Remove the use of the iptrie crate

[mvachhar]

iptrie is poorly tested and very buggy. This series of commits replaces it with PrefixMap. It does so by adding an lpm package that contains the Prefix implementation and a TrieMap trait that the implementation can hide behind so we can swap to different implementations later.

In the interest of time, there aren't many additional tests for the new stuff, but instead we rely on PrefixMap's tests which appear far superior to iptrie (with 90+% coverage) and the routing and nat tests. It is a better use of time to add fuzzing to routing which will uncover any trie issues and cover the routing crate.

Fixes #546

[Copilot]

Pull Request Overview

This PR removes the use of the iptrie crate and replaces it with a new lpm package providing Prefix and a TrieMap trait. Key changes include:

• Introducing the lpm crate with prefix and trie implementations.
• Swapping all iptrie::map::RTrieMap usages for PrefixMapTrieWithDefault and updating lookup and length calls.
• Updating imports across routing, NAT, and management modules to use lpm::prefix instead of the old prefix or iptrie paths.

Reviewed Changes

Copilot reviewed 38 out of 39 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
routing/src/rpc_adapt.rs	Swapped crate::prefix::Prefix import for lpm::prefix::Prefix
routing/src/rib/vrf.rs	Replaced RTrieMap with PrefixMapTrieWithDefault and updated initialization and LPM calls
routing/src/lib.rs	Removed the now-obsolete prefix module export
routing/src/fib/fibtype.rs	Updated FIB trie fields and lookups to use the new lpm trie API
nat/src/stateless/...	Updated NAT prefix trie to use lpm::prefix and lpm::trie types
mgmt/src/...	Updated management and GRPC converter modules to import from lpm
lpm/src/...	Added full implementation of prefix and trie logic in the lpm crate
Cargo.toml files	Added lpm as a workspace dependency and removed iptrie

Comments suppressed due to low confidence (2)

routing/src/rib/vrf.rs:184

• Consider adding a unit test that verifies the default routes are properly initialized for both IPv4 and IPv6, and that lookups on an empty VRF do not panic.

            unsafe { PrefixMapTrieWithDefault::with_capacity(Vrf::DEFAULT_IPV6_CAPACITY) };

mgmt/src/models/internal/natconfig/collapse.rs:64

• The original break in this exclusion case was correct: once an exclusion prefix removes a matching allowed prefix, you should stop processing further allowed prefixes for this exclusion and move to the next exclusion. Changing it to continue alters that logic and may lead to incorrect results.

                continue;

[qmonnet]

Looks good, thanks a lot for this work! Nice abstraction work for the trie map.

I've got some comments (see inline), but it's only minor things, the changes look good overall.

[Fredi-raspall]

Overall LGTM @mvachhar .
I would, for simplicity and clarity:

• have a single trait by removing TrieMapNew and adding a new() method to Triemap, removing all the with_capacity and with_root, etc..
• let the users of the crate decide and handle if there has to be a root or not, including the handling of the removal of objects for 0.0.0.0/0 or ::0/0. IMO, this is not something that this crate should decide and this avoids "dangerous" situations, which are better handled by the users of the crate.
• rename TrieMapWithDefault and the module as just Trie.

[daniel-noland]

I made comments but I think they are all outdated at this point.

Regardless, I did a careful review and I'm baiscally happy with where this is at

[daniel-noland]

Ideally, lookup, get, and get_mut, and remove would accept a Q: Borrow<Self::Prefix> in the same way that std's HashMap et al do.

Addressed the comments, but dismissing due to timezone so we can merge today.