One of the most useful things that vcenter can do is the live virtual machine migrations between ESXi hosts and storage. You can move a vm while it's running from one host to another, even if the storage is not shared. There might be a tiny blip during the cut off, but the vm is still up and serving traffic throughout the entire process which could at times take hours. Hopefully Bhyve could have some of that same functionality, maybe even jails too?

Does bhyve allow USB pass-through?

tuaris: I'm sure it's in the plans, but it could take a while. But, it would certainly be a nice feature!

farhan: Yes, it does.

farhan: Well, I haven't tried it, but it does allow PCI pass-through. So, I don't see why it wouldn't.

redis.io/blog/agplv3 those linux folks indoctrinating people into thinking that code which isnt GPL can be "stolen"

Yeah. That's been a problem for a very long time. Although, to be fair, if you don't want your code used anywhere that isn't OSS, maybe don't distribute as such?

Licensing gets tricky, but I've always just allowed anyone to use anything I've written at their own free will. If I didn't want to allow that, I wouldn't make it open source.

farhan: no it does not. You can PCI-Passthrough only

badkat: You can't pass-through the device the USB is connected to?

ek: Yes the usb bus is enumerated as a PCI device, but is not the same concept

you give the vm the full PCI bus, not only the usb sub device, as you can do on virtualbox/vmware (linux/windows)

in freebsd for some silly reason we cant handle usb devices inside virtual machines, not even on QEMU

Well, dang. That seems... strange. Like, even giving the full PCI bus to the VM, it doesn't see it as it's own and the devices attached?

at the funny part is that ESXi hypervisors have based their USB passtrough kernel module on the freebsd BSD usb stack lol

I'm struggling to understand how that even makes sense. Like, I'd mentioned, I've never tried it, but that sure is a shame.

ek: yes, giving the full pci bus works, but is not a usb-passthrough by term

Wait... It seems to have been working 2.5 years ago? forums.freebsd.org/threads/usb-pass…hrough-bhyve-windows-10-guest.86669

Is this a regression?

people tell usb-passtrough but its not. its pci

usb-passtrough is more flexible, they just make it to put the usb pci bus in the VM, not the same

also you have to reboot the HOST if you want to get back the pci bus from bhyve, that is crap in a lot of scenarios. In usb-passtrough you can attach/dettach on the VM but it can return to the host when not in use from the VM

I suppose that's fair. But, you *ARE* able to access USB devices from a VM on a Bhyve host.

but its not called USB-Passthrough, that was my point.

Yeah. The reboot and dedication of the device/port is kind of a bummer.

Yep, yep. I gotcha'.

i would love to invest time on that

If I were only that useful!

i saw some folk in illumos kernel that also uses bhyve trying to get something alike

I can see that.

Bhyve is a bit behind on some functionality, but boy-oh-boy does it perform better for me than ESXi, VirtualBox or Proxmox.

I've been extremely happy with it for what I need it for.

yes its really great, and plus ZFS even more :)

Exactly. I still snapshot my Bhyve dataset, but there's really isn't a reason to. Especially with the vm-bhyve auto-snapshot script which you can include for each startup (or shutdown? I can't remember.)

But, utilizing ZFS certainly gives me extra piece of mind. I've had some problems before with VMware storage. I was always able to recover (for the most part) but it wasn't easy.

I do with non-EFI grub stuff was easier with Bhyve, though. I sometimes have to test a lot of Linux-based images and it takes a while to figure it out (if I even can.) Grub is kinda rough sometimes.

s/with/wish/

i really like esxi for some use cases

also i had problems making sysadmins adopt bhyve , they are so biased with GUI based administration...

smart-os built a GUI for it IIRC

works with illumos/bhyve/zfs underneath

Yeah. They are. It's understandable considering everything has been for so long. There was a Bhyve webUI in development for a while (can't remember the name) and it seemed like it was going to be reasonable. I still haven't checked it out.

Nice. I hadn't heard of that. Whatever I'm thinking of was different.

FreeNAS had one too

production-ready

supports jails too

Hrm. Never heard that either. I'm wondering how far TrueNAS will take FBSD moving forward.

past year i worked on making esxi boots over bhyve, had problems after the kernel started

Esxi is bsd based?

The one I was thinking of was something designed by the cbsd team (I think?)

I thought it was Linux?

farhan: no, it WAS linux

farhan: No. It isn't FBSD based.

today last release is a own custom made kernel

that have linux compatible kernel modules for network cards

and the USB passthrough is based on freebsd usb stack

RIP Hans...

Hans? :O

Which Hans? Hans Reiser? lol

Reiser lma0

that guy was crazy as fuck

He wrote the USB stack on FreeBSD. Really great guy, spoke to him once.

It wasn't "MurderFS" for no reason!

Hans Reiser wrote usb stack? roflma0

ek: hahah

There's no way we're talking about the same Hans. lol

farhan: which Hans you mean lol

Ah... forums.freebsd.org/threads/in-memor…etter-william-sirevag-selasky.89697

Not even gonna try to type that name. Copy/paste was much easier.

Yes, him. Hans Selasky. Really great guy...

Sure sounds like it. That's a shame.

Oh, wait a minute! I remember that dude! I worked with him on some of the Infiniband drivers many, MANY years ago.

Trying to get BSD storage as an alternative in mainstream HPC setups. ConnectX was a BASTARD!

Man, that really is a bummer. He was an incredibly nice guy. Great to work with as he was crazy patient and understanding.

Kind of the opposite of Torvalds.

anyways, bhyve doesn't do USB passthrough just yet. That's really unfortunate.

maybe someday I can help sovle that.

i would love too, just need to find the right hardware to make comfy tests

Currently working on a USB driver. The lack of pass-through is a bummer.

That it is.

farhan: i use esxi for that :P

so i can boot multiple OSes and rotate the passthrough with an script

this new non-linux esxi is pretty tight and works solid enough.

It's own proprietary kernel?

Uhubctl locks up. It turns off power to a USB port.

farhan: yes

If I want to install a package with different options then the defaults, I have to use ports for that, yes?

mns: Or drive it with something like Poudriere, but yes.

thanks mason

maybe i forgot: is there any reason PIDs are capped at 99'999?

... oh, because pid 100'000 represents "no process group", sayeth the headers

MelMalik: i don't think that's the reason since NO_PID is just PID_MAX+1. i believe the answer is a lot of old code assumes 16-bit pid_t and tries to sprintf pids into a 6-byte buffer

ah

is there a way to list the bhyve instances running?

crb: sysutils/vm-bhyve is a great wrapper around bhyve with this and a lot more functionality. Sadly, bhyve itself doesn't expose any API/ABI

crb: That said, you can inspect bhyve processes using ps(1)

ls /dev/vmm

it tells you the created instances, it cannot know if something is running

the bhyve process will exit with a code corresponding to the vm exit state, but if the vm doesn't exit it might still be there

hi

please help, I have a problem with a k55d asus laptop, I loaded acpi_wmi and acpi_asus_wmi, but cpu fan never starts. Thanks!

?

any hints?

Posterdati: is the computer hot?

"fan in my laptop isn't running" isn't usually a problem, but what people want

I wish the fan would stop in my framework

very

cpu temp around 90C

what is telling you the temperature?

sysctl -a | grep temperature

[tj]: reading...

no asusctl on freebsd

sounds like linux does something it doesn't have to, but maybe we need to do something to get the fans into bios control

Posterdati: can you upload a dmesg to demsgd.nycbug.org

[tj]: submitted

that is quite an old computer, does the fan work?

yes

only thing that stuck out to me in that dmesg is this line ACPI: \134_SB.PCI0.WMI1.WQXM: 1 arguments were passed to a non-method ACPI object (Buffer) (20221020/nsarguments-361)

but I have no idea what it means

Posterdati: did it work in freebsd before?

I had debian 12 on it

Posterdati: are there fan settings in the bios?(would be unusual for a laptop IME) Could be some way to set it at a constant speed or something, it would be loud but at least it wouldn't self destruct while troubleshooting

no settings

in bios

the age probably means that asusctl isn't aimed at it

unless it has an anime led screen on the lid

:(

Posterdati: this may take some hacking, I'm not sure if that is what you want as an answer

I would figure out how linux is controlling the fan, which kernel module or whatever is responsible

crb: yes, check with ls -l /dev/vmm/*

is there a version of the 2025 FBF survey in PDF?

[tj]: fixed

[tj]: I had to remove the CMOS battery

and put it back, I assume? or replace it?

I just installed freebsd on a machine that has a mediatek wifi device. Noticed there is a mediatek mt76 module in src / contrib. But it doesnt seem to have a module compiled. Looking for guidance for where to find documentation for how to compile contrib drivers.

drobban: which mediatek card do you have? exact model

badkat: no details, it just tells me MediaTek when running usbconfig

attached to ugen.

first thing to do when dealing with network card drivers its to get an certain idea of the hardware identificators ok?

check dmesg and usbconfig/pciconf on verbose mode

grab vendor ID,product ID, all the details you can

mostly stunned that im not able to compile the modules found in contrib. but, yea.

you dont even know the model and want to compile stuff

good luck :)

badkat: you are right, I would assume the kernel know enough to attach it to ugen0 and call it MediaTek instead of showing ids as with unknown devices.

but I might remember incorrectly. and if its able to attach the driver is another story, I guessed that if could compile the modules it would be interesting to see if they would attach or not. But they seem unable to compile to begin with, further investigation into what id's i have is also waste of time.

as I have no code that could drive it.

Posterdati: wild, well done

drobban: are you sure the module isn't in your system, some kernel modules get compiled directly into the kernel and aren't there as loadable modules (no idea about that specific one)

any usb module should be loaded automatically

drobban: what's the directory in src/contrib where you spotted meditek?

I guess sys not src, so sys/contrib/dev/mediatek/mt76/

what does ugen say? anything about the model of device?

badkat: that's exactly what I was looking for, thank you!

rtprio: yea thats the one.

or yea, nimaje thats the one.