QUESTION: accessing pcscd from muvm / fex

[Rad0van]

I don't know if this is question for muvm or fex. I am running Ubuntu Asahi and need to run software for x86-64 that accesses pcscd servive (reading smart cards). The program runs but complains pcscd service is not running / available. Not sure what mechanism is used on x86-64 for the programs to communicate with the daemon and whether it is somehow possible to bridge these in muvm / fex scenario. Any help is appreciated. Thank you.

[slp]

IIRC, pcscd listens for clients on a socket. You would need to expose that socket to the guest. While muvm already exposes some sockets to the guest, it doesn't support having the user to specify additional sockets to be exposed, so you would need to extend it to be able to do that. Should be relatively simple using the existing implementation for pulseaudio as a reference.

[Rad0van]

IIRC, pcscd listens for clients on a socket. You would need to expose that socket to the guest. While muvm already exposes some sockets to the guest, it doesn't support having the user to specify additional sockets to be exposed, so you would need to extend it to be able to do that. Should be relatively simple using the existing implementation for pulseaudio as a reference.

Hi Sergio,

thank you for answer - this is exactly what I have been looking for. I would probably be able to do some simple copy-paste with changing few constants but I am not familiar with either Rust neither muvm internals. Can you point me a little closer as to where to look for this? Thank you.

[slp]

This is what we do in the host:

muvm/crates/muvm/src/bin/muvm.rs

Line 294 in f00a4ee

if let Ok(run_path) = env::var("XDG_RUNTIME_DIR") {

And this is what we do in the guest:

muvm/crates/muvm/src/guest/bin/muvm-guest.rs

Line 128 in f00a4ee

setup_socket_proxy(pulse_path, PULSE_SOCKET)?;

[Rad0van]

Guest seems easy ;-) Host not so much but I'll give it a try... Thanks a lot.

[Rad0van]

@slp with the help of Claude I managed to build a version with pcscd support. However I am facing another challenge: the application should start a local http server on port 15480 but it doesn't seem to be available from the host machine. I have used -p=15480:15480/TCP parameter however I am getting something like this:

wget http://localhost:15480/icon
--2026-03-24 09:50:46--  http://localhost:15480/icon
Resolving localhost (localhost)... 127.0.0.1
Connecting to localhost (localhost)|127.0.0.1|:15480... connected.
HTTP request sent, awaiting response... Read error (Connection reset by peer) in headers.
Retrying.

So it clearly tries to connect but fails later on. Because when the application is not running it returns:
Connecting to localhost (localhost)|127.0.0.1|:15480... failed: Connection refused.

Any idea what might be going on? Thanks a lot.

[Rad0van]

I've had Claude analyze this and it's working now:
Summary of the bug and fix:

Root cause: eID_Client inside the VM listens only on 127.0.0.1:15480 (loopback). passt forwards TCP connections from the host to the VM's eth0 IP address — but nothing listens there, so the guest kernel sends TCP RST, causing "Connection reset by peer".

Fix: After the guest network is configured (DHCP gives eth0 an IP), we now start a socat TCP proxy inside the VM for each published port:

TCP-LISTEN:<host_port>,bind=<eth0_ip>,fork,reuseaddr → TCP:127.0.0.1:<guest_port>
This proxy binds on the eth0 IP (where passt delivers connections) and forwards them to loopback (where the application listens). The bind to the specific eth0 IP avoids conflicts with the application's 127.0.0.1 binding.

Files changed:

• utils/launch.rs — added publish_ports field to GuestConfiguration
• bin/muvm.rs — passes publish_ports into GuestConfiguration
• guest/net.rs — configure_network() now returns the assigned Ipv4Addr; added tcp_port_pairs() parser
• guest/socket.rs — added setup_loopback_tcp_proxies()
• guest/bin/muvm-guest.rs — calls the proxy setup after network config

If there's interest I can create a PR for this. @slp