[Bug]: VirtualBox 7.2: Crash when screens coming back from sleep

[alxandr3]

Version

7.2.0

Host OS Type

Windows

Host OS name + version

Windows 11 Build 26100.4652

Host Architecture

x86

Guest OS Type

Linux

Guest Architecture

x86

Guest OS name + version

Manjaro (rolling - kernel 6.12.41-1-MANJARO)

Component

Other

What happened?

Many times when my screens go to sleep and I wake them up, I get a popup saying

VirtualBoxVM.exe Application Error - instruction 0x00007FFC6DCE1FC7 referenced memory at 0x0E0 could not be written.

Attaching logs only for the initial report, but I activated local dumps so I will upload it + fresh logs the next time this happens.

aquila-2025-08-25-15-05-43.log

How can we reproduce this?

This doesn't happen every time the screens go to sleep or wake up (not sure which), but I suspect it's related to my monitor setup. I have one screen with adaptive sync and other advanced features that make it slower to wake up than my other screen, which is just a regular 4K 60Hz display. Both are connected via DisplayPort cables (e.g. the OS will detect if a screen is off and transfer its windows to the remaining screen, which doesn't happen when using HDMI cables, on my setup at least).

Since the secondary screen wakes up before the primary screen (where VirtualBox runs), this causes display glitches (in general, whether vbox is running or not) during the first few seconds of waking up before everything settles.

Both monitors are connected to an RTX 4080 with NVIDIA drivers 580.88. This issue has been happening for well over a year and also occurred with VirtualBox 7.1 and earlier versions.

Did you upload all of your necessary log files, screenshots, etc.?

• Yes, I've uploaded all pertinent files to this issue.

[irwir]

Somewhat similar configuration: single monitor, integrated Intel graphics, DisplayPort, sleep mode configured.
Crashes on display wake up had been observed for over 2 years in earlier versions of VirtualBox (Windows 11 22H2, 24H2).
There were cases when VirtualBox Manager also crashed.
Guests mostly crash if it was in a foreground window of the host, otherwise guests tended to keeps going but guest's window got maximized (not using full screen mode).
Edit.
VB does not crash every time display goes to sleep.
There are less chances for a crash if the guest's window size was fixed, but still possible (which leads to a guess that on display wake up VB receives weird display sizes which it does not know how to handle - or VB looses display device completely for a while).

[andiz81]

I have the same problem with a general error message: "VirtualBox manager does niot work anymore. ..."
Crash: VirtualBox.exe.
VirtualBox version: 7.2 + Guest additions
OS: Windows 11 Pro 24H2
Graphic: NIVIDA

It can be reproduced if VirtualBox Window is visible on screen and the screen gets turned off.
(I have a second monitor, where i can see the error pops up after the monitor is switched off)

This error happend since i replaced the HDMI cable with an DisplayPort (DVI) cable.
This can't be a coincidence.

[Dsen-orcl]

Hello guys, thank you for your reports!

Unfortunately, so far I was not able to reproduce that trouble tho I have more or less similar setup (Windows 11 host with NVIDIA card and DisplayPort cable connected to my single 4k monitor). I was trying to reproduce it with 7.1 and 7.2 both. What I was trying is: 1. to send host to sleep mode and back, 2. to send only screen to sleep mode and back, 3. even detaching my single monitor and attaching it back (because I suspect that might be a Qt issue with zero-monitor handling). But so far, no progress reproducing the problem. BTW, we had such reports previously as well, the problem is only to reproduce it with debug build.

Wondering which guests you have, whether guest additions installed and whether 3D is enabled or not. Or maybe any additional hints to make the crash more stable way?
Thanks in advance!

[irwir]

Host is never sleeping, only display.
Guest OS probably any: Windows, Ubuntu...
Guest graphics VMVGA without 3D, VMSVGA with or without 3D.
Guest additions installed.

that might be a Qt issue with zero-monitor handling

There are a few records in the Event Viewer. Exception codes are in pairs in this order: 0xc0000005, then 0xc000041d.

Faulting application name: VirtualBox.exe, version: 7.2.0.20228, time stamp: 0x689cee7e
Faulting module name: Qt6GuiVBox.dll, version: 6.8.0.0, time stamp: 0x673e228a
Exception code: 0xc000041d
Fault offset: 0x00000000000d59c0
Faulting application path: C:\Program Files\Oracle\VirtualBox\VirtualBox.exe
Faulting module path: C:\Program Files\Oracle\VirtualBox\Qt6GuiVBox.dll

[alxandr3]

I'm running mostly just Manjaro, but I remember once I had a manjaro VM booted as well as a Rocky Linux booted at the same time, and they both crashed simultaneously when my screens woke up. Guest additions are installed, 3D is enabled.

@Dsen-orcl: My feeling is that you'd need two screens to reproduce though, if my theory is correct. The behavior of my screens with DVI cables (not a theory) is that if I power off one screen, all my windows will move to the second screen. Because one of my screen is slower at powering on/waking up, every time my pc comes back from sleep all my windows will move to the second screen for just 1-2 seconds before coming back to their original screen once it's available.
I believe we are facing a race condition where virtualbox tries to render the VM on the second screen and bring it back to the first screen before it finished its initial rendering, or something along those lines. I'm working on a powershell script to force monitors going to sleep to help reproduce. What do you propose we do if you only have one screen though, is there something I could run on my machine to get you more information if I'm able to reproduce?

Also, to further pinpoint the issue: as per the error in the logs attached in the original description:

VMSVGA: command 1165: VERR_INVALID_STATE

Command 1165 on the host side (if my understanding is correct) pertains to vmsvga3dCmdDXDefineQuery() in src/VBox/Devices/Graphics/DevVGA-SVGA-cmd.cpp, which only defines the query if VMSVGA3D_DX is defined, further pointing to a bug with 3D acceleration in the VMSVGA driver.

@irwir - I would like to draw you attention on the fact that this ticket is for VirtualBoxVM.exe, not VirtualBox.exe - although both issues look alike, they are from two different executables. The file which I referenced just above is not used by VirtualBox.exe (the manager), but only the VM itself (VirtualBoxVM.exe). The issue you should be commenting on is #121 as it pertains to the executable you are referencing. I also got the crash of the manager and was able to provide logs and crash dumps in #121 , too. It was closed yesterday, but if nobody picks up on it within a few days, I'll try to find another way to have it reopened.

[irwir]

Single 4K display, 200% scaling.
VM Adapter is VMSVGA, no 3D, scale factor is 200%.

Turned off display with a button while Ubuntu guest was running minimized; manager probably was running too. When the display was turned on, there was a VM window at the upper left corner sized like credit card, and this message box:
Caption: ubuntu [Running] - Oracle VirtualBox: VirtualBoxVM.exe - Application Error
Text: The instruction at 0x00007FFDD1B75750 referenced memory at 0x0000000000000008. The memory could not be read.

This is from manager dump, but the same offset "8" - apparently, null pointer dereference (object pointer in RCX is null)
00007FFDD1B759C0 48 8B 41 08 mov rax,qword ptr [rcx+8]
00007FFDD1B759C4 0F 10 80 88 00 00 00 movups xmm0,xmmword ptr [rax+88h]
00007FFDD1B759CB 48 8B C2 mov rax,rdx
00007FFDD1B759CE 0F 11 02 movups xmmword ptr [rdx],xmm0
00007FFDD1B759D1 C3 ret

No events, no crash dumps for VM itself.

[alxandr3]

I am no longer able to reproduce the issue, and the only thing that changed is an upgrade of nvidia drivers (though downgrading to latest -1 still didn't cause the crashes, maybe older versions would be different).

@irwir do you want me to keep this ticket open? If it helps, troubleshooting steps I'd see would be:

• upgrading nvidia drivers if this applies to you (it never changed anything for me until last week...)
• double checking your registry in HKEY_LOCAL_MACHINE\Software\Microsoft\Windows\Windows Error Reporting\LocalDumps to make sure you don't have an entry for VirtualBoxVM.exe that would have it dump to a different location that the default %LOCALAPPDATA%\CrashDumps, perhaps explaining why you don't see any dumps for it when you get the application error
• uploading the full VM logs containing a crash in this ticket

[irwir]

@alxandr3
My suggestion would be to leave this ticket open for a while.

There are no VM crash dumps, only manager's dumps.
Crashes do not happen every time when display power state changes.

Intel driver was updated regurlarly, host OS changed from Windows 11 22H2 to 24H2 and now 25H2, a few VB versions changed too.
Collected data from a number of differnt VMs indicate that the issue is mostly unrelated to:

• integrated Intel video or nVidia chips
• host Hyper-V on or off
• guest OS
• guest graphics adapter and 3D support

Could be that the idea of the crash originating somewhere in Qt6 was correct.

The latest manager dump was with the fresh 7.2.2 (not pair of dumps as was usual for 7.2.0):

Sequence of records in the event viewer is:
System log
Information 11.09.2025 19:03:35 Kernel-Power 566 (268)
Information 11.09.2025 19:03:42 Kernel-Power 566 (268)
Application log
Error 11.09.2025 19:03:46 Application Error 1000 Application Crashing Events

The crash happened after display powered on, not when it turned off.

Faulting application name: VirtualBox.exe, version: 7.2.2.20484, time stamp: 0x68c1bd6d
Faulting module name: Qt6GuiVBox.dll, version: 6.8.0.0, time stamp: 0x673e228a
Exception code: 0xc0000005
Fault offset: 0x00000000000d69aa

Possibly in QScreen::virtualSiblings while accessing something at [rcx+8] - null pointer for QScreen object.

[irwir]

VMs keep failing from time to time.
Again, one dump for manager and no dumps for VM; the event log simply copies message box contents:

Application popup: ubuntu [Running] - Oracle VirtualBox: VirtualBoxVM.exe - Application Error : The instruction at 0x00007FF911135750 referenced memory at 0x0000000000000008. The memory could not be read.

[Dsen-orcl]

VMs keep failing from time to time. Again, one dump for manager and no dumps for VM; the event log simply copies message box contents:

Application popup: ubuntu [Running] - Oracle VirtualBox: VirtualBoxVM.exe - Application Error : The instruction at 0x00007FF911135750 referenced memory at 0x0000000000000008. The memory could not be read.

Could you please upload that one dump for 7.2.2 VirtualBox.exe process? Thx in advance!

[irwir]

Sorry, that one dump had been purged already.
Event log

Faulting application name: VirtualBox.exe, version: 7.2.2.20484, time stamp: 0x68c1bd6d
Faulting module name: Qt6GuiVBox.dll, version: 6.8.0.0, time stamp: 0x673e228a
Exception code: 0xc0000005
Fault offset: 0x00000000000d69aa

Crashed in the same QScreen::virtualSiblings

?virtualSiblings@QScreen@@QEBA?AV?$QList@PEAVQScr
0D69A0: 4C 8B DC  mov    r11,rsp
0D69A3: 56        push   rsi
0D69A4: 4155      push   r13
0D69A6: 4883EC68  sub    rsp,068
0D69AA: 488B4108  mov    rax,[rcx][8]

VirtualBox.exe.dmp.zip

Manager crashes appear to be easily reproducible.
Attached is another dump, but seems to be identical in crash type and address.

[Dsen-orcl]

Ok, I've parsed new one dump (for VirtualBox.exe, i.e. manager app), thank you! It seems to be a bug in Qt itself. We have possibility to patch Qt as well of course, but to check my guess faster I'd like to test a workaround on the GUI side. I'll update latest Windows test-build and upload it soon (hopely today). If my guess is correct and workaround works => I'll explain it properly.