> Jakub Klama wrote:
>> The purpose of virtio-console is to create arbitrary bidirectional, 
>> host-to-guest communication
>> channels that bypass guest's network stack (don't require working networking 
>> in the guest).
> thanks. i had no idea that the existing console support required a networking 
> stack; it works for the loader and i can watch the kernel load and 
> initialize, so i always assumed it bypassed "networking".

It didn't. What I meant is that virtio-console can be used as a replacement for 
TCP/IP communication between host and guest (at least in some applications). 
For example, it can be used by the "guest additions" code to talk to the host.

>> But even for using it as the system console, it's a bit better than emulated 
>> serial port, because
>> the protocol supports passing console resize events from host to guest.
> nmdm could theoretically (as pty and pts both do) support TIOCGWINSZ and 
> SIGWINCH, though? or perhaps bhyve's virtio_console device could offer pts(4) 
> support?

Yeah, virtio_console totally deserves support for binding virtual console to a 

> is the protocol spoken over the virtio_console socket documented online? i'd 
> like to add support to rtty. currently i've been running the "resize" command 
> at the guest shell, which feels very primitive.

Right now the socket just speaks raw data. It would need some multiplexing or 
ideally an IPC mechanism to send the resize events.

freebsd-virtualization@freebsd.org mailing list
To unsubscribe, send any mail to 

Reply via email to