in the 2016-003 update
* Unless turned off, the bug symptoms will result in multiple and continuing dumps.

* The bug does not seem to gobble up CPU resources
* The suggestion that this is an arbitration or race issue has a lot going for it.

Latest ElCap upgrade hanging 10.11.6 - "Stalling for detach"

Welcome to Apple Support Community
A forum where Apple customers help each other with their products. Get started with your Apple ID.