Windows 10 guest crashes after boot since VBox 7 update

[Greygeek]

Had to bail for a great dinner & show last night - not a slow modem And today, had to go where there was a working Windoze system so I could actually get some work done.

I downloaded another fresh copy of VB 7.0.10 and Guest Additions, and I re-installed Virtual Box app on my Linux Host. (same Ver 7.0.10), rebooted everything, then started VirtualBox immediately after the host reboot. (max RAM available)
No change. Still won't boot (except in Safe Mode) with Guest Additions installed.

I recall reading that trying to run a Windows guest additions version that doesn't match the currently installed Virtualbox version installed on the host is asking for trouble. Can you confirm that? (i.e. is it worth trying to install a 7.0.4 or 7.0.6 version of guest additions to see if it behaves better?)

If I revert to VBox Version 6.xxx, then I'll have to manually edit all of my Linux OS security updates to remove updates to the "old" version of Virtual Box back to the broken version 7.0.10. If I can avoid that PITA, it'd be great.

[scottgus1]

I recall reading that trying to run a Windows guest additions version that doesn't match the currently installed Virtualbox version installed on the host is asking for trouble.

Most of the time, no. Unless a new feature requires up-to-date GAs, the GAs can usually be older. 3D acceleration may require up-to-date GAs, if changes to 3D were made. Changes to 3D are being made just now, so older GAs may or may not work. But you can try. 7.0.8 GAs had some bugs, but the link to 7.0.6 to get the GAs is above.

The Extension Pack is what needs to match the Virtualbox version.

[fth0]

Since the previously provided VBox.log file exhibited really unexpected content, perhaps you could provide a new (zipped) VBox.log file for comparison.

[Greygeek]

Hi -
I've attached a log file dump from last evening after the clean re-installs and boot. Same symptoms: The Win 10 VM boots up, I get the splash screen where I use space bar to reveal the login (password entry) box, where I type the Windows login password. I get the spinning "wait" signal for about 4 seconds before the VM is aborted (visually, it just disappears from the screen and I'm left looking at the Linux Host desktop with the Linux Virtual Box application waiting for me to launch the VM.

I have not had time today to try loading a different (7.0.06 for example) version of Guest Additions. I do know that the Ver 6.1.40 guest additions I ran before updating it did not work right. Will try to do that soon.

[Greygeek]

Good Morning ...
I got the VM up in Safe Mode so I could un-install Guest Additions 7.0.10 and then got GuestAdditions 7.0.06 into my Linux Host's directory and set up to install. I booted up Windows (with no Guest Additions installed, it booted up in normal full mode), then ran the installer for GuestAdditions 7.0.06, and the installer failed to finish before the abort took place. So it's same-old, same-old, error except that my Host Linux spewed out these messages. Likely triggered by the incompatibility of the VBox 7.0.10 application and the 7.0.6 GAs since this is the first time I've seen these.

Message from syslogd@GlacierMX at Aug 24 00:30:00 ...
kernel:[45950.332756]

Message from syslogd@GlacierMX at Aug 24 00:30:00 ...
kernel:[45950.332756] !!Assertion Failed!!

Message from syslogd@GlacierMX at Aug 24 00:30:00 ...
kernel:[45950.332756] Expression: RT_SUCCESS(rc)

Message from syslogd@GlacierMX at Aug 24 00:30:00 ...
kernel:[45950.332756] Location : /build/virtualbox-7.0.10-dfsg/src/VBox/VMM/VMMAll/PGMAll.cpp(3850) int pgmLock(PVMCC, bool)

Message from syslogd@GlacierMX at Aug 24 00:30:00 ...
kernel:[45950.332764] pVM=fffface28ab80000 pCritSect=fffface28ac86760: -39

I much appreciate any further insights you can provide.
I don't know at what point I can actually get Oracle's attention to apply some engineering bandwidth to get their product working again, but my Virtual Windows 10 has been down for nearly a month now. All this has been working so darned well for over 2.5 years, then an update to VirtualBox just blew it all up.

I attached another log file with an August 24 date (after installing GAs ver 7.0.6). All of these log files end with an entry "EHCI: USB Suspended". So what is VB actually *doing* when it is trying to "suspend USB" ?

Thank you!

[scottgus1]

ran the installer for GuestAdditions 7.0.06, and the installer failed to finish before the abort took place.

Since the VM aborted before the GAs installer completed with the required OS reboot, it appears that the Guest Additions are not causing the abort. You did succeed in running the VM for some time without Guest Additions, maybe the OS didn't do the thing that causes the abort. Good, we can uncheck GAs as a cause.

Next possibility: there could be something wrong in the OS itself. From your OP I don't see if you have any other VMs? If you do , do they exhibit this occasional abort?

Also, please try making a new Windows 10 VM, no need to activate the OS, just see if the new VM aborts over time.

BTW, do you have backups of the data in the VM? If you can get it to run long enough, a Macrium Reflect Free disk image from inside the VM OS may not come amiss. Recovering data from outside a snapshotted VM can be a bit hairy if an inside-the-OS disk image is not available.

[fth0]

I'll try to catch-up with the new info in this thread one by one:

I've attached a log file dump from last evening after the clean re-installs and boot.

00:00:01.237391 VirtualBox VM 7.0.10_Debian r158379 linux.amd64 (Aug 19 2023 13:35:49) release log

First of all, you're not using a VirtualBox version provided by Oracle, but a VirtualBox version provided by Debian, and I don't know the differences. Since it is ~1 month newer, I'd guess that it already contains modifications developed after Oracle's release. Those could be bugfixes provided by Oracle, though.

00:00:01.261466 Guest OS type: 'Windows2016_64'

The wrong guest OS type could play a role, too. But you know that already.

00:00:01.308496   MonitorCount     <integer> = 0x0000000000000002 (2)

If you provide 2 virtual screens to a VM and if the second virtual screen is already disabled when the VirtualBox Guest Additions 7.0.10 are started, the first virtual screen is erroneously disabled, too, and this results in a black screen. This is a known issue, solved in newer test builds (since ~2023-07-27) and perhaps solved in the Debian VirtualBox version.

---

All of these log files end with an entry "EHCI: USB Suspended".

You could try if providing the USB 3.0 Controller to the guest OS makes a difference. It is allegedly better even when only capturing USB 1.1 or USB 2.0 devices. Additionally, if you have any optional USB devices attached to the host, you could (for a test) remove them before starting the VM.

---

Please provide a zip file with the contents of the VirtualBox global configuration folder (~/.config/VirtualBox).

[Greygeek]

scotgus1 -
Thank you for your post and advice to install / run Macrium to create and image from within Windows. I do have backups of the important data (Cloud backup service), but I've used Macrium Reflect in the past and it works well.

You wrote:
Since the VM aborted before the GAs installer completed with the required OS reboot, it appears that the Guest Additions are not causing the abort. You did succeed in running the VM for some time without Guest Additions, maybe the OS didn't do the thing that causes the abort. Good, we can uncheck GAs as a cause.

Whoa - not so fast. GAs interacts with the VBox linux host app, and that host app is keenly aware of the status (*Aborted*) of the VM. A Windoze crash doesn't leave an *Aborted* status moniker in the VBox linux app. (there's also zero entries in Windows event logs indicating a crash - it's just like a power fail - instant death, not a crash that records errors in log files). I can install & run GAs from Safe Mode no problem, (i.e. mounted host drives appear, the 3D graphics controller is running, etc.) but when I do, any subsequent reboot to full mode terminates right after login (as Windoze is loading modules required after a login ..)

Creating another test VM provides data, but won't fix the problem. I'm thinking that my next step is to re-install VirtualBox ver 6.1.40 which was running fine before this VBox update to version 7 blew it app up. If that doesn't work using a VirtualBox .iso file downloaded from the Oracle site, then creating a new VM (if it works) might be a path toward getting back to work. I just don't have enough disc space to do that properly for longer-term use.

[scottgus1]

Shortness of disk space could be a problem, yes. Have you 30GB you can borrow for a few days? The purpose of the new VM is to see if it also aborts. If it does, then the problem is likely not inside your existing VM and possibly on the host. If it does not abort, it could be your existing VM OS is trying something very unusual that Virtualbox really doesn't like. Finding out if the problem is with the VM OS is an important part of troubleshooting this problem.

Re Aborts vs BSODs:
The VM actually consists of a "hardware" container and the OS installed within the container. The OS can crash horribly while the container stays up and running.

An "abort" is not the same as a Windows BSOD. BSODs can happen and the VM container stays running. An abort is when the container crashed. Aborts can happen, though from both host troubles and VM OS's doing things far out of bounds, if I understand correctly. So the fact that your VM aborts is not a definitive-enough data point to decide why this is happening. The test VM can clear the matter up with higher probability.

Of course, if you want to go back to 6.1.40 and leave the question unanswered, that's fine. However, you may not be able to stay on 6.1.40 for long. Linux keeps upgrading the kernel, and new kernels require new Virtualbox versions.

[Greygeek]

I created a new, small Windows 10 VM and the Version 7 VirtualBox still aborts it when guest additions is installed. Seems there is something just plain incompatible between Virtual Box 7.x.x and my Debian 11 Linux host and Windows 10 virtual machine. I'll replace VBox 7.0.10 with the older version of the entire VirtualBox Linux app and GAs when I get back from a short trip on Wednesday.

[Greygeek]

Sorry gents, I'm a bit overwhelmed with month-end accounting. I had to purchase a Windows laptop and set it up so I could get my work done while this darned Virtual Machine is down. I hope to be able to return the new device during the trial period ...

Meanwhile, I'm trying to make sense of how to manually install VBox on my Debian 11 System. I've found this part of the VBox manual quite confusing. I've obviously used package installers to get it installed, but now want to revert to Ver 6.1.40. I've downloaded the .deb file from Oracle/VBox site. I think it has all of the required files / drivers etc. and even Guest Additions.

https://www.virtualbox.org/manual/ch02. ... nux-prereq

More later ...

[scottgus1]

The biggest problems we've seen with Linux installs of Virtualbox are:

• failure to uninstall/purge the Linux fork of Virtualbox the distro preinstalls
• not installing the kernel headers for the Linux OS.

Usually, it seems that the Linux OS has Qt, SDL, gcc, and make already, though some posters have reported one or more of these missing on their host too. The second point about the kernel headers is done differently depending on the distro, thus the manual's confusing-sounding wordage.

Here's what I would do:
If you have the Extension Pack installed inside Virtualbox, run Virtualbox and uninstall the EP.
Uninstall/purge the existing Virtualbox. Your VMs and configuration will remain behind.
Run the prerequisites: make sure Qt et al are installed; regrettably, you'll have to figure out how to do the kernel headers. Don't ask me how, I'm a Windows guy.
Install the 6.1.40 Virtualbox from www.virtualbox.org, and the matching Extension Pack if you want it.
Try a VM and hope it works!

[Greygeek]

scottgus1 - thank you for your reply.

Finally ... had some time to work on this.
I un-installed the problematic VirtualBox Ver 7.0.10 using the Package Manager that is part of my distro, which enabled me to remove all of the files that were part of the install. I did leave the config files intact.

Then I re-installed Version 6.1.40 from a .deb file downloaded from the Oracle site. Ver 6.1.40 had been working just fine for me. At first it complained about the mis-matched extension pack, so I went back and installed the same version EP and I now have a completely functioning Windows 10 VM again. At last. The Windows VM shows no sign of corruption thank goodness. And I didn't have to futz with kernel headers, thankfully.

This is not an ideal solution, because I'll have to manually tweak each system update so it doesn't try to "upgrade" VBox. But I'll be migrating to Debian 12 with a new release of my distro later this year, so I'll get another chance at making VirtualBox work then. Meanwhile, I've added notes of my journey to that of other users on my distro's support forums who also couldn't make VBox Ver 7.x work right. Most have also given up and gone back to VBox Ver 6.x.x

I appreciate the tips and ideas you both have provided me. I just don't have any more time to debug Oracle software right now. I have a bootable Windows 10 VM again, after 33 days of messing around after VBox Ver7 was installed as part of a Host system update.

[scottgus1]

We're really glad you got things running again!

Maybe once the headers are installed for a later Virtualbox version they're good enough for earlier versions. Nice that whatever headers were there worked for your host.

If you can't turn off that automatic update thingy so you don't get side-swiped again, you may want to institute some sort of backup routine for your host. A couple of ideas:

Disk imaging (while all the VMs are fully shut down not save stated): Have a couple backup media disks you can plug in and image the host disk. Practice restoring the backup image to another drive, so the whole process is demonstrated to work.

Disk cloning: If the main disk is or can be made relatively easy to remove, make identical copies of the whole host disk on replacement disks. Then a restore is a direct disk swap.