bhyve Booted Debian and FreeBSD 15 using qemu accelerated with bhyve/vmm for the first time. This is an epic milestone for the community !

memory bandwidth seems odd. My host is FreeBSD 15.0 Release and I get 18.3 GB/s. In my bhyve guests FreeBSD 15.0 Release reaches 17.1 GB/s and Ubuntu 24.04 12 GB/s.
 
I will repeat the tests when I have enabled SMP on FreeBSD. At the moment it seems that this task is complicated to achieve. When I say to qemu/bhyve to use more than 1 CPU,it crashes every time.
 
ZioMario said:
I'm also nostalgic. I've been enjoying qemu/kvm since the 1990s. I cut my teeth on it. Virtualization has always been my favorite thing to do, first on Linux. Then on FreeBSD. I also owe the Nvidia GPU passthru feature to me, working behind the scenes with Corvin.
Click to expand...
Qemu was released in 2003 and kvm in 2006 according to Wikipedia.
 
freezr said:
I miss the goal.
You made it because you can get HW acceleration only through QEMU? Can't you?
Click to expand...

QEMU has many features that Bhyve does not have.

For example with QEMU you can change ISO on the fly:
- https://www.linux-kvm.org/page/Change_cdrom

You may even connect USB devices over IP network from other computers with QEMU:
- https://zachary.com/freebsd-usbip/
- https://freshports.org/net/usbredir/

You can also passthru single USB device with QEMU - but only entire USB controller (and must support MSI/MSI-X interrupts to work) on Bhyve.

So I welcome having Bhyve acceleration for QEMU as a VERY WANTED feature :)

Hope that helps.

Regards,
vermaden
 
---> You can also passthru single USB device with QEMU

thanks to this feature,I'm working on a parallel project that will fix once for all the problems that FreeBSD has with most (if not all) BT chipsets. I had an idea and I'm developing it. Because Qemu allows me to do it thanks to libusb.
 
With QEMU on top, bhyve will gain :
  • advanced BIOS/UEFI
  • modern virtual device
  • sophisticated PCI topology
  • evoluted virtio
  • a serious USB emulation
  • advanced NVMe
  • advanced snapshot
  • migration
  • advanced block layer
  • complete qcow2
  • a mature SPICE/VNC
  • vfio-like orchestration
  • complete libvirt
Compatibility with :
  • appliance cloud
  • OpenStack images
  • Proxmox images
  • libvirt images
  • tooling CI/CD
  • Kubernetes virtualization
  • Terraform
  • Packer
Qemu already has :
  • virtio-gpu
  • virglrenderer
  • rutabaga
  • gfxstream
  • vhost-user-gpu

How many of these features are also offered by bhyve ? it is useful to know this,for further development. The idea might be to not invest time and money on bhyve for those features that are offered by qemu,but only for its stabilization (of bhyve).
 
Love it 💪.... Keep at it ZioMario all of your post regarding Bhyve passthru through the past 2-5 years got me excited to game again...

Much appreciation, I'll keep lurking around through your efforts help out when I can.
 
loveydovey said:
There may be security reasons bhyve is not implementing some things that qemu might.
Click to expand...

More than for security reasons, I think development is slow by design and that there are few developers capable of working there 'cause FreeBSD cannot boast a large number of developers working on it, generally speaking, as it does not have the same social penetration capacity that Linux has. Anyway,security measures can be implemented, right ?
 
rootbert said:
bhyve is slick and does it's job for 95% of standard usecases.
Click to expand...

Standard yes. 95% should be decreased. Above I have listed some functions not offered by the slick bhyve that makes the 95% decrease until 70% or even less. The average users are only a part of the equation.
 
ZioMario , your signature says "System admins could be defined,somehow, as OS psychologists. Infact in the future a specific OS will be integrated in the human brains,so there will be no difference between fixing an OS or a human brain anymore. As system admins / psychologists we are just ahead the times that will come."

Is that your personal hell pipe dream? Because that's not happening with smart people who know how fallible OSes are and have always been. So unless you want your brain hacked and violated in some nefarious ways, you should not be a proponent of any such fantasy.

Moreover, 95% human and 5% artificial OS is not really human anymore. It's someone's strong desire to make that a sliding scale and single-handedly redefine what it means to be human, but it is not human to not be human to some extent. There are some fundamental reasons why that's so...it's reductive to give up some percent of humanity, to say the least. Even 0.05% non-human has unfathomable effects of profound complexity.

And then you'll have these reductive non-humans trying to run things and make decisions about what's good for humans, having nothing to do with being human anymore. I bet some people will want to do it secretly so that there's no debate about it...and do it slowly because incremental change is harder to identify and easier to shove down people's throat. But the end result will be immediately "the machines have taken over."

The machines may have already taken over, if you analyze what you see around you. How would one know for sure?
 
You must log in or register to reply here.
Back
Top