VM Ware Fusion potentially causes macOS 10.15.6 to crash

[Alex_Ma]

I recently upgraded to macOS 10.15.6. Now if I leave VM Ware Fusion running while leaving my mac for some time, e.g. - for a night, the system crashes. This did not happen with the older macOS 10.15.5 version.

VM Ware Fusion version: Professional Version 11.5.5 (16269456)

macOS Catalina version: 10.15.6 (19G73)

[deepybee]

Alex_Ma​

Really? I thought Linux Kernel is being monitored pretty well. By nature of being open source, it is much less likely to catch a bug like this. And if it does, it will be fixed quicker.

The fact that macOS is also a Unix version, in fact it is BSD Unix for large part, will make the transition easier.

The xnu kernel in macOS and iOS has always been open source​, along with a load of other components, some lifted from FreeBSD and others Apple authored. Specific userland libraries and frameworks are closed source and what makes macOS distinct from vanilla Darwin.

I know that Apple actually open sourced it's kernel, but there is a catch - they never open source current code, only up to a version before.

The 10.15.6 kernel source is here ​and kernel development documentation here.

You can easily browse the lot at Open Source - Releases

[Alex_Ma]

Yes, I also had a hard crash while running Time Machine, and VM as not running actually. But I do not keep it mounted. I do backups regularly, every couple of weeks or so.

[pelletierr]

I had crashes on 4 different Macs (3 VMs and 1 Mac Mini 2014). All of them are without Fusion and all of them run Time Machine.

[kigoi]

fingers crossed that mitigating this becomes a lost art soon, there's two ways time machine or spotlight could be involved:

1. they can use a lot of memory when they work hard, and this could tip a very leak-sick session over the crash cliff
2. run full time, hourly, over the course of a session, they might also be triggering the leak itself

edit: yes i think tm/spotlight also trigger the leak, so i'm unmounting the local time machine volume during unattended sessions (overnight).

[synriga]

Had a kernel panic event yesterday that invalidated my Time Machine backup (several years worth) . Of darn!

I'm hoping Apple releases a fix for this soon.

[sergiorru]

Almost four days in a row and no crashes using Big Sur + Fusion Experimental.

The only thing that i noticed is the Guest VM is a little bit slow, but nothing compared with the anoying watchdog kernel panic issue!

[kigoi]

reposting me from virtualbox boards:

execute sudo zprint -d. If the output contains kalloc.32 every few seconds, then your scenario is affected.
Up to now, it is known that Oracle VirtualBox and VMware Fusion are affected, and it is assumed that Parallels Desktop and the macOS Hypervisor.framework are not affected.

this is not a safe assumption. i just ran the zprint, watched it sit quiet awhile, then intiated a time machine backup: kalloc.32, kalloc.32, kalloc.32.

can see the dumb headlines now: DO BACKUPS DAMAGE YOUR COMPUTER?

or: TEENAGER CRACKS NSA WITH MYSTERIOUS BKUP.NOW MALWARE

[Mikero · Accepted solution]

Looks like Apple has a fix for us:

https://www.macrumors.com/2020/08/12/apple-releases-macos-10-16-5-supplemental-update/

macOS Catalina‌ 10.15.6 supplemental update includes bug fixes for your Mac.

- Fixes a stability issue that could occur when running virtualization apps
- Resolves an issue where an ‌iMac‌ (Retina 5K, 27-inch, 2020) may appear washed out after waking from sleep

[kigoi]

seems to be working right. i watched various things, including zprint.

1. big post-update time machine didn't trigger problem
2. running vm awhile didn't trigger problem
3. wired memory returned to original level after those jobs quit

what a frikkin relief this is!!!

still before i reinstate my full setup, i'm gonna watch memory pressure a day or so, until convinced.

[Alex_Ma]

Terrific! Running the update now. Thanks VM Ware team for help. Great job.

[mctavish]

Yes, thank you VMware. Thank you, Apple too, that was a fairly quick fix.

[amjbecker]

That's an awfully big update (2.6 GB by my measurements) just to downgrade the kernel:

19G73 (10.15.6):

Darwin Kernel Version 19.6.0: Sun Jul  5 00:43:10 PDT 2020; root:xnu-6153.141.1~9/RELEASE_X86_64

19G2021 (10.15.6 w/supplemental):

Darwin Kernel Version 19.6.0: Thu Jun 18 20:49:00 PDT 2020; root:xnu-6153.141.1~1/RELEASE_X86_64

[kigoi]

looks like they had an interesting few weeks

[ColoradoMarmot]

Ditto here.  And from now on I'm keeping a full installer before doing a combo update.

[ktalley_tn]

Installed the supplemental upgrade (for 10.15.6) yesterday at 6pm CST.  This morning, I was running latest VMWare Fusion (and a Linux Mint v/m) for about 1.5 hours -- suddenly fan started blowing loudly, and my 2019 MBPro shut itself down after 2-3 minutes.  Was not able to even do any diagnostic during the fan blowing because the system was virtually unresponsive.

My older 2015 MBPro (running Big Sur beta and VMWare Fusion Tech Preview) is humming along quite nicely for a couple of days now....

Ultimately, I may simply have to send the 2019 MBPro back to my company and have them return to Apple for diagnostics (just to rule out some hardware issue).  I still strongly suspect o/s, but if this latest update didn't fix it, who knows...

[amjbecker]

Did it generate a panic log?

If so, would be interested to see what was in it.

[kigoi]

from now on I'm keeping a full installer before doing a combo update.

it's one of those few times i truly wish apple was more forthcoming.

obviously coronavirus countermeasures interfere with team QA, adding pressure, complexity & plain old distraction. like trying to do your normal business from the street after a building evacuation. you do eventually get better at it but the situation's not helping you.

is this, as i saw claimed, the first monster kernel bug they've had? is there reason to think that's a coincidence, given upheaval in workflows?

[ktalley_tn]

Here is a screenshot of the top part -- not sure if this gives you enough to "see" or not (to be useful).  After the restart, I *did* uninstall Virtualbox completely.  I see kext[s] loaded for this in the dump, and since I don't use it, there is really no need for it to be adding to any confusion.  I am about to start up Fusion again, so we will see what happens -- I will report back with any useful results or updates:

[amjbecker]

Obvious conjecture, but with the fact they reverted to an actual earlier build of the kernel it makes me wonder if:

1) They couldn't find the source of the bug and just reverted to a known good build.

2) It would be too much work to correct the code and issue a new build so decided to use an old one.

Obviously there were 8 more builds and ~2 weeks between the two versions. So I wonder if we're missing anything else they improved/enhanced.

[amjbecker]

Well it's not the same kind of crash as previous (good thing maybe?). It's a page fault. Let's hope it's not another regression in the kernel.