[PDM 1.1.7] PDM in VM inside PVE 9: Odd Total RAM reporting inside VM?

[SInisterPisces]

Hello,

I've configured my PDM VM with 3 GiB of RAM. (Yes, this is an odd config, but this is an extremely RAM-limited PVE node that I'm using for testing.)
In short, I'm showing about 2.77-2.8 GiB of total RAM inside the VM, which isn't what I've assigned.

I recall reading threads like this before, but I can't remember this ever happening on one of my VMs on other nodes. Have I done something wrong?

I suspect that this is actually normal behavior. Something about the difference between total host memory usage and usage inside the VM? In any case, I'd like to understand this better.



Meanwhile, inside the VM, I'm getting reports of ... a good bit less total RAM. Instead of 3.0 GiB, I'm seeing 2.8 GiB or 2.77 GiB, depending on where I look.

Btop:


Htop:

 

[Impact]

It's pretty normal to have some memory unavailable for reserves and such. See journalctl -b0 -g "reserved" for example. There's many more things you can look into but this might help already.
Depending on how it's configured, the behavior also differs. For Q35 it appears to reserve a bit more, for example. Compare free -m now and after changing it to i440fx. It's similar for BIOS but that might be harder to test/change.
Any specific reason you use a VM for this and allocate it "so much" memory just for PDM? My PDM CT uses ~120M/^150M. Granted, I use it very sparingly so it might need more in your case.