ptp: Listen with different sockets on individual interfaces

This allows us to portably know on which interface a multicast packet
arrived, and to send back any packets for that clock on the correct
interface.

Fixes https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/2728

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/5232>

subprojects/gstreamer/libs/gst/helpers/ptp/ffi.rs

@@ -41,6 +41,7 @@ pub mod unix {
     pub const POLLNVAL: c_short = 0x20;
 
     pub const IPPROTO_IP: c_int = 0;
+
     #[cfg(any(
         target_os = "freebsd",
         target_os = "openbsd",
@@ -54,6 +55,19 @@ pub mod unix {
     #[cfg(any(target_os = "solaris", target_os = "illumos"))]
     pub const IP_ADD_MEMBERSHIP: c_int = 19;
 
+    #[cfg(any(
+        target_os = "freebsd",
+        target_os = "openbsd",
+        target_os = "netbsd",
+        target_os = "dragonfly",
+        target_os = "macos",
+    ))]
+    pub const IP_MULTICAST_IF: c_int = 9;
+    #[cfg(target_os = "linux")]
+    pub const IP_MULTICAST_IF: c_int = 32;
+    #[cfg(any(target_os = "solaris", target_os = "illumos"))]
+    pub const IP_MULTICAST_IF: c_int = 16;
+
     #[cfg(any(
         target_os = "solaris",
         target_os = "illumos",
@@ -822,7 +836,9 @@ pub mod windows {
     }
 
     pub const IPPROTO_IP: u32 = 0u32;
+
     pub const IP_ADD_MEMBERSHIP: u32 = 12u32;
+    pub const IP_MULTICAST_IF: u32 = 9u32;
 
     pub const SOL_SOCKET: u32 = 65535;
     pub const SO_REUSEADDR: u32 = 4;

subprojects/gstreamer/libs/gst/helpers/ptp/io.rs

@@ -8,25 +8,59 @@
 //
 // SPDX-License-Identifier: MPL-2.0
 
+use std::net::UdpSocket;
+
+#[derive(Debug, PartialEq, Eq, Clone, Copy)]
+pub enum SocketType {
+    EventSocket,
+    GeneralSocket,
+}
+
 /// Result of polling the inputs of the `Poll`.
 ///
-/// Any input that has data available for reading will be set to `true`, potentially multiple
-/// at once.
-///
 /// Note that reading from the sockets is non-blocking but reading from stdin is blocking so
 /// special care has to be taken to only read as much as is available.
-pub struct PollResult {
+pub struct PollResult<'a> {
-    pub event_socket: bool,
+    ready_sockets: &'a [(usize, SocketType, &'a UdpSocket)],
-    pub general_socket: bool,
+    sockets: &'a [(UdpSocket, UdpSocket)],
-    pub stdin: bool,
+    stdin: Option<&'a Stdin>,
+    stdout: &'a Stdout,
+}
+
+impl<'a> PollResult<'a> {
+    /// Returns the sockets that are currently ready for reading.
+    pub fn ready_sockets(&self) -> &[(usize, SocketType, &UdpSocket)] {
+        &self.ready_sockets
+    }
+
+    /// Returns the event socket.
+    pub fn event_socket(&self, idx: usize) -> &UdpSocket {
+        &self.sockets[idx].0
+    }
+
+    /// Returns the general socket.
+    pub fn general_socket(&self, idx: usize) -> &UdpSocket {
+        &self.sockets[idx].1
+    }
+
+    /// Returns standard input if there is data to read.
+    pub fn stdin(&self) -> Option<&Stdin> {
+        self.stdin
+    }
+
+    /// Returns standard output.
+    pub fn stdout(&self) -> &Stdout {
+        self.stdout
+    }
 }
 
 #[cfg(unix)]
 mod imp {
-    use super::PollResult;
+    use super::{PollResult, SocketType};
 
     use std::{
         io::{self, Read, Write},
+        mem,
         net::UdpSocket,
         os::unix::io::{AsRawFd, RawFd},
     };
@@ -37,10 +71,11 @@ mod imp {
     ///
     /// This carries the event/general UDP socket and stdin/stdout.
     pub struct Poll {
-        event_socket: UdpSocket,
+        sockets: Vec<(UdpSocket, UdpSocket)>,
-        general_socket: UdpSocket,
         stdin: Stdin,
         stdout: Stdout,
+        pollfd: Vec<pollfd>,
+        results_cache: Vec<(usize, SocketType, &'static UdpSocket)>,
     }
 
     #[cfg(test)]
@@ -83,7 +118,7 @@ mod imp {
     }
 
     #[cfg(test)]
-    impl Read for Pipe {
+    impl<'a> Read for &'a Pipe {
         fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> {
             // SAFETY: read() requires a valid fd and a mutable buffer with the given size.
             // The fd is valid by construction as is the buffer.
@@ -122,25 +157,27 @@ mod imp {
 
     impl Poll {
         /// Name of the input based on the `struct pollfd` index.
-        fn fd_name(idx: usize) -> &'static str {
+        fn fd_name(idx: usize, len: usize) -> &'static str {
             match idx {
-                0 => "event socket",
+                i if i == len - 1 => "stdin",
-                1 => "general socket",
+                i if i % 2 == 0 => "event socket",
-                2 => "stdin",
+                i if i % 2 == 1 => "general socket",
                 _ => unreachable!(),
             }
         }
 
-        /// Create a new `Poll` instance from the two sockets.
+        /// Create a new `Poll` instance from the pairs of sockets.
-        pub fn new(event_socket: UdpSocket, general_socket: UdpSocket) -> Result<Self, Error> {
+        pub fn new(sockets: Vec<(UdpSocket, UdpSocket)>) -> Result<Self, Error> {
             let stdin = Stdin::acquire();
             let stdout = Stdout::acquire();
+            let n_sockets = sockets.len();
 
             Ok(Self {
-                event_socket,
+                sockets,
-                general_socket,
                 stdin,
                 stdout,
+                pollfd: Vec::with_capacity(n_sockets * 2 + 1),
+                results_cache: Vec::with_capacity(n_sockets * 2),
             })
         }
 
@@ -157,57 +194,63 @@ mod imp {
 
             Ok((
                 Self {
-                    event_socket,
+                    sockets: vec![(event_socket, general_socket)],
-                    general_socket,
                     stdin: Stdin(stdin.read),
                     stdout: Stdout(stdout.write),
+                    pollfd: Vec::with_capacity(3),
+                    results_cache: Vec::with_capacity(2),
                 },
                 stdin,
                 stdout,
             ))
         }
 
-        /// Mutable reference to the event socket.
+        /// Reference to the event socket.
-        pub fn event_socket(&mut self) -> &mut UdpSocket {
+        #[allow(unused)]
-            &mut self.event_socket
+        pub fn event_socket(&self, iface: usize) -> &UdpSocket {
+            &self.sockets[iface].0
         }
 
-        /// Mutable reference to the general socket.
+        /// Reference to the general socket.
-        pub fn general_socket(&mut self) -> &mut UdpSocket {
+        #[allow(unused)]
-            &mut self.general_socket
+        pub fn general_socket(&self, iface: usize) -> &UdpSocket {
+            &self.sockets[iface].1
         }
 
-        /// Mutable reference to stdin for reading.
+        /// Reference to stdin for reading.
-        pub fn stdin(&mut self) -> &mut Stdin {
+        #[allow(unused)]
-            &mut self.stdin
+        pub fn stdin(&self) -> &Stdin {
+            &self.stdin
         }
 
-        /// Mutable reference to stdout for writing.
+        /// Reference to stdout for writing.
-        pub fn stdout(&mut self) -> &mut Stdout {
+        pub fn stdout(&self) -> &Stdout {
-            &mut self.stdout
+            &self.stdout
         }
 
         /// Poll the event socket, general socket and stdin for available data to read.
         ///
         /// This blocks until at least one input has data available.
-        pub fn poll(&mut self) -> Result<PollResult, Error> {
+        pub fn poll<'a>(&'a mut self) -> Result<PollResult<'a>, Error> {
-            let mut pollfd = [
+            self.pollfd.clear();
-                pollfd {
+            for (event_socket, general_socket) in &self.sockets {
-                    fd: self.event_socket.as_raw_fd(),
+                self.pollfd.push(pollfd {
+                    fd: event_socket.as_raw_fd(),
                     events: POLLIN,
                     revents: 0,
-                },
+                });
-                pollfd {
+                self.pollfd.push(pollfd {
-                    fd: self.general_socket.as_raw_fd(),
+                    fd: general_socket.as_raw_fd(),
                     events: POLLIN,
                     revents: 0,
-                },
+                });
-                pollfd {
+            }
+
+            self.pollfd.push(pollfd {
                 fd: self.stdin.0,
                 events: POLLIN,
                 revents: 0,
-                },
+            });
-            ];
 
             // SAFETY: Polls the given pollfds above and requires a valid number to be passed.
             // A negative timeout means that it will wait until at least one of the pollfds is
@@ -219,7 +262,7 @@ mod imp {
             // On EINTR polling should be retried.
             unsafe {
                 loop {
-                    let res = poll(pollfd[..].as_mut_ptr(), pollfd.len() as _, -1);
+                    let res = poll(self.pollfd[..].as_mut_ptr(), self.pollfd.len() as _, -1);
                     if res == -1 {
                         let err = std::io::Error::last_os_error();
                         if err.kind() == std::io::ErrorKind::Interrupted {
@@ -233,20 +276,60 @@ mod imp {
             }
 
             // Check for errors or hangup first
-            for (idx, pfd) in pollfd.iter().enumerate() {
+            for (idx, pfd) in self.pollfd.iter().enumerate() {
                 if pfd.revents & (POLLERR | POLLNVAL) != 0 {
-                    bail!("Poll error on {}", Self::fd_name(idx));
+                    bail!(
+                        "Poll error on {} for interface {}",
+                        Self::fd_name(idx, self.pollfd.len()),
+                        idx / 2
+                    );
                 }
 
                 if pfd.revents & POLLHUP != 0 {
-                    bail!("Hang up during polling on {}", Self::fd_name(idx));
+                    bail!(
+                        "Hang up during polling on {} for interface {}",
+                        Self::fd_name(idx, self.pollfd.len()),
+                        idx / 2
+                    );
+                }
+            }
+
+            self.results_cache.clear();
+            // SAFETY: References have the same memory representation independent of lifetime
+            let ready_sockets = unsafe {
+                mem::transmute::<
+                    &mut Vec<(usize, SocketType, &'static UdpSocket)>,
+                    &mut Vec<(usize, SocketType, &'a UdpSocket)>,
+                >(&mut self.results_cache)
+            };
+
+            for (idx, pfd) in self.pollfd.iter().enumerate() {
+                if pfd.revents & POLLIN != 0 {
+                    if idx == self.pollfd.len() - 1 {
+                        break;
+                    }
+                    if idx % 2 == 0 {
+                        ready_sockets.push((
+                            idx / 2,
+                            SocketType::EventSocket,
+                            &self.sockets[idx / 2].0,
+                        ));
+                    } else {
+                        ready_sockets.push((
+                            idx / 2,
+                            SocketType::GeneralSocket,
+                            &self.sockets[idx / 2].1,
+                        ));
+                    }
                 }
             }
 
             Ok(PollResult {
-                event_socket: pollfd[0].revents & POLLIN != 0,
+                ready_sockets: &*ready_sockets,
-                general_socket: pollfd[1].revents & POLLIN != 0,
+                sockets: &self.sockets,
-                stdin: pollfd[2].revents & POLLIN != 0,
+                stdin: (self.pollfd[self.pollfd.len() - 1].revents & POLLIN != 0)
+                    .then_some(&self.stdin),
+                stdout: &self.stdout,
             })
         }
     }
@@ -263,6 +346,12 @@ mod imp {
     }
 
     impl Read for Stdin {
+        fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> {
+            <&Stdin as Read>::read(&mut &*self, buf)
+        }
+    }
+
+    impl<'a> Read for &'a Stdin {
         fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> {
             // SAFETY: read() requires a valid fd and a mutable buffer with the given size.
             // The fd is valid by construction as is the buffer.
@@ -290,6 +379,16 @@ mod imp {
     }
 
     impl Write for Stdout {
+        fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
+            <&Stdout as Write>::write(&mut &*self, buf)
+        }
+
+        fn flush(&mut self) -> io::Result<()> {
+            <&Stdout as Write>::flush(&mut &*self)
+        }
+    }
+
+    impl<'a> Write for &Stdout {
         fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
             // SAFETY: write() requires a valid fd and a mutable buffer with the given size.
             // The fd is valid by construction as is the buffer.
@@ -326,6 +425,16 @@ mod imp {
     }
 
     impl Write for Stderr {
+        fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
+            <&Stderr as Write>::write(&mut &*self, buf)
+        }
+
+        fn flush(&mut self) -> io::Result<()> {
+            <&Stderr as Write>::flush(&mut &*self)
+        }
+    }
+
+    impl<'a> Write for &'a Stderr {
         fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
             // SAFETY: write() requires a valid fd and a mutable buffer with the given size.
             // The fd is valid by construction as is the buffer.
@@ -348,7 +457,7 @@ mod imp {
 
 #[cfg(windows)]
 mod imp {
-    use super::PollResult;
+    use super::{PollResult, SocketType};
 
     use std::{
         cmp,
@@ -371,12 +480,12 @@ mod imp {
     ///
     /// This carries the event/general UDP socket and stdin/stdout.
     pub struct Poll {
-        event_socket: UdpSocket,
+        sockets: Vec<(UdpSocket, UdpSocket)>,
-        event_socket_event: EventHandle,
+        events: Vec<(EventHandle, EventHandle)>,
-        general_socket: UdpSocket,
-        general_socket_event: EventHandle,
         stdin: Stdin,
         stdout: Stdout,
+        handles: Vec<HANDLE>,
+        results_cache: Vec<(usize, SocketType, &'static UdpSocket)>,
     }
 
     /// Helper struct for a WSA event.
@@ -456,7 +565,7 @@ mod imp {
     }
 
     #[cfg(test)]
-    impl Read for Pipe {
+    impl<'a> Read for &'a Pipe {
         fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> {
             // SAFETY: Reads the given number of bytes into the buffer from the stdin handle.
             unsafe {
@@ -510,21 +619,31 @@ mod imp {
     impl Poll {
         /// Internal constructor.
         pub fn new_internal(
-            event_socket: UdpSocket,
+            sockets: Vec<(UdpSocket, UdpSocket)>,
-            general_socket: UdpSocket,
             stdin: Stdin,
             stdout: Stdout,
         ) -> Result<Self, Error> {
             // Create event objects for the readability of the sockets.
-            let event_socket_event = EventHandle::new().context("Failed creating WSA event")?;
+            let events = sockets
-            let general_socket_event = EventHandle::new().context("Failed creating WSA event")?;
+                .iter()
+                .map(|(_event_socket, _general_socket)| -> Result<_, Error> {
+                    Ok((
+                        EventHandle::new().context("Failed creating WSA event")?,
+                        EventHandle::new().context("Failed creating WSA event")?,
+                    ))
+                })
+                .collect::<Result<Vec<_>, Error>>()?;
 
+            for ((event_socket, general_socket), (event_socket_event, general_socket_event)) in
+                Iterator::zip(sockets.iter(), events.iter())
+            {
                 // SAFETY: WSAEventSelect() requires a valid socket and WSA event, which are both
                 // passed here, and the bitflag of events that should be selected for.
                 //
                 // On error a non-zero value is returned.
                 unsafe {
-                if WSAEventSelect(event_socket.as_raw_socket(), event_socket_event.0, FD_READ) != 0
+                    if WSAEventSelect(event_socket.as_raw_socket(), event_socket_event.0, FD_READ)
+                        != 0
                     {
                         bail!(
                             source: io::Error::from_raw_os_error(WSAGetLastError()),
@@ -544,23 +663,26 @@ mod imp {
                         );
                     }
                 }
+            }
+
+            let n_sockets = sockets.len();
 
             Ok(Self {
-                event_socket,
+                sockets,
-                event_socket_event,
+                events,
-                general_socket,
-                general_socket_event,
                 stdin,
                 stdout,
+                handles: Vec::with_capacity(n_sockets * 2 + 1),
+                results_cache: Vec::with_capacity(1),
             })
         }
 
         /// Create a new `Poll` instance from the two sockets.
-        pub fn new(event_socket: UdpSocket, general_socket: UdpSocket) -> Result<Self, Error> {
+        pub fn new(sockets: Vec<(UdpSocket, UdpSocket)>) -> Result<Self, Error> {
             let stdin = Stdin::acquire().context("Failure acquiring stdin handle")?;
             let stdout = Stdout::acquire().context("Failed acquiring stdout handle")?;
 
-            Self::new_internal(event_socket, general_socket, stdin, stdout)
+            Self::new_internal(sockets, stdin, stdout)
         }
 
         #[cfg(test)]
@@ -579,45 +701,53 @@ mod imp {
             let stdout =
                 Stdout::from_handle(stdout_pipe.write).context("Failed acquiring stdout handle")?;
 
-            let poll = Self::new_internal(event_socket, general_socket, stdin, stdout)?;
+            let poll = Self::new_internal(vec![(event_socket, general_socket)], stdin, stdout)?;
 
             Ok((poll, stdin_pipe, stdout_pipe))
         }
 
-        /// Mutable reference to the event socket.
+        /// Reference to the event socket.
-        pub fn event_socket(&mut self) -> &mut UdpSocket {
+        #[allow(unused)]
-            &mut self.event_socket
+        pub fn event_socket(&self, iface: usize) -> &UdpSocket {
+            &self.sockets[iface].0
         }
 
-        /// Mutable reference to the general socket.
+        /// Reference to the general socket.
-        pub fn general_socket(&mut self) -> &mut UdpSocket {
+        #[allow(unused)]
-            &mut self.general_socket
+        pub fn general_socket(&self, iface: usize) -> &UdpSocket {
+            &self.sockets[iface].1
         }
 
-        /// Mutable reference to stdin for reading.
+        /// Reference to stdin for reading.
-        pub fn stdin(&mut self) -> &mut Stdin {
+        #[allow(unused)]
-            &mut self.stdin
+        pub fn stdin(&self) -> &Stdin {
+            &self.stdin
         }
 
-        /// Mutable reference to stdout for writing.
+        /// Reference to stdout for writing.
-        pub fn stdout(&mut self) -> &mut Stdout {
+        pub fn stdout(&self) -> &Stdout {
-            &mut self.stdout
+            &self.stdout
         }
 
         /// Poll the event socket, general socket and stdin for available data to read.
         ///
         /// This blocks until at least one input has data available.
-        pub fn poll(&mut self) -> Result<PollResult, Error> {
+        pub fn poll<'a>(&'a mut self) -> Result<PollResult<'a>, Error> {
-            let handles = [
+            self.handles.clear();
-                self.event_socket_event.0,
+
-                self.general_socket_event.0,
+            for (event_socket_event, general_socket_event) in &self.events {
+                self.handles.push(event_socket_event.0);
+                self.handles.push(general_socket_event.0);
+            }
+
+            self.handles.push(
                 // If stdin is a pipe then we use the signalling event, otherwise stdin itself.
                 if let Some(ref thread_state) = self.stdin.thread_state {
                     thread_state.event
                 } else {
                     self.stdin.handle
                 },
-            ];
+            );
 
             // If stdin is a pipe and currently no data is pending on it then signal
             // the reading thread to try reading one byte and blocking for that long.
@@ -641,8 +771,12 @@ mod imp {
             // On error u32::MAX is returned, otherwise an index into the array of handles is
             // returned for the handle that became ready.
             let res = unsafe {
-                let res =
+                let res = WaitForMultipleObjects(
-                    WaitForMultipleObjects(handles.len() as _, handles[..].as_ptr(), 0, u32::MAX);
+                    self.handles.len() as _,
+                    self.handles[..].as_ptr(),
+                    0,
+                    u32::MAX,
+                );
                 if res == u32::MAX {
                     bail!(
                         source: io::Error::from_raw_os_error(WSAGetLastError()),
@@ -651,21 +785,30 @@ mod imp {
                 }
 
                 assert!(
-                    (0..=2).contains(&res),
+                    (0..self.handles.len()).contains(&(res as usize)),
                     "Unexpected WaitForMultipleObjects() return value {}",
                     res,
                 );
 
-                res
+                res as usize
+            };
+
+            self.results_cache.clear();
+            // SAFETY: References have the same memory representation independent of lifetime
+            let ready_sockets = unsafe {
+                mem::transmute::<
+                    &mut Vec<(usize, SocketType, &'static UdpSocket)>,
+                    &mut Vec<(usize, SocketType, &'a UdpSocket)>,
+                >(&mut self.results_cache)
             };
 
             // For the sockets, enumerate the events that woke up the waiting, collect any errors
             // and reset the event objects.
-            if (0..=1).contains(&res) {
+            if res < self.handles.len() - 1 {
-                let (socket, event) = if res == 0 {
+                let (socket, event) = if res % 2 == 0 {
-                    (&self.event_socket, &self.event_socket_event)
+                    (&self.sockets[res / 2].0, &self.events[res / 2].0)
                 } else {
-                    (&self.general_socket, &self.general_socket_event)
+                    (&self.sockets[res / 2].1, &self.events[res / 2].1)
                 };
 
                 // SAFETY: Requires a valid socket and event, which is given by construction here.
@@ -681,8 +824,9 @@ mod imp {
                     {
                         bail!(
                             source: io::Error::from_raw_os_error(WSAGetLastError()),
-                            "Failed enumerating network events on {} socket",
+                            "Failed enumerating network events on {} socket for interface {}",
-                            if res == 0 { "event" } else { "general" },
+                            if res % 2 == 0 { "event" } else { "general" },
+                            res / 2,
                         );
                     }
 
@@ -692,8 +836,9 @@ mod imp {
                 if networkevents.ierrorcode[FD_READ_BIT] != 0 {
                     bail!(
                         source: io::Error::from_raw_os_error(networkevents.ierrorcode[FD_READ_BIT]),
-                        "Error on {} socket while waiting for events",
+                        "Error on {} socket for interface {} while waiting for events",
                         if res == 0 { "event" } else { "general" },
+                        res / 2,
                     );
                 }
 
@@ -707,12 +852,22 @@ mod imp {
                         res
                     );
                 }
+                ready_sockets.push((
+                    res / 2,
+                    if res % 2 == 0 {
+                        SocketType::EventSocket
+                    } else {
+                        SocketType::GeneralSocket
+                    },
+                    socket,
+                ));
             }
 
             Ok(PollResult {
-                event_socket: res == 0,
+                ready_sockets: &*ready_sockets,
-                general_socket: res == 1,
+                sockets: &self.sockets,
-                stdin: res == 2,
+                stdin: (res == self.handles.len() - 1).then_some(&self.stdin),
+                stdout: &self.stdout,
             })
         }
     }
@@ -918,6 +1073,12 @@ mod imp {
     }
 
     impl Read for Stdin {
+        fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> {
+            <&Stdin as Read>::read(&mut &*self, buf)
+        }
+    }
+
+    impl<'a> Read for &'a Stdin {
         fn read(&mut self, mut buf: &mut [u8]) -> io::Result<usize> {
             if buf.is_empty() {
                 return Ok(0);
@@ -926,7 +1087,7 @@ mod imp {
             // If a read byte is pending from the readiness signalling thread then
             // read that first here before reading any remaining data.
             let mut already_read = 0;
-            if let Some(ref mut thread_state) = self.thread_state {
+            if let Some(ref thread_state) = self.thread_state {
                 let mut guard = thread_state.buffer.lock().unwrap();
                 assert!(!guard.fill_buffer);
                 if guard.buffer_filled {
@@ -1013,6 +1174,16 @@ mod imp {
     }
 
     impl Write for Stdout {
+        fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
+            <&Stdout as Write>::write(&mut &*self, buf)
+        }
+
+        fn flush(&mut self) -> io::Result<()> {
+            <&Stdout as Write>::flush(&mut &*self)
+        }
+    }
+
+    impl<'a> Write for &'a Stdout {
         fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
             // SAFETY: Writes the given number of bytes to stdout or at most u32::MAX. On error
             // zero is returned, otherwise the number of bytes written is set accordingly and
@@ -1107,6 +1278,16 @@ mod imp {
     }
 
     impl Write for Stderr {
+        fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
+            <&Stderr as Write>::write(&mut &*self, buf)
+        }
+
+        fn flush(&mut self) -> io::Result<()> {
+            <&Stderr as Write>::flush(&mut &*self)
+        }
+    }
+
+    impl<'a> Write for &'a Stderr {
         fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
             if self.0 == INVALID_HANDLE_VALUE {
                 return Ok(buf.len());
@@ -1145,6 +1326,8 @@ pub use self::imp::{Poll, Stderr, Stdin, Stdout};
 mod test {
     #[test]
     fn test_poll() {
+        use super::{Poll, SocketType};
+
         use std::io::prelude::*;
 
         let event_socket = std::net::UdpSocket::bind(std::net::SocketAddr::from((
@@ -1167,8 +1350,7 @@ mod test {
         )))
         .unwrap();
 
-        let (mut poll, mut stdin, _stdout) =
+        let (mut poll, mut stdin, _stdout) = Poll::new_test(event_socket, general_socket).unwrap();
-            super::Poll::new_test(event_socket, general_socket).unwrap();
 
         let mut buf = [0u8; 4];
 
@@ -1177,30 +1359,33 @@ mod test {
                 .send_to(&[1, 2, 3, 4], (std::net::Ipv4Addr::LOCALHOST, event_port))
                 .unwrap();
             let res = poll.poll().unwrap();
-            assert!(res.event_socket);
+            assert_eq!(res.ready_sockets().len(), 1);
-            assert!(!res.general_socket);
+            assert_eq!(res.ready_sockets()[0].0, 0);
-            assert!(!res.stdin);
+            assert_eq!(res.ready_sockets()[0].1, SocketType::EventSocket);
-            assert_eq!(poll.event_socket().recv(&mut buf).unwrap(), 4);
+            assert_eq!(res.ready_sockets()[0].2.recv(&mut buf).unwrap(), 4);
             assert_eq!(buf, [1, 2, 3, 4]);
 
             send_socket
                 .send_to(&[1, 2, 3, 4], (std::net::Ipv4Addr::LOCALHOST, general_port))
                 .unwrap();
             let res = poll.poll().unwrap();
-            assert!(!res.event_socket);
+
-            assert!(res.general_socket);
+            assert_eq!(res.ready_sockets().len(), 1);
-            assert!(!res.stdin);
+            assert_eq!(res.ready_sockets()[0].0, 0);
-            assert_eq!(poll.general_socket().recv(&mut buf).unwrap(), 4);
+            assert_eq!(res.ready_sockets()[0].1, SocketType::GeneralSocket);
+            assert_eq!(res.ready_sockets()[0].2.recv(&mut buf).unwrap(), 4);
             assert_eq!(buf, [1, 2, 3, 4]);
 
             stdin.write_all(&[1, 2, 3, 4]).unwrap();
             let res = poll.poll().unwrap();
-            assert!(!res.event_socket);
+            assert!(res.ready_sockets().is_empty());
-            assert!(!res.general_socket);
+            {
-            assert!(res.stdin);
+                let mut stdin = res.stdin();
-            poll.stdin().read_exact(&mut buf).unwrap();
+                let stdin = stdin.as_mut().unwrap();
+                stdin.read_exact(&mut buf).unwrap();
                 assert_eq!(buf, [1, 2, 3, 4]);
             }
+        }
 
         drop(poll);
     }

subprojects/gstreamer/libs/gst/helpers/ptp/main.rs

@@ -73,12 +73,8 @@ fn create_socket(port: u16) -> Result<UdpSocket, Error> {
     Ok(socket)
 }
 
-/// Join the multicast groups for PTP on the configured interfaces.
+/// Retrieve the list of interfaces based on the available ones and the arguments.
-fn join_multicast(
+fn list_interfaces(args: &args::Args) -> Result<Vec<net::InterfaceInfo>, Error> {
-    args: &args::Args,
-    event_socket: &UdpSocket,
-    general_socket: &UdpSocket,
-) -> Result<[u8; 8], Error> {
     let mut ifaces = net::query_interfaces().context("Failed to query network interfaces")?;
     if ifaces.is_empty() {
         bail!("No suitable network interfaces for PTP found");
@@ -115,15 +111,28 @@ fn join_multicast(
         }
     }
 
+    Ok(ifaces)
+}
+
+fn run() -> Result<(), Error> {
+    let args = args::parse_args().context("Failed parsing commandline parameters")?;
+
+    let ifaces = list_interfaces(&args).context("Failed listing interfaces")?;
+
+    let mut sockets = vec![];
     for iface in &ifaces {
         info!("Binding to interface {}", iface.name);
-    }
 
-    for socket in [&event_socket, &general_socket].iter() {
+        let event_socket = create_socket(PTP_EVENT_PORT).context("Failed creating event socket")?;
-        for iface in &ifaces {
+        let general_socket =
+            create_socket(PTP_GENERAL_PORT).context("Failed creating general socket")?;
+
+        for socket in [&event_socket, &general_socket] {
             net::join_multicast_v4(socket, &PTP_MULTICAST_ADDR, iface)
                 .context("Failed to join multicast group")?;
         }
+
+        sockets.push((event_socket, general_socket));
     }
 
     let clock_id = if args.clock_id == 0 {
@@ -140,27 +149,13 @@ fn join_multicast(
     } else {
         args.clock_id.to_be_bytes()
     };
-
     info!("Using clock ID {:?}", clock_id);
 
-    Ok(clock_id)
-}
-
-fn run() -> Result<(), Error> {
-    let args = args::parse_args().context("Failed parsing commandline parameters")?;
-
-    let event_socket = create_socket(PTP_EVENT_PORT).context("Failed creating event socket")?;
-    let general_socket =
-        create_socket(PTP_GENERAL_PORT).context("Failed creating general socket")?;
-
     thread::set_priority().context("Failed to set thread priority")?;
 
     privileges::drop().context("Failed dropping privileges")?;
 
-    let clock_id = join_multicast(&args, &event_socket, &general_socket)
+    let mut poll = io::Poll::new(sockets).context("Failed creating poller")?;
-        .context("Failed joining multicast groups")?;
-
-    let mut poll = io::Poll::new(event_socket, general_socket).context("Failed creating poller")?;
 
     // Write clock ID first
     {
@@ -184,27 +179,20 @@ fn run() -> Result<(), Error> {
     // We assume that stdout never blocks and stdin receives a complete valid packet whenever it is
     // ready and never blocks in the middle of a packet.
     let mut socket_buffer = [0u8; 8192];
-    let mut stdinout_buffer = [0u8; 8192 + 3 + 8];
+    let mut stdinout_buffer = [0u8; 8192 + 4 + 8];
 
     loop {
         let poll_res = poll.poll().context("Failed polling")?;
 
         // If any of the sockets are ready, continue reading packets from them until no more
         // packets are left and directly forward them to stdout.
-        'next_socket: for idx in [poll_res.event_socket, poll_res.general_socket]
+        'next_socket: for (idx, type_, socket) in poll_res.ready_sockets() {
-            .iter()
+            let idx = *idx;
-            .enumerate()
+            let type_ = *type_;
-            .filter_map(|(idx, r)| if *r { Some(idx) } else { None })
-        {
-            let idx = idx as u8;
 
             // Read all available packets from the socket before going to the next socket.
             'next_packet: loop {
-                let res = match idx {
+                let res = socket.recv_from(&mut socket_buffer);
-                    MSG_TYPE_EVENT => poll.event_socket().recv_from(&mut socket_buffer),
-                    MSG_TYPE_GENERAL => poll.general_socket().recv_from(&mut socket_buffer),
-                    _ => unreachable!(),
-                };
 
                 let (read, addr) = match res {
                     Err(err) if err.kind() == std::io::ErrorKind::WouldBlock => {
@@ -213,12 +201,7 @@ fn run() -> Result<(), Error> {
                     Err(err) => {
                         bail!(
                             source: err,
-                            "Failed reading from {} socket",
+                            "Failed reading from {:?} socket for interface {}", type_, idx,
-                            if idx == MSG_TYPE_EVENT {
-                                "event"
-                            } else {
-                                "general"
-                            }
                         );
                     }
                     Ok((read, addr)) => (read, addr),
@@ -227,13 +210,10 @@ fn run() -> Result<(), Error> {
                 let recv_time = clock::time();
                 if args.verbose {
                     trace!(
-                        "Received {} bytes from {} socket from {} at {}",
+                        "Received {} bytes from {:?} socket for interface {} from {} at {}",
                         read,
-                        if idx == MSG_TYPE_EVENT {
+                        type_,
-                            "event"
+                        idx,
-                        } else {
-                            "general"
-                        },
                         addr,
                         recv_time,
                     );
@@ -275,18 +255,25 @@ fn run() -> Result<(), Error> {
                 }
 
                 {
-                    let mut buf = &mut stdinout_buffer[..(read + 3 + 8)];
+                    let mut buf = &mut stdinout_buffer[..(read + 4 + 8)];
-                    buf.write_u16be(read as u16 + 8)
+                    buf.write_u16be(read as u16 + 1 + 8)
                         .expect("Too small stdout buffer");
-                    buf.write_u8(idx).expect("Too small stdout buffer");
+                    buf.write_u8(if type_ == io::SocketType::EventSocket {
+                        MSG_TYPE_EVENT
+                    } else {
+                        MSG_TYPE_GENERAL
+                    })
+                    .expect("Too small stdout buffer");
+                    buf.write_u8(idx as u8).expect("Too small stdout buffer");
                     buf.write_u64be(recv_time).expect("Too small stdout buffer");
                     buf.write_all(&socket_buffer[..read])
                         .expect("Too small stdout buffer");
                     assert!(buf.is_empty(), "Too big stdout buffer",);
                 }
 
-                let buf = &stdinout_buffer[..(read + 3 + 8)];
+                let buf = &stdinout_buffer[..(read + 4 + 8)];
-                poll.stdout()
+                poll_res
+                    .stdout()
                     .write_all(buf)
                     .context("Failed writing to stdout")?;
             }
@@ -294,8 +281,8 @@ fn run() -> Result<(), Error> {
 
         // After handling the sockets check if a packet is available on stdin, read it and forward
         // it to the corresponding socket.
-        if poll_res.stdin {
+        if let Some(ref mut stdin) = poll_res.stdin() {
-            poll.stdin()
+            stdin
                 .read_exact(&mut stdinout_buffer[0..3])
                 .context("Failed reading packet header from stdin")?;
 
@@ -305,7 +292,7 @@ fn run() -> Result<(), Error> {
             }
             let type_ = stdinout_buffer[2];
 
-            poll.stdin()
+            stdin
                 .read_exact(&mut stdinout_buffer[0..size as usize])
                 .context("Failed reading packet body from stdin")?;
 
@@ -314,23 +301,31 @@ fn run() -> Result<(), Error> {
                 continue;
             }
 
+            if size < 1 + 8 + 34 {
+                bail!("Invalid packet body size");
+            }
+
+            let buf = &mut &stdinout_buffer[..(size as usize)];
+
+            let idx = buf.read_u8().expect("Too small stdin buffer");
+            if idx as usize >= ifaces.len() {
+                warn!("Unexpected stdin message interface index {}", idx);
+                continue;
+            }
+
             if args.verbose {
                 trace!(
-                    "Received {} bytes for {} socket from stdin",
+                    "Received {} bytes for {} socket for interface {} from stdin",
                     size,
                     if type_ == MSG_TYPE_EVENT {
                         "event"
                     } else {
                         "general"
                     },
+                    idx,
                 );
             }
 
-            if size < 8 + 34 {
-                bail!("Invalid packet body size");
-            }
-
-            let buf = &mut &stdinout_buffer[..(size as usize)];
             let main_send_time = buf.read_u64be().expect("Too small stdin buffer");
 
             // We require that the main process only ever sends valid PTP messages with the clock
@@ -347,11 +342,11 @@ fn run() -> Result<(), Error> {
 
             let send_time = clock::time();
             match type_ {
-                MSG_TYPE_EVENT => poll
+                MSG_TYPE_EVENT => poll_res
-                    .event_socket()
+                    .event_socket(idx as usize)
                     .send_to(buf, (PTP_MULTICAST_ADDR, PTP_EVENT_PORT)),
-                MSG_TYPE_GENERAL => poll
+                MSG_TYPE_GENERAL => poll_res
-                    .general_socket()
+                    .general_socket(idx as usize)
                     .send_to(buf, (PTP_MULTICAST_ADDR, PTP_GENERAL_PORT)),
                 _ => unreachable!(),
             }
@@ -393,7 +388,8 @@ fn run() -> Result<(), Error> {
             }
 
             let buf = &stdinout_buffer[..(3 + 12)];
-            poll.stdout()
+            poll_res
+                .stdout()
                 .write_all(buf)
                 .context("Failed writing to stdout")?;
         }

subprojects/gstreamer/libs/gst/helpers/ptp/net.rs

@@ -399,6 +399,93 @@ mod imp {
                 }
             }
 
+            #[cfg(not(any(
+                target_os = "openbsd",
+                target_os = "dragonfly",
+                target_os = "netbsd",
+                target_os = "macos"
+            )))]
+            {
+                let mreqn = ip_mreqn {
+                    imr_multiaddr: in_addr {
+                        s_addr: u32::from_ne_bytes(Ipv4Addr::UNSPECIFIED.octets()),
+                    },
+                    imr_address: in_addr {
+                        s_addr: u32::from_ne_bytes(Ipv4Addr::UNSPECIFIED.octets()),
+                    },
+                    imr_ifindex: iface.index as _,
+                };
+
+                // SAFETY: Requires a valid ip_mreq or ip_mreqn struct to be passed together
+                // with its size for checking which of the two it is. On errors a negative
+                // integer is returned.
+                unsafe {
+                    if setsockopt(
+                        socket.as_raw_fd(),
+                        IPPROTO_IP,
+                        IP_MULTICAST_IF,
+                        &mreqn as *const _ as *const _,
+                        mem::size_of_val(&mreqn) as _,
+                    ) < 0
+                    {
+                        bail!(
+                            source: io::Error::last_os_error(),
+                            "Failed joining multicast group for interface {}",
+                            iface.name,
+                        );
+                    }
+                }
+            }
+            #[cfg(any(target_os = "openbsd", target_os = "dragonfly", target_os = "macos"))]
+            {
+                let addr = in_addr {
+                    s_addr: u32::from_ne_bytes(iface.ip_addr.octets()),
+                };
+
+                // SAFETY: Requires a valid in_addr struct to be passed together with its size for
+                // checking which of the two it is. On errors a negative integer is returned.
+                unsafe {
+                    if setsockopt(
+                        socket.as_raw_fd(),
+                        IPPROTO_IP,
+                        IP_MULTICAST_IF,
+                        &addr as *const _ as *const _,
+                        mem::size_of_val(&addr) as _,
+                    ) < 0
+                    {
+                        bail!(
+                            source: io::Error::last_os_error(),
+                            "Failed joining multicast group for interface {}",
+                            iface.name,
+                        );
+                    }
+                }
+            }
+            #[cfg(target_os = "netbsd")]
+            {
+                let idx = (iface.index as u32).to_be();
+
+                // SAFETY: Requires a valid in_addr struct or interface index in network byte order
+                // to be passed together with its size for checking which of the two it is. On
+                // errors a negative integer is returned.
+                unsafe {
+                    if setsockopt(
+                        socket.as_raw_fd(),
+                        IPPROTO_IP,
+                        IP_MULTICAST_IF,
+                        &idx as *const _ as *const _,
+                        mem::size_of_val(&idx) as _,
+                    ) < 0
+                    {
+                        bail!(
+                            source: io::Error::last_os_error(),
+                            "Failed joining multicast group for interface {}",
+                            iface.name,
+                        );
+                    }
+                }
+            }
+
             Ok(())
         }
         #[cfg(any(target_os = "solaris", target_os = "illumos"))]
@@ -414,6 +501,29 @@ mod imp {
                     )
                 })?;
 
+            let addr = in_addr {
+                s_addr: u32::from_ne_bytes(iface.ip_addr.octets()),
+            };
+
+            // SAFETY: Requires a valid in_addr struct to be passed together with its size for
+            // checking which of the two it is. On errors a negative integer is returned.
+            unsafe {
+                if setsockopt(
+                    socket.as_raw_fd(),
+                    IPPROTO_IP,
+                    IP_MULTICAST_IF,
+                    &addr as *const _ as *const _,
+                    mem::size_of_val(&addr) as _,
+                ) < 0
+                {
+                    bail!(
+                        source: io::Error::last_os_error(),
+                        "Failed setting multicast interface {}",
+                        iface.name,
+                    );
+                }
+            }
+
             Ok(())
         }
     }
@@ -838,6 +948,31 @@ mod imp {
             }
         }
 
+        let addr = IN_ADDR {
+            S_un: IN_ADDR_0 {
+                S_addr: u32::from_ne_bytes(Ipv4Addr::new(0, 0, 0, iface.index as u8).octets()),
+            },
+        };
+
+        // SAFETY: Requires a valid IN_ADDR struct to be passed together with its size for checking
+        // which of the two it is. On errors a negative integer is returned.
+        unsafe {
+            if setsockopt(
+                socket.as_raw_socket(),
+                IPPROTO_IP as i32,
+                IP_MULTICAST_IF as i32,
+                &addr as *const _ as *const _,
+                mem::size_of_val(&addr) as _,
+            ) < 0
+            {
+                bail!(
+                    source: io::Error::last_os_error(),
+                    "Failed joining multicast group for interface {}",
+                    iface.name,
+                );
+            }
+        }
+
         Ok(())
     }
 

subprojects/gstreamer/libs/gst/net/gstptpclock.c

@@ -239,8 +239,8 @@ typedef struct
 
 typedef enum
 {
-  TYPE_EVENT = 0,               /* 64-bit monotonic clock time and PTP message is payload */
+  TYPE_EVENT = 0,               /* 8-bit interface index, 64-bit monotonic clock time and PTP message is payload */
-  TYPE_GENERAL = 1,             /* 64-bit monotonic clock time and PTP message is payload */
+  TYPE_GENERAL = 1,             /* 8-bit interface index, 64-bit monotonic clock time and PTP message is payload */
   TYPE_CLOCK_ID = 2,            /* 64-bit clock ID is payload */
   TYPE_SEND_TIME_ACK = 3,       /* 64-bit monotonic clock time, 8-bit message type, 8-bit domain number and 16-bit sequence number is payload */
 } StdIOMessageType;
@@ -292,6 +292,7 @@ typedef struct
   GstClockTime receive_time;
 
   PtpClockIdentity master_clock_identity;
+  guint8 iface_idx;
 
   guint8 grandmaster_priority_1;
   PtpClockQuality grandmaster_clock_quality;
@@ -306,6 +307,7 @@ typedef struct
 typedef struct
 {
   PtpClockIdentity master_clock_identity;
+  guint8 iface_idx;
 
   GstClockTime announce_interval;       /* last interval we received */
   GQueue announce_messages;
@@ -322,6 +324,7 @@ typedef struct
   GstClockTime follow_up_recv_time_local;
 
   GSource *timeout_source;
+  guint8 iface_idx;
   guint16 delay_req_seqnum;
   GstClockTime delay_req_send_time_local;       /* t3, -1 if we wait for FOLLOW_UP */
   GstClockTime delay_req_recv_time_remote;      /* t4, -1 if we wait */
@@ -355,6 +358,7 @@ typedef struct
   /* Last selected master clock */
   gboolean have_master_clock;
   PtpClockIdentity master_clock_identity;
+  guint8 iface_idx;
   guint64 grandmaster_identity;
 
   /* Last SYNC or FOLLOW_UP timestamp we received */
@@ -722,8 +726,11 @@ compare_announce_message (const PtpAnnounceMessage * a,
     else if (a->master_clock_identity.port_number >
         b->master_clock_identity.port_number)
       return 1;
-    else
+
-      g_assert_not_reached ();
+    if (a->iface_idx < b->iface_idx)
+      return -1;
+    else if (a->iface_idx > b->iface_idx)
+      return 1;
 
     return 0;
   }
@@ -819,13 +826,15 @@ select_best_master_clock (PtpDomainData * domain, GstClockTime now)
 
   if (domain->have_master_clock
       && compare_clock_identity (&domain->master_clock_identity,
-          &best->master_clock_identity) == 0) {
+          &best->master_clock_identity) == 0
+      && domain->iface_idx == best->iface_idx) {
     GST_DEBUG ("Master clock in domain %u did not change", domain->domain);
   } else {
     GST_DEBUG ("Selected master clock for domain %u: 0x%016" G_GINT64_MODIFIER
-        "x %u with grandmaster clock 0x%016" G_GINT64_MODIFIER "x",
+        "x %u on interface %u with grandmaster clock 0x%016" G_GINT64_MODIFIER
-        domain->domain, best->master_clock_identity.clock_identity,
+        "x", domain->domain, best->master_clock_identity.clock_identity,
-        best->master_clock_identity.port_number, best->grandmaster_identity);
+        best->master_clock_identity.port_number, best->iface_idx,
+        best->grandmaster_identity);
 
     domain->have_master_clock = TRUE;
     domain->grandmaster_identity = best->grandmaster_identity;
@@ -833,9 +842,11 @@ select_best_master_clock (PtpDomainData * domain, GstClockTime now)
     /* Opportunistic master clock selection likely gave us the same master
      * clock before, no need to reset all statistics */
     if (compare_clock_identity (&domain->master_clock_identity,
-            &best->master_clock_identity) != 0) {
+            &best->master_clock_identity) != 0
+        || domain->iface_idx != best->iface_idx) {
       memcpy (&domain->master_clock_identity, &best->master_clock_identity,
           sizeof (PtpClockIdentity));
+      domain->iface_idx = best->iface_idx;
       domain->mean_path_delay = 0;
       domain->last_delay_req = 0;
       domain->last_path_delays_missing = 9;
@@ -865,7 +876,8 @@ select_best_master_clock (PtpDomainData * domain, GstClockTime now)
 }
 
 static void
-handle_announce_message (PtpMessage * msg, GstClockTime receive_time)
+handle_announce_message (PtpMessage * msg, guint8 iface_idx,
+    GstClockTime receive_time)
 {
   GList *l;
   PtpDomainData *domain = NULL;
@@ -925,7 +937,7 @@ handle_announce_message (PtpMessage * msg, GstClockTime receive_time)
     PtpAnnounceSender *tmp = l->data;
 
     if (compare_clock_identity (&tmp->master_clock_identity,
-            &msg->source_port_identity) == 0) {
+            &msg->source_port_identity) == 0 && tmp->iface_idx == iface_idx) {
       sender = tmp;
       break;
     }
@@ -936,6 +948,7 @@ handle_announce_message (PtpMessage * msg, GstClockTime receive_time)
 
     memcpy (&sender->master_clock_identity, &msg->source_port_identity,
         sizeof (PtpClockIdentity));
+    sender->iface_idx = iface_idx;
     g_queue_init (&sender->announce_messages);
     domain->announce_senders =
         g_list_prepend (domain->announce_senders, sender);
@@ -968,6 +981,7 @@ handle_announce_message (PtpMessage * msg, GstClockTime receive_time)
   announce->sequence_id = msg->sequence_id;
   memcpy (&announce->master_clock_identity, &msg->source_port_identity,
       sizeof (PtpClockIdentity));
+  announce->iface_idx = iface_idx;
   announce->grandmaster_identity =
       msg->message_specific.announce.grandmaster_identity;
   announce->grandmaster_priority_1 =
@@ -991,7 +1005,7 @@ handle_announce_message (PtpMessage * msg, GstClockTime receive_time)
 static gboolean
 send_delay_req_timeout (PtpPendingSync * sync)
 {
-  guint8 message[STDIO_MESSAGE_HEADER_SIZE + 8 + 44] = { 0, };
+  guint8 message[STDIO_MESSAGE_HEADER_SIZE + 1 + 8 + 44] = { 0, };
   GstByteWriter writer;
   gsize written;
   GError *err = NULL;
@@ -1003,8 +1017,9 @@ send_delay_req_timeout (PtpPendingSync * sync)
       gst_clock_get_time (observation_system_clock);
 
   gst_byte_writer_init_with_data (&writer, message, sizeof (message), FALSE);
-  gst_byte_writer_put_uint16_be_unchecked (&writer, 8 + 44);
+  gst_byte_writer_put_uint16_be_unchecked (&writer, 1 + 8 + 44);
   gst_byte_writer_put_uint8_unchecked (&writer, TYPE_EVENT);
+  gst_byte_writer_put_uint8_unchecked (&writer, sync->iface_idx);
   gst_byte_writer_put_uint64_be_unchecked (&writer, send_time);
   gst_byte_writer_put_uint8_unchecked (&writer, PTP_MESSAGE_TYPE_DELAY_REQ);
   gst_byte_writer_put_uint8_unchecked (&writer, 2);
@@ -1058,6 +1073,7 @@ send_delay_req (PtpDomainData * domain, PtpPendingSync * sync)
   }
 
   domain->last_delay_req = now;
+  sync->iface_idx = domain->iface_idx;
   sync->delay_req_seqnum = domain->last_delay_req_seqnum++;
 
   /* IEEE 1588 9.5.11.2 */
@@ -1481,7 +1497,8 @@ out:
 }
 
 static void
-handle_sync_message (PtpMessage * msg, GstClockTime receive_time)
+handle_sync_message (PtpMessage * msg, guint8 iface_idx,
+    GstClockTime receive_time)
 {
   GList *l;
   PtpDomainData *domain = NULL;
@@ -1523,15 +1540,20 @@ handle_sync_message (PtpMessage * msg, GstClockTime receive_time)
 
   /* If we have a master clock, ignore this message if it's not coming from there */
   if (domain->have_master_clock
-      && compare_clock_identity (&domain->master_clock_identity,
+      && (compare_clock_identity (&domain->master_clock_identity,
-          &msg->source_port_identity) != 0)
+              &msg->source_port_identity) != 0
+          || domain->iface_idx != iface_idx)) {
+    GST_TRACE ("SYNC msg not from current clock master. Ignoring");
     return;
+  }
 
 #ifdef USE_OPPORTUNISTIC_CLOCK_SELECTION
   /* Opportunistic selection of master clock */
-  if (!domain->have_master_clock)
+  if (!domain->have_master_clock) {
     memcpy (&domain->master_clock_identity, &msg->source_port_identity,
         sizeof (PtpClockIdentity));
+    domain->iface_idx = iface_idx;
+  }
 #else
   if (!domain->have_master_clock)
     return;
@@ -1613,7 +1635,8 @@ handle_sync_message (PtpMessage * msg, GstClockTime receive_time)
 }
 
 static void
-handle_follow_up_message (PtpMessage * msg, GstClockTime receive_time)
+handle_follow_up_message (PtpMessage * msg, guint8 iface_idx,
+    GstClockTime receive_time)
 {
   GList *l;
   PtpDomainData *domain = NULL;
@@ -1643,8 +1666,9 @@ handle_follow_up_message (PtpMessage * msg, GstClockTime receive_time)
 
   /* If we have a master clock, ignore this message if it's not coming from there */
   if (domain->have_master_clock
-      && compare_clock_identity (&domain->master_clock_identity,
+      && (compare_clock_identity (&domain->master_clock_identity,
-          &msg->source_port_identity) != 0) {
+              &msg->source_port_identity) != 0
+          || domain->iface_idx != iface_idx)) {
     GST_TRACE ("FOLLOW_UP msg not from current clock master. Ignoring");
     return;
   }
@@ -1709,7 +1733,8 @@ handle_follow_up_message (PtpMessage * msg, GstClockTime receive_time)
 }
 
 static void
-handle_delay_resp_message (PtpMessage * msg, GstClockTime receive_time)
+handle_delay_resp_message (PtpMessage * msg, guint8 iface_idx,
+    GstClockTime receive_time)
 {
   GList *l;
   PtpDomainData *domain = NULL;
@@ -1740,9 +1765,12 @@ handle_delay_resp_message (PtpMessage * msg, GstClockTime receive_time)
 
   /* If we have a master clock, ignore this message if it's not coming from there */
   if (domain->have_master_clock
-      && compare_clock_identity (&domain->master_clock_identity,
+      && (compare_clock_identity (&domain->master_clock_identity,
-          &msg->source_port_identity) != 0)
+              &msg->source_port_identity) != 0
+          || domain->iface_idx != iface_idx)) {
+    GST_TRACE ("DELAY_RESP msg not from current clock master. Ignoring");
     return;
+  }
 
   if (msg->log_message_interval == 0x7f) {
     domain->min_delay_req_interval = GST_SECOND;
@@ -1809,7 +1837,8 @@ handle_delay_resp_message (PtpMessage * msg, GstClockTime receive_time)
 }
 
 static void
-handle_ptp_message (PtpMessage * msg, GstClockTime receive_time)
+handle_ptp_message (PtpMessage * msg, guint8 iface_idx,
+    GstClockTime receive_time)
 {
   /* Ignore our own messages */
   if (msg->source_port_identity.clock_identity == ptp_clock_id.clock_identity &&
@@ -1818,20 +1847,20 @@ handle_ptp_message (PtpMessage * msg, GstClockTime receive_time)
     return;
   }
 
-  GST_TRACE ("Message type %d receive_time %" GST_TIME_FORMAT,
+  GST_TRACE ("Message type %d iface idx %d receive_time %" GST_TIME_FORMAT,
-      msg->message_type, GST_TIME_ARGS (receive_time));
+      msg->message_type, iface_idx, GST_TIME_ARGS (receive_time));
   switch (msg->message_type) {
     case PTP_MESSAGE_TYPE_ANNOUNCE:
-      handle_announce_message (msg, receive_time);
+      handle_announce_message (msg, iface_idx, receive_time);
       break;
     case PTP_MESSAGE_TYPE_SYNC:
-      handle_sync_message (msg, receive_time);
+      handle_sync_message (msg, iface_idx, receive_time);
       break;
     case PTP_MESSAGE_TYPE_FOLLOW_UP:
-      handle_follow_up_message (msg, receive_time);
+      handle_follow_up_message (msg, iface_idx, receive_time);
       break;
     case PTP_MESSAGE_TYPE_DELAY_RESP:
-      handle_delay_resp_message (msg, receive_time);
+      handle_delay_resp_message (msg, iface_idx, receive_time);
       break;
     default:
       break;
@@ -1941,12 +1970,14 @@ have_stdout_body (GInputStream * stdout_pipe, GAsyncResult * res,
     case TYPE_EVENT:
     case TYPE_GENERAL:{
       GstClockTime receive_time = gst_clock_get_time (observation_system_clock);
+      guint8 iface_idx;
       GstClockTime helper_receive_time;
       PtpMessage msg;
 
-      helper_receive_time = GST_READ_UINT64_BE (stdout_buffer);
+      iface_idx = GST_READ_UINT8 (stdout_buffer);
+      helper_receive_time = GST_READ_UINT64_BE (stdout_buffer + 1);
 
-      if (parse_ptp_message (&msg, (const guint8 *) stdout_buffer + 8,
+      if (parse_ptp_message (&msg, (const guint8 *) stdout_buffer + 1 + 8,
               CUR_STDIO_HEADER_SIZE)) {
         dump_ptp_message (&msg);
         if (helper_receive_time != 0) {
@@ -1955,7 +1986,7 @@ have_stdout_body (GInputStream * stdout_pipe, GAsyncResult * res,
                       helper_receive_time)));
           receive_time = helper_receive_time;
         }
-        handle_ptp_message (&msg, receive_time);
+        handle_ptp_message (&msg, iface_idx, receive_time);
       }
       break;
     }