Send outgoing packets for every address by anti-social · Pull Request #462 · keepsimple1/mdns-sd

anti-social · 2026-04-23T15:01:07Z

Partially fixes #459

keepsimple1 · 2026-04-25T23:36:39Z

Thanks for submitting the PR! I understand it's still in Draft, but I have a general comment: can we split this PR into smaller ones that are relatively independent (or built on-top) of each other? That will make review and merge much easier.

For example, those cleanup changes can be in one PR:

println! → debug! in DnsHostInfo::write.
u32-overflow fix in get_expiration_time.
add Debug derives on DnsOutPacket, etc.

anti-social · 2026-04-27T16:40:53Z

I've fixed u32 overflow in a PR that optimizes dns packet serialization because I caught it in tests that were written for the PR

#465 #466 #467

anti-social · 2026-04-27T19:55:50Z

-    /// Resend unregister packet.
-    UnregisterResend(Vec<u8>, u32, bool), // (packet content, if_index, is_ipv4)
+    /// Resend unregister packets.
+    UnregisterResend(Vec<DnsOutPacket>, u32, bool), // (packets to resend, if_index, is_ipv4)


We changed this variant thus DnsOutPacket should implement std::fmt::Debug

keepsimple1

As this is somewhat non-trivial change, could you please add a detailed PR description:

What's the problem? (motivation)
Changes (approach or the solution)

That will also help the reader understand which part of #459 is addressed by this PR.

And could you please also add new test case to cover the multi-address-per-interface case that motivated this change?

Thanks!

keepsimple1 · 2026-05-02T04:33:44Z

-
-    send_dns_outgoing_impl(out, if_name, my_intf.index, if_addr, sock, port)
+    let mut invalid_interfaces = HashSet::new();
+    let packets = out.to_packets();


It seems it would return packets even when no address was actually sent on.

keepsimple1 · 2026-05-02T04:35:29Z

-        };
-        sent_vec.into_iter().next().unwrap_or_default()
+        let (sent_packets, invalid_intf_addrs) = send_dns_outgoing(&out, intf, sock, self.port);
+        let _ = self.send_cmd_to_self(Command::InvalidIntfAddrs(invalid_intf_addrs));


This sends the InvalidIntfAddrs command even if invalid_intf_addrs is empty.

keepsimple1 · 2026-05-02T04:40:07Z

    let is_ipv4 = sock.domain() == Domain::IPV4;
    if let Some(out) = prepare_announce(info, intf, dns_registry, is_ipv4) {
-        let _ = send_dns_outgoing(&out, intf, sock, port, None)?;
+        let _ = send_dns_outgoing(&out, intf, sock, port);


This call used to propagate the error back to the caller. Now it silently ignores the error.

As I can see we need a little deeper refactoring to track status for pair of an interface and an address.

Maybe we could change this function's return type to match the new return type of send_dns_outgoing so that it can bubble it up?

Something like this?

fn announce_service_on_intf( dns_registry: &mut DnsRegistry, info: &ServiceInfo, intf: &MyIntf, sock: &PktInfoUdpSocket, port: u16, ) -> (bool, HashSet<Interface>) { let is_ipv4 = sock.domain() == Domain::IPV4; if let Some(out) = prepare_announce(info, intf, dns_registry, is_ipv4) { let (sent_packets, invalid) = send_dns_outgoing(&out, intf, sock, port); return (!sent_packets.is_empty(), invalid); } (false, HashSet::new()) }

The call sites need to change to handle a HashSet but I think doable?

keepsimple1 · 2026-05-02T04:45:41Z


    /// Returns a list of actual DNS packet data to be sent on the wire.
-    pub fn to_data_on_wire(&self) -> Vec<Vec<u8>> {
+    pub(crate) fn _to_data_on_wire(&self) -> Vec<Vec<u8>> {


This change seems odd as it's still used. If you removed it completely and replace the call sites with direct alternative calls, I'm ok with that.

keepsimple1 · 2026-05-15T04:16:15Z

@anti-social friendly bump: will you be able to move forward with this PR?

keepsimple1 · 2026-05-18T03:12:25Z

    let is_ipv4 = sock.domain() == Domain::IPV4;
    if let Some(out) = prepare_announce(info, intf, dns_registry, is_ipv4) {
-        let _ = send_dns_outgoing(&out, intf, sock, port, None)?;
+        let _ = send_dns_outgoing(&out, intf, sock, port);


Maybe we could change this function's return type to match the new return type of send_dns_outgoing so that it can bubble it up?

Something like this?

fn announce_service_on_intf( dns_registry: &mut DnsRegistry, info: &ServiceInfo, intf: &MyIntf, sock: &PktInfoUdpSocket, port: u16, ) -> (bool, HashSet<Interface>) { let is_ipv4 = sock.domain() == Domain::IPV4; if let Some(out) = prepare_announce(info, intf, dns_registry, is_ipv4) { let (sent_packets, invalid) = send_dns_outgoing(&out, intf, sock, port); return (!sent_packets.is_empty(), invalid); } (false, HashSet::new()) }

The call sites need to change to handle a HashSet but I think doable?

keepsimple1 · 2026-05-18T03:21:13Z

-            self.notify_monitors(DaemonEvent::Respond(if_name));
+                self.increase_counter(Counter::Respond, 1);
+                self.notify_monitors(DaemonEvent::Respond(if_name));
+            }


I think we should have a fallback to fanout when matched_source is None (current diff silently drops responses in such case).

something like

else { // No matched source — fan out across all addresses on the interface so // the querier may hear us from one subnet. let (_, invalid) = send_dns_outgoing(&out, intf, &sock.pktinfo, self.port); if !invalid.is_empty() { let _ = self.send_cmd_to_self(Command::InvalidIntfAddrs(invalid)); } }

Also this branch now needs a rebase (sorry, it's a bit more work) as unicast_dest was added through send_dns_outgoing, and the fanout logic needs to be gated on unicast_dest.is_none().

and if we could do that, we can move the counter update out after if..else.. block, i.e. same as before

let if_name = intf.name.clone(); self.increase_counter(Counter::Respond, 1); self.notify_monitors(DaemonEvent::Respond(if_name));

anti-social force-pushed the multi-addr-if branch 4 times, most recently from d157e97 to 555d157 Compare April 23, 2026 17:41

anti-social mentioned this pull request Apr 23, 2026

fix: use a source IP matching the querier's subnet when responding #460

Merged

anti-social force-pushed the multi-addr-if branch from 5fcd11e to 392a3ac Compare April 27, 2026 19:50

anti-social commented Apr 27, 2026

View reviewed changes

anti-social force-pushed the multi-addr-if branch 3 times, most recently from 95ca00f to 0814fba Compare May 1, 2026 22:26

Send outgoing packets for every address

4bf97bc

anti-social force-pushed the multi-addr-if branch from 0814fba to 4bf97bc Compare May 1, 2026 22:30

anti-social marked this pull request as ready for review May 1, 2026 22:34

keepsimple1 reviewed May 2, 2026

View reviewed changes

Some fixes

3d4f26e

keepsimple1 reviewed May 18, 2026

View reviewed changes

Conversation

anti-social commented Apr 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

keepsimple1 commented Apr 25, 2026

Uh oh!

anti-social commented Apr 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

keepsimple1 left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

keepsimple1 commented May 15, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

keepsimple1 May 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

anti-social commented Apr 23, 2026 •

edited

Loading

anti-social commented Apr 27, 2026 •

edited

Loading

keepsimple1 left a comment •

edited

Loading

keepsimple1 May 18, 2026 •

edited

Loading