Components residing in the same domain (`qrexec-client-vm` to `qrexec-agent`, `qrexec-client` to `qrexec-daemon`) use local sockets as the underlying transport medium.
Components in separate domains (`qrexec-daemon` to `qrexec-agent`, data channel between `qrexec-agent`s) use vchan links.
Because of [vchan limitation](https://github.com/qubesos/qubes-issues/issues/951), it is not possible to establish qrexec connection back to the source domain.
*`-d target-domain-name`: Specifies the target for the execution/service request.
*`-l local-program`: Optional. If present, `local-program` is executed and its stdout/stdin are used when sending/receiving data to/from the remote peer.
*`-e`: Optional. If present, stdout/stdin are not connected to the remote peer. Only process creation status code is received.
*`-c <request-id,src-domain-name,src-domain-id>`: used for connecting a VM-VM service request by `qrexec-policy`. Details described below in the service example.
*`cmdline`: Command line to pass to `qrexec-daemon` as the execution/service request. Service request format is described below in the service example.
(If `local_program` is set, `qrexec-client` executes it and uses that child's stdin/stdout in place of its own when exchanging data with `qrexec-agent` later.)
`qrexec-client` translates that request into a `MSG_EXEC_CMDLINE` message sent to `qrexec-daemon`, with `connect_domain` set to 0 (connect to **dom0**) and `connect_port also set to 0 (allocate a port).
- **dom0**: `qrexec-daemon` allocates a free port (in this case 513), and sends a `MSG_EXEC_CMDLINE` back to the client with connection parameters (**domX** and 513) and with command field empty.
`qrexec-client` disconnects from the daemon, starts a vchan server on port 513 and awaits connection.
Then, `qrexec-daemon` passes on the request as `MSG_EXEC_CMDLINE` message to the `qrexec-agent` running in **domX**. In this case, the connection parameters are **dom0** and 513.
- **domX**: `qrexec-agent` receives `MSG_EXEC_CMDLINE`, and starts the command (`user:cmd`, or `cmd` as user `user`). If possible, this is actually delegated to a separate server (`qrexec-fork-server`) also running on domX.
After starting the command, `qrexec-fork-server` connects to `qrexec-client` in **dom0** over the provided vchan port 513.
- Data is forwarded between the `qrexec-client` in **dom0** and the command executed in **domX** using `MSG_DATA_STDIN`, `MSG_DATA_STDOUT` and `MSG_DATA_STDERR`.
Empty messages (with data `len` field set to 0 in `msg_header`) are an EOF marker. Peer receiving such message should close the associated input/output pipe.
When `cmd` terminates, **domX**'s `qrexec-fork-server` sends `MSG_DATA_EXIT_CODE` header to `qrexec-client` followed by the exit code (**int**).
### domX: request execution of service `admin.Service` in dom0
(If `local_program` is set, it will be executed in **domX** and connected to the remote command's stdin/stdout).
`qrexec-client-vm` connects to `qrexec-agent` and requests service execution (`admin.Service`) in **dom0**.
`qrexec-agent` assigns an internal identifier to the request. It's based on a file descriptor of the connected `qrexec-client-vm`: in this case, `SOCKET11`.
- **dom0**: `qrexec-daemon` receives the request and triggers `qrexec-policy` program, passing all necessary parameters: source domain **domX**, target domain **dom0**, service `admin.Service` and identifier `SOCKET11`.
`qrexec-policy` evaluates if the RPC should be allowed or denied, possibly also launching a GUI confirmation prompt.
(If the RPC is denied, it returns with exit code 1, in which case `qrexec-daemon` sends a `MSG_SERVICE_REFUSED` back).
- **dom0**: If the RPC is allowed, `qrexec-policy` will launch a `qrexec-client` with the right command:
The `-c domX,X,SOCKET11` are parameters indicating how connect back to **domX** and pass its input/output.
The command parameter describes the RPC call: it contains service name (`admin.Service`), source domain (`domX`) and target description (`name dom0`, could also be e.g. `keyword @dispvm`). The target description is important in case the original target wasn't dom0, but the service is executing in dom0.
`qrexec-client` connects to a `qrexec-daemon` for **domX** and sends a `MSG_SERVICE_CONNECT` with connection parameters (**dom0**, and port 0, indicating a port should be allocated) and request identifier (`SOCKET11`).
`qrexec-daemon` allocates a free port (513) and sends back connection parameters to `qrexec-client` (**domX** port 513).
`qrexec-client` starts the command, and tries to connect to **domX** over the provided port 513.
Then, `qrexec-daemon` forwards the connection request (`MSG_SERVICE_CONNECT`) to `qrexec-agent` running in **domX**, with the right parameters (**dom0** port 513, request `SOCKET11`).
- **dom0**: Because the command has the form `QUBESRPC: ...`, it is started through the `qubes-rpc-multiplexer` program with the provided parameters (`admin.Service domX name dom0`). That program finds and executes the necessary script in `/etc/qubes-rpc/`.
- **domX**: `qrexec-agent` receives the `MSG_SERVICE_CONNECT` and passes the connection parameters back to the connected `qrexec-client-vm`. It identifies the `qrexec-client-vm` by the request identifier (`SOCKET11` means file descriptor 11).
`qrexec-client-vm` starts a vchan server on 513 and receives a connection from `qrexec-client`.
- Data is forwarded between **dom0** and **domX** as in the previous example (dom0-VM).
### domX: invoke execution of qubes service `qubes.Service` in domY
- The request is forwarded as `MSG_TRIGGER_SERVICE3` to `qrexec-daemon` running in **dom0**, then to `qrexec-policy`, then (if allowed) to `qrexec-client`.
`qrexec-client` will then send a `MSG_EXEC_CMDLINE` message to `qrexec-daemon` for **domY**. The message will be with port number 0, requesting port allocation.
`qrexec-daemon` for **domY** will allocate a port (513) and send it back. It will also send a `MSG_EXEC_CMDLINE` to its corresponding agent. (It will also translate `DEFAULT` to the configured default username).
Then, `qrexec-client` will also send `MSG_SERVICE_CONNECT` message to **domX**'s agent, indicating that it should connect to **domY** over port 513.
Having notified both domains about a connection, `qrexec-client` now exits.
- **domX**: `qrexec-agent` receives a `MSG_SERVICE_CONNECT` with connection parameters (**domY** port 513) and request identifier (`SOCKET11`). It sends the connection parameters back to the right `qrexec-client-vm`.
`qrexec-client-vm` starts a vchan server on port 513. note that this is different than in the other examples: `MSG_SERVICE_CONNECT` means you should start a server, `MSG_EXEC_CMDLINE` means you should start a client.
- **domY**: `qrexec-agent` receives a `MSG_EXEC_CMDLINE` with the command to execute (`user:QUBESRPC...`) and connection parameters (**domX** port 513).
It forwards the request to `qrexec-fork-server`, which handles the command and connects to **domX** over the provided port.
Because the command is of the form `QUBESRPC ...`, `qrexec-fork-server` starts it using `qubes-rpc-multiplexer` program, which finds and executes the necessary script in `/etc/qubes-rpc/`.
`qrexec-policy` is a mechanism for evaluating whether an RPC call should be allowed. For introduction, see [Qubes RPC administration](/doc/qrexec/#qubes-rpc-administration).
### `qrexec-policy-daemon`
This is a service running in dom0. It is called by `qrexec-daemon` and is responsible for evaluating the request and possibly launching an action.
The daemon listens on a socket (`/var/run/qubes/policy.sock`). It accepts requests in the format described in [qrexec-policy-daemon.rst](https://github.com/QubesOS/qubes-core-qrexec/blob/master/doc/qrexec-policy-daemon.rst) and replies with `result=allow/deny`.
