Latest ElCap upgrade hanging 10.11.6 - "Stalling for detach"

Heeday D. wrote:

Sorry chrde-san, I didn't recognize your post was in reply to aprilten-san...

No problem!

I did some further digging. I observed that a huge amount of IOAccelMemoryInfoUserClient resources is allocated by a process called UserEventAgent.

ioreg -lfr -c IntelAccelerator -w0

in Terminal shows it all. Replace IntelAccelerator with the name of your graphics card if not Intel.

This morning I was up to:

+-o IntelAccelerator <class IntelAccelerator, id 0x100000432, registered, matched, active, busy 0 (10922 ms), retain 1044>

and the problem came up again.

I then killed the UserEventAgent.

sudo kill -9 <pid>

and the system recovered! The UserEventAgent restarts immediately but the resources are freed.

You have to insert the correct pid for the UserEventAgent owned by root. There are running two (or more) instances of UserEventAgent.

While I am not sure about the side effects you can at least regain control of the machine until you are willing to reboot.

Anybody knows what UserEventAgent does and why it allocates so many IOAccelMemoryInfoUserClient?

What I have also noticed in console today is when the hang was happening I have a System Diagnostic report for each time it happened under the name "WindowServer_2017-01-05-090712_Seans-MacBook-Pro.userspace_watchdog_timeout"

There's seven of those leading up to when I rebooted. Apple support as you said will not be able to resolve this. We escalated it as well and I gave them this thread which they read, so they're onto our plight.

I went through the security update doc to try and pinpoint what I thought it could be (kernel) but still haven't been able to logically find a reason.

These are the ones I am looking more into:

AppleGraphicsPowerManagement

CoreCapture (this, Bluetooth & LibreSSL are the only ones that have 10.11.6 in the availability)

CoreGraphics

Intel Graphics Driver

IOHIDFamily

As you say, best bet is to try to narrow it down. I'll monitor my system closely to see if it happens again.

I think we can all agree this is a weird one. Let's hope its resolved soon.

P.S. have a look at this thread. I think this may help.

Mac Pro locks up, Watchdog Timeout Spin report?

I have the same symptoms, i.e. stalling for a minute or so, every few minutes,

on an MacBook Air (mid-2013) running OS X 10.11.6 (15G1212, so it's after the Security Update 2016-003),

rendering the MBA useless. Console shows "stalling for detach from IntelAccelerator" every time.

For me, it seems that the stalling happens when I change "Locations" (LAN configurations) from

a wired-LAN (sometimes connected to a M$FT Windows server) environment to

a Wi-Fi environment...

At the end of December, when this first started, I initially thought

newly installed Brother printer software and/or Firefox was the culprit, but it happens without either now.

In "Console," I now have 60 or so "WindowServer_201X_MM_DD-HHMMSS_ComputerName.userspace_watchdog_timeout.spin" reports dating back to December 20, 2016,

with 6 or 7 different dates -- meaning on days it doesn't happen, it doesn't.

Once it happens, the only cure is a restart -- I thought logging out and/or changing accounts solved it, but not...

This happens with a "fresh" account as well

FYI, there is an article that references this Apple Support Discussions thread.

https://www.jamf.com/jamf-nation/discussions/22595/avoid-security-update-2016-00 3-for-os-x-el-capitan-10-11-6

Just in case, no Photoshop, no VMware...

Sent "Bug Report" through Apple Feedback page:

http://www.apple.com/feedback/macbookair.html

Hope an upcoming software update fixes this...

I got too used to "trouble-free" Software Updates over these few years -- maybe I should get back to being more cautious with updates...

this is how it's done, but the results, it's a **** of a file, > 30 MB.

If you file a bug report, adding this file may help apple to debug/find the problem

QUOTE: instructions for an extended Bug Report:

Please provide sysdiagnose output from the affected system(s). Please collect this either while the issue is occurring, or as soon after as possible (for example, take it after a reboot if necessary).

Please let us know what you were doing that caused the issue, so we can correlate your actions with the diagnostic output in sysdiagnose. Please try to focus on one or two actions that cause an issue. If you perform many different actions, it will be difficult for us to determine which action caused the sysdiagnose output.

Note the date and time of the failure also, please.

sysdiagnose Instructions:

https://developer.apple.com/services-account/download?path=/OS_X/OS_X_Logs/sysdi agnose_Logging_Instructions.pdf

For a complete list of logging instructions visit:

https://developer.apple.com/bug-reporting/profiles-and-logs/

Some observations on Macs running 15G1212.

1. On my work Mac Mini 2012 the problem appears every 6 days. Only reboot fixes it.

2. On my second Mac, Mini 2011 AMD Graphics used as a Server, it takes about 8 days to occur. There is no VM installed, no browser running on the Server. So I believe the problem is not related to any application at all. It also appears not to depend on the graphics card driver.

3. Regarding to the kernel code mentioned by hughmcintyre there is this in:

https://opensource.apple.com/source/xnu/xnu-3789.21.4/iokit/Kernel/IOServicePriv ate.h.auto.html

kIOServiceBusyMax = 1023,

That looks really suspicious raised from 255 in 10.11.6 kernel. Could it be that they just forgot to raise the limit while backporting the check.

4. Apple needs to fix this really soon.

Another thread about this same problem:

Re: AMD Graphics stalling on 3 Macs with latest El Capitan udpate

It's all just me-toos, no discussion of possible causes. But it started after the last security update.

@ lotlorien

(1) If you check the "installations" under System Information (see About System Information on your Mac - Apple Support) you can sort the installations by date or name and check them.

(2) I can only report the current status, before the installation restart does not help, now it runs as smooth as before.

If it get worse again I will report this here ...

Same problem here, since the recent Apple security update. Most frequently the stall is triggered by one of my Java-based apps (a banking app, and JOSM, both running smoothly so far), and occasionally by Safari and others. Updated the Java JRE to the latest 1.8.0_111 with no improvement.

The freeze only affects the graphical session, i.e. the WindowServer. While frozen, I can connect from remote via ssh, and the shell remains fully responsive. Once I 'kill -9' the triggering process, the machine becomes responsive again.

Machine is a mid-2015 MBP retina, 10.11.6.

I installed 10.11.6 last week and this has been nailing me. The system is somewhat catatonic - mouse pointer moves, but clicks don't respond until they all get blasted out a few minutes later. Hot keys like cmd-left/right don't work, but pointer moves. sudo dmesg shows

stalling for detach from nvAccelerator

stalling for detach from nvAccelerator

stalling for detach from nvAccelerator

stalling for detach from nvAccelerator

Google Chrome He[65602] triggered unnest of range 0x7fff85200000->0x7fff85400000 of DYLD shared region in VM map 0x99653af5143537f7. While not abnormal for debuggers, this increases system memory footprint until the target exits. (Yeah, chrome is bloaty. This isn't a newsflash. But 10.11.6 is new in my world. Chrome hasn't changed.)

stalling for detach from nvAccelerator

stalling for detach from nvAccelerator

The messages correspond with the hangs. They're tripping me out because the symptoms match when I had a failing HD, but this message is new and smartctl -a disk1 isn't showing an avalanche of unrecoverable I/O errors.

There's some mention of VirtualBox in the threads. I have been running Parallels most of this week as I've been trying to make a new release of GPSBabel and this Mac is how I test all three OSes.

Everyone that's listed OS version has listed 10.11.6. Perhaps something was incompletely backported from 10.12. Is anyone running 10.12 affected?

I'm on a 2012 27-inch iMac. 3.4Ghz i7, 32GB of core and a 680MX.

Seems the frequency of stalls during a session increases, from triggered by a particular action in an app, to being triggered by any action, up to stalling with no interaction at all. At one point, the login session crashes:

Jan 3 02:15:31 hostname kernel[0]: stalling for detach from IntelAccelerator
Jan 3 02:15:46 hostname loginwindow[23724]: ERROR | -[LoginApp windowServerExited] | Window Server exited, closing down the session immediately
Jan 3 02:15:46 hostname thunderbird[24945]: HIToolbox: received notification of WindowServer event port death.

(and so on all user apps being terminated, system falls back to login screen)

I also have El Capitan 10.11.6 (15G1212) installed.

I hat this stalled error three times now. Only way to fix is a reboot. Also just for the record the machine was not under high usage that time.

03.01.17 14:28:04,000 kernel[0]: stalling for detach from nvAccelerator

03.01.17 14:28:07,000 kernel[0]: stalling for detach from nvAccelerator

03.01.17 14:28:22,000 kernel[0]: stalling for detach from nvAccelerator

03.01.17 14:28:41,000 kernel[0]: stalling for detach from nvAccelerator

03.01.17 14:29:10,000 kernel[0]: stalling for detach from nvAccelerator

03.01.17 14:29:25,000 kernel[0]: stalling for detach from nvAccelerator

03.01.17 14:37:29,701 watchdogd[345]: [watchdog_daemon] @(_wd_daemon_service_thread) - service (com.apple.WindowServer) reported as unresponsive

03.01.17 14:37:34,000 kernel[0]: stalling for detach from nvAccelerator

I already removed VirtualBox from the system by using the provided uninstaller script.

any help would be appreciated

Update from my above post.

Both my MBP and Mini have "OS X 10.11.6 (15G1212)" and I'm seeing the issue on them.

A colleague with a very similar Mini has "OS X 10.11.6 (15G1108)" and is not having this issue.

My wife has a MBA and it also has "OS X 10.11.6 (15G1108)" and not having this issue.

Two other colleagues have macs, but they're running 10.12 and aren't seeing it.

I am experiencing this on my "iMac (27-inch, Late 2013)". No VirtualBox. Haven't been able to pinpoint an exact cause, but when it happens I need to reboot ASAP or risk losing my work because the kernel won't stop stalling.

Looking at gbbgu's and other's build numbers, those with "OS X 10.11.6 (15G1212)" do have the bug, including mine.

gbbgu's "OS X 10.11.6 (15G1108)" is apparently the build before the December update, not having the bug.

As for other updates installed simultaneously, we can exclude iTunes, as I do not use nor update it on my affected machine, and Xcode, which I still have in v.7.2 from 2015. I updated Safari and the Xcode-cmd-line-tools, however.

Since I rebooted after the last user session crash, uptime is about 24h, the first symptoms are starting again.