* Add general-purpose packet cache
This commit adds a caching system that will save packet data in a much
more compact form than the regular MeshPacket protobuf. It cannot be
worked with directly to the same degree (although the packet header is
available), but consumes *much* less memory, and as a result can be used
to temporarily store large numbers of packets.
Cached packets can be retrieved either by their (from, id) tuple, or by
their hash.
This cache is a pre-requisite for the upcoming packet replay feature.
* Remove debug initialiser
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
* Fix ordering
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
* Add missing size assignment
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
* Add comments for hash & bucket macros
* Make it clear that this field stores a map of the original data
---------
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: Ben Meadors <benmmeadors@gmail.com>
* If a packet is heard multiple times, rebroadcast using the highest hop limit
Sometimes a packet will be in the TX queue waiting to be transmitted,
when it is overheard being rebroadcast by another node, with a higher
hop limit remaining. When this occurs, modify the pending packet in
the TX queue to avoid unnecessarily wasting hops.
* Reprocess instead of modifying queued packet
In order to ensure that the traceroute module works correctly, rather
than modifying the hop limnit of the existing queued version of the
packet, simply drop it ifrom the queue and reprocess the version of the
packet with the superior hop limit.
* Update protobufs submodule
* Merge upstream/develop into overheard-hoptimisation branch
Resolved conflicts in:
- src/mesh/FloodingRouter.cpp: Integrated hop limit optimization with refactored duplicate handling
- src/mesh/MeshPacketQueue.h: Kept both hop_limit_lt parameter and new find() method
* Improve method naming and code clarity
- Rename findPacket() to getPacketFromQueue() for better clarity
- Make code DRY by having find() use getPacketFromQueue() internally
- Resolves method overloading conflict with clearer naming
* If a packet is heard multiple times, rebroadcast using the highest hop limit
Sometimes a packet will be in the TX queue waiting to be transmitted,
when it is overheard being rebroadcast by another node, with a higher
hop limit remaining. When this occurs, modify the pending packet in
the TX queue to avoid unnecessarily wasting hops.
* Improve router role checking using IS_ONE_OF macro
- Replace multiple individual role checks with cleaner IS_ONE_OF macro
- Add CLIENT_BASE support as suggested in PR #7992
- Include MeshTypes.h for IS_ONE_OF macro
- Makes code more maintainable and consistent with other parts of codebase
* Apply IS_ONE_OF improvement to NextHopRouter.cpp
- Replace multiple individual role checks with cleaner IS_ONE_OF macro
- Add CLIENT_BASE support for consistency
- Include MeshTypes.h for IS_ONE_OF macro
- Matches the pattern used in FloodingRouter.cpp
* Create and apply IS_ROUTER_ROLE() macro across codebase
- Add IS_ROUTER_ROLE() macro to meshUtils.h for consistent router role checking
- Update FloodingRouter.cpp to use macro in multiple locations
- Update NextHopRouter.cpp to use macro
- Include CLIENT_BASE role support
* Core Changes:
- Add hop_limit field to PacketRecord (17B→20B due to alignment)
- Extend wasSeenRecently() with wasUpgraded parameter
- Enable router optimization without duplicate app delivery
- Handle ROUTER_LATE delayed transmission properly
Technical Details:
- Memory overhead: ~4000 bytes for 1000 records
- Prevents duplicate message delivery while enabling routing optimization
- Maintains protocol integrity for ACK/NAK handling
- Supports upgrade from hop_limit=0 to hop_limit>0 scenarios
* Delete files accdentally added for merge
* Trunk formatting
* Packets are supposed to be unsigned. Thankfully, it's only a log message.
* Upgrade all packets, not just 0 hop packets.
* Not just unsigned, but hex. Updating packet log IDs.
* Fixed order of operations issue that prevented packetrs from being removed from the queue
* Fixing some bugs after testing. Only storing the maximum hop value in PacketRecord which makes sense. Also, updating messaging to make more sense in the logs.
* Fixed flow logic about how to handle re-inserting duplicate packets. Removed IS_ROUTER_ROLE macro and replaced it with better isRebroadcaster().
* Add logic to re-run modules, but avoid re-sending to phone.
* Refactor how to process the new packet with hops. Only update nodeDB and traceRouteModule.
* - Apply changes to both FloodingRouter and NextHopRouter classes to make packets mutable for traceroute
- MESHTASTIC_EXCLUDE_TRACEROUTE guard for when we don't want traceroute
* Allow MeshPacket to be modified in-place in processUpgradePacket
* let's not make a copy where a copy is unncessary.
---------
Co-authored-by: Clive Blackledge <clive@ansible.org>
Co-authored-by: Clive Blackledge <git@ansible.org>
Co-authored-by: Ben Meadors <benmmeadors@gmail.com>
* Correct function comment
* Enqueue the intended packet, not the pointer to what we just dropped!
* Add some log output when we drop packets due to a full queue
* Make it clear when a non-late packet is dropped
* Remove from queue before release, not after
* Erase dropped packet from queue
* Declared type
* Log TX queue length after every send
* Fix operand order
* Add worst-case cap on TX delay vs current time
Will always rebroadcast packets, but will do so after all other modes.
Intended for router nodes that are there to provide additional coverage
in areas not already covered by other routers, or to bridge around
problematic terrain, but should not be given priority over other routers
in order to avoid unnecessaraily consuming hops.
By default, this role will rebroadcast during the normal client window.
However, if another node is overheard rebroadcasting the packet, then it
will be moved to a second window *after* the normal client one, with the
same timing behaviour.
EraydandBen Meadors
