Merge pull request #127 from DLR-FT/dev/syscalls2

Prepare syscall logic

Cargo.toml

@@ -49,3 +49,5 @@ polling = "3.4"
 itertools = "0.13"
 once_cell = "1.19"
 bytesize = "1.1"
+bincode = "1.3"
+serde = { version = "1.0", features = ["derive"] }

core/Cargo.toml

@@ -20,15 +20,14 @@ procfs.workspace = true
 polling.workspace = true
 itertools.workspace = true
 anyhow.workspace = true
+bincode.workspace = true
+serde.workspace = true
 
 log = "0"
 walkdir = "2.3"
-serde = { version = "1.0", features = ["derive"] }
 memfd = "0.6"
-bincode = "1.3"
 thiserror = "1.0"
 bytesize = {workspace = true, features = ["serde"]}
-byteorder = "1.4.3"
 enum_primitive = "0.1"
 ptr_meta = "0.2.0"
 

core/src/syscall/mod.rs

@@ -0,0 +1,171 @@
+//! Common definitions for the execution of system calls
+
+use anyhow::Result;
+
+pub const SYSCALL_SOCKET_PATH: &str = "/syscall-a653";
+
+pub mod receiver;
+pub mod sender;
+pub mod syscalls;
+mod ty;
+
+pub use ty::SyscallType;
+
+// This is the data type that is transferred to the hypervisor when a
+// syscall request is made by a partition. The parameter data is stored as an
+// already serialized `Vec<u8>`, so that the receiver can deserialize the
+// SyscallType without knowing the parameter's types.
+type SyscallRequest = (SyscallType, Vec<u8>);
+
+// This is the data type that is returned from the hypervisor to the partition
+// when a syscall was handled. In contrast to [`SyscallRequest`], a generic can
+// be used for the return value's type.
+type SyscallResponse<T> = Result<T, a653rs::bindings::ErrorReturnCode>;
+
+#[cfg(test)]
+mod tests {
+    use std::collections::VecDeque;
+    use std::os::unix::net::UnixDatagram;
+    use std::thread;
+    use std::time::Duration;
+
+    use a653rs::bindings::ApexSystemTime;
+    use a653rs::prelude::{QueueOverflow, QueuingPortId};
+
+    use super::SyscallType;
+    use crate::syscall::receiver::{self, SyscallReceiver};
+    use crate::syscall::sender::SyscallSender;
+    use crate::syscall::syscalls;
+
+    fn new_sender_receiver_pair() -> (SyscallSender, SyscallReceiver) {
+        let (sender, receiver) = UnixDatagram::pair().unwrap();
+        (
+            SyscallSender::from_datagram(sender),
+            SyscallReceiver::from_datagram(receiver),
+        )
+    }
+
+    #[test]
+    pub fn single_syscall() {
+        let (sender, receiver) = new_sender_receiver_pair();
+
+        let receiver_thread = thread::spawn(move || {
+            let syscall_handler = |ty: SyscallType, serialized_params: &[u8]| -> Vec<u8> {
+                assert_eq!(ty, SyscallType::SendQueuingMessage);
+
+                receiver::wrap_serialization::<syscalls::SendQueuingMessage, _>(
+                    serialized_params,
+                    |params| {
+                        assert_eq!(
+                            params,
+                            (
+                                0 as QueuingPortId,
+                                [1u8, 2, 3].as_slice(),
+                                0 as ApexSystemTime
+                            )
+                        );
+
+                        Ok(())
+                    },
+                )
+                .expect("serialization to succeed")
+            };
+
+            let syscall_was_handled = receiver
+                .receive_one(Some(Duration::from_secs(1)), syscall_handler)
+                .unwrap();
+
+            assert!(syscall_was_handled);
+        });
+
+        // Make a syscall
+        let response: Result<(), a653rs::bindings::ErrorReturnCode> = sender
+            .execute::<syscalls::SendQueuingMessage>((
+                0 as QueuingPortId,
+                &[1, 2, 3],
+                0 as ApexSystemTime,
+            ))
+            .expect("sending and receiving a response to succeed");
+
+        assert_eq!(response, Ok(()));
+
+        // join the receiver thread just to be safe
+        receiver_thread.join().unwrap();
+    }
+
+    #[test]
+    pub fn two_syscalls() {
+        let (sender, receiver) = new_sender_receiver_pair();
+
+        // The receiver thread represents the hypervisor.
+        let receiver_thread = thread::spawn(move || {
+            // A simulated queuing port. This represents the hypervisor state.
+            let mut queuing_port_state: VecDeque<Vec<u8>> = VecDeque::new();
+
+            let mut syscall_handler = |ty: SyscallType, serialized_params: &[u8]| -> Vec<u8> {
+                match ty {
+                    SyscallType::SendQueuingMessage => {
+                        receiver::wrap_serialization::<syscalls::SendQueuingMessage, _>(
+                            serialized_params,
+                            |params| {
+                                queuing_port_state.push_back(params.1.to_owned());
+
+                                Ok(())
+                            },
+                        )
+                        .expect("serialization to succeed")
+                    }
+                    SyscallType::ReceiveQueuingMessage => {
+                        receiver::wrap_serialization::<syscalls::ReceiveQueuingMessage, _>(
+                            serialized_params,
+                            |_params| {
+                                queuing_port_state
+                                    .pop_front()
+                                    .map(|msg| (false as QueueOverflow, msg))
+                                    .ok_or(a653rs::bindings::ErrorReturnCode::NotAvailable)
+                            },
+                        )
+                        .expect("serialization to succeed")
+                    }
+                    _ => unimplemented!("this test only implements two syscalls"),
+                }
+            };
+
+            // Let's handle exactly three syscalls
+            for _ in 0..3 {
+                let syscall_was_handled = receiver
+                    .receive_one(Some(Duration::from_secs(1)), &mut syscall_handler)
+                    .unwrap(); // TODO log error and ignore syscall
+                assert!(syscall_was_handled);
+            }
+        });
+
+        // Send one message into the queuing port
+        let response = sender
+            .execute::<syscalls::SendQueuingMessage>((
+                0 as QueuingPortId,
+                &[4, 3, 2, 1],
+                0 as ApexSystemTime,
+            ))
+            .unwrap();
+        assert_eq!(response, Ok(()));
+
+        // Receive the previous message from the queuing port
+        let response = sender
+            .execute::<syscalls::ReceiveQueuingMessage>((0 as QueuingPortId, 0 as ApexSystemTime))
+            .expect("sending and receiving a response to succeed");
+        assert_eq!(response, Ok((false as QueueOverflow, vec![4, 3, 2, 1])));
+
+        // Now the queuing port should be empty
+        let response = sender
+            .execute::<syscalls::ReceiveQueuingMessage>((0 as QueuingPortId, 0 as ApexSystemTime))
+            .expect("sending and receiving a response to succeed");
+        assert_eq!(
+            response,
+            Err(a653rs::bindings::ErrorReturnCode::NotAvailable)
+        );
+
+        // join the receiver thread just to be safe
+        receiver_thread.join().unwrap();
+    }
+}