[BRLTTY] redirect braille display to virtual machine

[Samuel Thibault]

Hello,

Aura Kelloniemi, le sam. 21 mars 2026 09:41:32 +0200, a ecrit:
> On 2026-03-20 at 18:32 +0100, Samuel Thibault <samuel.thibault at ens-lyon.org> wrote:
>  > Paul Scott, le ven. 20 mars 2026 12:09:07 -0500, a ecrit:
>  > > Mabey forward the brlapi address from Linux to windows? Abd connect
>  > > brapi to the shared address.
> 
>  > That is technically possible, yes, but the essential problem that
>  > remains is to switch automatically as needed. The libvirt-native case is
>  > the same question, so better fix it that way.
> 
> I have never tried, but I assume that controlling BRLTTY (or something else)
> with an emulated Baum display would be very difficult and limited (especially
> because Baum displays are the ones which only a few keys). Therefore I would
> really appreciate if there would be a way to pass BrlAPI through the VM
> boundary – that would allow using my existing bindings on the guest side and
> allow raw access to my host display.

We have been thinking about doing this for some time actually. Not
specifically for qemu, but also for the spice protocol, containerized
applications, etc., that'd just need to forward the brlapi packets,
adding the tty path along the way. That could also support
detach/reattach, the guest wouldn't know about ttys etc., it's the
virtualization front-end that would add the tty.

That being said, you can already do that by just forwarding the brlapi
4101 TCP port to the guest and use the brlapi driver in it.  You'd just
need to set WINDOWPATH=n inside the guest when starting brltty, to make
its output show up on VT n.

> I'd actually want to run DOS guests which would require either a working
> network stack in DOS or BrlAPI over serial port (or something similar).

With a bit of work (giving libbrlapi the already-opened file rather than
letting it connect to TCP) that could work over serial, yes.

(and qemu can probably forward the serial port to a TCP connection)

Samuel