bhyve dev

[bhyve dev @bhyve_dev]

@chmari_t What guest o/s is this ?

[bhyve dev @bhyve_dev]

@jmmv It was a somewhat arbitrary compromise between static allocation of vCPU structs, and common CPU counts when the change was done.

[bhyve dev @bhyve_dev]

@_bramandityo @FreeBSDHelp Tiger, VNC Viewer, tightvnc should all work fine.

[bhyve dev @bhyve_dev]

Debian bhyve guest running Blender, Nvidia RTX2080 passthru.
Thanks to Corvin Köhne and Mario Marietto for getting this working.

[bhyve dev @bhyve_dev]

@lattera So you're claiming that because a patch exists, it has to be committed ?

[bhyve dev @bhyve_dev]

@lattera Such as ?

[bhyve dev @bhyve_dev]

@lattera What are you doing to help ?

[bhyve dev @bhyve_dev]

@jmmv "total number of vm exits" is a good place to start. If that's e.g. in the 100ks/sec, the other exit counters can be used to narrow down the issue.

[bhyve dev @bhyve_dev]

@allanjude @jmmv bhyve doesn't expose the kvm pvclock.

[bhyve dev @bhyve_dev]

The win11 release ISO requires the install-time regedit TPM workaround.

[Olivier @PCTuning_OW]

Just tested #Windows11 in #FreeBSD #BHYVE. Seems to run, without stupid requirements:
•TPM 2.0
• Secure Boot
• RAM ≥ 4GB

[Warner Losh @bsdimp]

So I'm running Venix/86 16bit a.out binaries using vm86 mode on my FreeBSD/i386 VM running under bhyve on my FreeBSD/amd64 build box. Trippy.

[bhyve dev @bhyve_dev]

@maackintosh @lattera Not yet.

[bhyve dev @bhyve_dev]

Windows11 #bhyve

[bhyve dev @bhyve_dev]

@ctuffli NVMe

[bhyve dev @bhyve_dev]

@Stefar77 The overhead of the vm-exit itself is going to be difficult to overcome.

[bhyve dev @bhyve_dev]

@Stefar77 An easy win is to remove the verify_gla() call.

[bhyve dev @bhyve_dev]

@Stefar77 There was some GSoC work done on this a while back: https://t.co/XoPCIliebH

[bhyve dev @bhyve_dev]

@Stefar77 If these are all instr-emul exits, it could be worth some dtrace to determine the guest phys addr.