memtest86 test 3 fails on guest with multiple threads

[ramidzkh]

I had a standard Linux guest running, a freshly installed Ubuntu 21.10 Desktop, and I shut the machine down to increase the core count from the initial 1. Once rebooting, everything was very unstable. I could not get past the Ubuntu splash image without entering grub first and a lot of other things randomly kept failing, so I ran memtest86 and it was failing as soon as it reached test 3. I checked my host system and that had no memtest86 errors. I made a new machine with 1 core and went straight into memtest86, but it's also failing test 3. I have not checked if the guest fails anything other than test 3.

Host: Processor Intel(R) Core(TM) i7-8700 CPU @ 3.20GHz 3.19 GHz, RAM: 16GB, Windows 10 Home 21H1 19043.1586
Guest: Standard Linux/Ubuntu 64 bit guest, 6GB RAM, 6 cores, 100GB hard drive, Ubuntu 21.10 Desktop

[mpack]

I don't really see much point in looking for physical errors in virtual RAM or virtual disks. Look to the host PC for that.

[scottgus1]

I agree with Mpack, that you should only run memory tests on the physical host, not inside the VM. The VM is using the host's memory, so the host's memory range should be tested. FWIW, any run of the VM may shift to a different range of physical host memory cells, depending on what other programs are running at the start of the VM. So the RAM you memtest in the VM today extremely likely won't be the RAM your VM uses tomorrow. And we have no idea if or how the RAM handling routine in Virtualbox may or may not be compatible with memory testers.

Please try a fresh new VM set to 2 processors before you do the install. Is the new VM stable? If it is not, post a vbox.log:

Start the VM from full normal shutdown, not save-state. Run until you see the problem happen, then shut down the VM from within the VM's OS if possible. If not possible, close the Virtualbox window for the VM with the Power Off option set.

Right-click the VM in the main Virtualbox window's VM list, choose Show Log. Save the far left tab's log, zip it, and post the zip file, using the forum's Upload Attachment tab.

[ramidzkh]

I already ran memtest86 on my host PC, it got through all the tests multiple times with no failures. I've also tested QEMU and that seems to work, so I'm pretty sure it's VirtualBox.

I made a completely fresh VM, ubuntu 64 bit, 8192MB ram and 10gb fresh disk, set it to 6 cores and went straight into memtest86 from grub

Here's the log https://gist.github.com/ramidzkh/532bb1 ... 0cf24d3891
Here's a picture https://i.imgur.com/Y0AxJeO.png

[mpack]

Why are you persisting in looking for physical defects in simulated RAM? That quite obviously makes no sense.

There are a number of hardware tests and benchmarks that make no sense at all to run in virtual environments. The results will only serve to confuse.

[fth0]

FWIW, memtest86 inside a VirtualBox VM works fine, if VirtualBox uses its own hardware virtualization (VT-x/AMD-V).

In the case of VirtualBox 6.1.32 on a Windows host with Hyper-V active ...

[scottgus1]

Here's the log (some 3rd-party link)

Save the far left tab's log, zip it, and post the zip file, using the forum's Upload Attachment tab.

Following instructions will give you the best help. Not following instructions will get you delays...

[ramidzkh]

Why are you persisting in looking for physical defects in simulated RAM? That quite obviously makes no sense.

There are a number of hardware tests and benchmarks that make no sense at all to run in virtual environments. The results will only serve to confuse.

I ran memtest because the guest was being very unstable, such as not being able to boot properly, which is not present if I use the same iso on a physical machine

[mpack]

00:00:06.688190 Host RAM: 16206MB (15.8GB) total, 7335MB (7.1GB) available
00:00:07.549815 CPUM: Physical host cores: 6
...
00:00:06.816970 RamSize <integer> = 0x0000000200000000 (8 589 934 592, 8 192 MB, 8.0 GB)
00:00:06.817158 VRamSize <integer> = 0x0000000001000000 (16 777 216, 16 MB)
00:00:06.816968 NumCPUs <integer> = 0x0000000000000006 (6)

You are over allocating host CPU cores and RAM (and under-allocating guest graphics RAM).

Normally Linux will allocate RAM dynamically, rather like how dynamic VDI works. Ironically, by running a memory test you exercise all regions of RAM causing it all to be fully allocated and thus forcing the error.


Reduce allocated RAM to 4096MB, CPU cores to 2, and increase graphics RAM to 128MB.

I see you have also allocated just 10GB to the hard disk drive. That isn't going to be enough for a modern desktop OS. It might be best to scrap this VM and start again with the settings mentioned above, and a 64GB dynamic VDI drive.