Windows 10 Guest on Linux host dies seconds after boot

[fth0]

FWIW, VirtualBox 6.1.34 does not support Linux kernel 5.18. The VirtualBox test builds 6.1.35r151573 (and newer) are supposed to support Linux kernel 5.18 on the host and in the guest.

Perhaps the actual border is between 5.17.11 and 5.17.12 ...

Please report back after you've tried VirtualBox 6.1.35r151573 (or newer).

[clecastilho]

I am experiencing the same problem with Fedora 36 host, 6.1.34vbox, and win10 guest. Tried reinstalling win10, getting the same results.

Not just Windows guests, but I don't think it's working for any OS. I tried CentOS 7, CentOS 9 Stream, Windows 10, 11 and also Fedora 36. None of them worked. The status bar at the bottom of the screen does not show any activity.

[DrWatson]

Hello,

I did some background search regarding this issue.
This issue is already discussed in the Linux Kernel and Virtual Box Bug tracker since beginning of May.
It is related to Kernel commit 6e8ec2552c7d which fixes security vulnerabilities.

As Red Hat user with 5.17.x. facing this issue now, they seem to got a backport patch from Fedora Kernel team. Anyway...installing test versions ( VB here ) as end user is always the worst way you can go and should be only done for a short test.


Patches for this issue are available since around 2 weeks. This are patches for latest stable release !

Patch provided by Suse Linux ( recommended ! )
build.opensuse[.]org/package/view_file/Virtualization/virtualbox/fixes_for_kernel_5.18.patch?expand=1

Patch provided by VirtualBox Devs directly ( some say this is still a bit buggy )
virtualbox[.]org/attachment/ticket/20914/vbox-linux-5.18.patch

There are 3 solutions available now:
1. Change back boot order to latest working Kernel
2. Patch Virtual Box ( or work with testing version if you are fine to work with unstable version )
3. Wait for new distribution maintainer package which include the needed patches ( check if it's maybe already in specific repo )

Greetings,
Henrik

[AGMS]

One oddity is that this bug only happens on some computers. I've updated the related bug report at https://www.virtualbox.org/ticket/20976 with some details.

By the way, if it is the CPU (not motherboard architecture or something else weird), what kind of CPUs do you see the problem on?

AMD X2 Athlon - works (circa 2008)
Intel i5-750 - fails (circa 2009)
Intel i7-4820K - fails (circa 2014)
Intel i5-10500H - works (circa 2020)

Anyway, nice to see from the previous message by DrWatson (viewtopic.php?f=7&t=106071&start=15#p519086) that the kernel people are working on it.

[nixuser]

For me, Linux (various) guests fail on anything later than kernel-5.17.11, all my h/w is failing.

[1] ASUS G750JS 1 x Intel(R) Core(TM) i7-4700HQ CPU @ 2.40GHz
Intel Corporation 4th Gen Core Processor Integrated Graphics Controller (rev 06)
NVIDIA Corporation GK104M [GeForce GTX 870M] (rev a1)
Fedora 36


[2] Intel NUC i7 NUC11PAH 1 x 11th Gen Intel(R) Core(TM) i7-1165G7 @ 2.80GHz
Intel Corporation TigerLake-LP GT2 [Iris Xe Graphics] (rev 01)
Fedora 36

[psfales]

Another data point is that it seems the problem may also be related to the host processor.

I spent a lot of time determining (as stated above) that the problem showed up between 5.17.11 and 5.17.12 on a system where cpuinfo reports:

model name : Intel(R) Core(TM) i7 CPU 870 @ 2.93GHz

Part of what confused me is that when the failing VM is copied to another machine, it works fine, even on 5.17.12 and 5.17.13. That machine reports

model name : Intel(R) Core(TM) i5-2520M CPU @ 2.50GHz

There are other differences between those machines, of course, so I don't know for sure that the processor is what makes the difference. But it seems like the most obvious one.

[fth0]

Perhaps the actual border is between 5.17.11 and 5.17.12 ...

I can confirm that Linux kernel 5.17.12 got the backport of the extensive random number handling changes (~100 commits, including 6e8ec2552c7d13991148e551e3325a624d73fac6), so I'd expect that VirtualBox 6.1.34 won't reliably run VMs on a host with Linux kernels 5.17.12+.

[AGMS]

Besides ticket https://www.virtualbox.org/ticket/20976 there's a possibly better one at https://www.virtualbox.org/ticket/20914 about this problem.

[jdholland]

There are 3 solutions available now:
1. Change back boot order to latest working Kernel
2. Patch Virtual Box ( or work with testing version if you are fine to work with unstable version )
3. Wait for new distribution maintainer package which include the needed patches ( check if it's maybe already in specific repo )

I wish to share my personal experience with others who come to this forum as I did. I hope that it will calm some nerves and save others time..

I currently run Debian Stable (Bullseye 11.3) on several machines as my bare-metal OS, as I have done for the past twelve years (not entirely true, as I usually upgrade to testing roughly six months before becoming stable).
I do have the Oracle Vbox repository in my sources list so that I run the latest version, currently 6.1.34-150636.1).

The morning of 11 June 2022 I was presented with a kernel upgrade from 5.10.0-14 to 5.10.0-15 on all my machines, which I installed. After rebooting, which I only do after a kernel upgrade or an upgrade to WINE, I found that all of my guests were misbehaving badly. My guests include multiple instances of Windows 10 and Windows 8.1 (my preferred Windows dd).

After spending much more time than I can afford to waste in troubleshooting, I booted Debian back into kernel 5.10.0-14. All was fine, and no damage was done to the guests.

My course Of action:
I agree with DrWatson. It seems to be a security patch that Oracle has not yet caught up with.
I am going to sit on my hands, continue to run 5.10.0-14, and wait for the next Vbox update from Oracle before I attempt to reboot into 5.10.0-15. Hopefully, it comes soon.

Cheers to all,
JD

[JR1234]

I'll say ditto to all the reported issues. The one twist I have to add while debugging: I'm on a laptop with a Synaptics TouchPad. With no other pointing devices attached via USB, things seem to run fine. However, once I attach a USB pointing devices AND USE IT, that's when the crashes start when interacting with the guest. I've tried 3 different devices - 2 external touch pads, 1 mouse. If it's attached, but I only use the Synaptics TouchPad, no crashes. I'm am not doing any magic in the USB settings in VirtualBox; issue occurs with both USB 2 and USB 3 support.

So for me it's:
Host kernel 5.17.11 regardless of pointing devices attached, guest Windows 10 is stable (present work-around for me)
Host kernel 5.17.12 or 5.17.13 plus USB pointing drive, guest Windows 10 is not so stable - especially using Chrome/Edge.
Host kernel 5.17.12 or 5.17.13 with just the built in Synaptics, guest Windows 10 seems stable.

[ingo2]

My guess is that the latest linux kernel received a greater security enhancement by a redesigned random number generator. That improvement is described in detail here: https://www.theregister.com/2022/03/21/ ... _improved/.

The improvement seems to be so important that it has been ported back to older kernels still widely used. To my understanding it has been applied in kernel 5.10.119+ (Debian) and 5.17.12+ (Fedora, SUSE).
For those of you familiar with C-sources the "random.c" module is said to contain extensive documentation.
Seems VirtualBox has to adapt something which will be done in 6.1.36?

[fth0]

The VirtualBox test builds 6.1.35r151864 (and newer) are supposed to support Linux kernels 5.18, 5.17.12+, 5.15.44+, 5.10.119+ and more on the host and in the guest. Please report back if it solves the problem.

[TNZ]

Hi all,
Virtualbox 6.1.34 doesn't work since 5.17.12 because of the 90 kernel commits on random module.
I must keep 5.17.11 to run my VMs.

About Linux 5.18 issue, it is due to the remove of a function in kernel (maybe moved I think) :
Headers 5.17.9 :

root:/usr/src# grep netif_rx_ni $(find linux-headers-5.17.9-amd64 -type f)
linux-headers-5.17.9-amd64/include/linux/netdevice.h:int netif_rx_ni(struct sk_buff *skb);
linux-headers-5.17.9-amd64/include/trace/events/net.h:DEFINE_EVENT(net_dev_rx_verbose_template, netif_rx_ni_entry,
linux-headers-5.17.9-amd64/include/trace/events/net.h:DEFINE_EVENT(net_dev_rx_exit_template, netif_rx_ni_exit,
linux-headers-5.17.9-amd64/Module.symvers:0x675aef42 netif_rx_ni vmlinux EXPORT_SYMBOL

Headers 5.18.0 :

root:/usr/src# grep netif_rx_ni $(find linux-headers-5.18.0-amd64 -type f)
root:/usr/src#

Even if Virtualbox 6.1.35 works with kernel 5.18, I wonder about random module issue (commits on 5.18.1) ... If one of you has tried 6.1.35 on kernel 5.18 ? Let's know

[MikeP ITC]

W10 Guest Crashing issue

VB 6.1.34
Fedora 5.17.13-300.fc36.x86_64
Intel(R) Core(TM) i7-4771 CPU @ 3.50GHz

Replaced with Testing VB 6.1.35 - Stable again and usable. Have run 2 W10 guests for 24 Hours no issue.

Only issue is 6.1.35 Guest Additions fail to install

I am now deffering kernel updates until 6.1.35 goes to full release

[fth0]

Only issue is 6.1.35 Guest Additions fail to install

Please try the separately downloadable GA ISO instead of the included one.