[Bugfix][TE] Reject MC_IB_PORT=0 instead of disabling every RNIC

[SongOf]

Description

loadGlobalConfig validated MC_IB_PORT with val >= 0, so 0 passed and became globalConfig().port. IB port numbers are 1-based: RdmaContext::openRdmaDevice then calls ibv_query_port(ctx, 0, ...) on every device, which fails, so
construct() returns ERR_CONTEXT and each RNIC is disableDeviced. With the topology empty, TransferEngine::init() fails with ERR_DEVICE_NOT_FOUND — one mistyped environment variable takes down all RDMA on the node, and no log
line names the value as the cause.

atoi() returning 0 for unparseable input made this reachable by typo as well: MC_IB_PORT=abc selected port 0 instead of being ignored.

Fix: require val > 0, matching the other bounded knobs in the same function. Out-of-range and unparseable values now keep the default port 1 and log the existing warning.

Module

• Transfer Engine (mooncake-transfer-engine)
• Mooncake Store (mooncake-store)
• Mooncake EP (mooncake-ep)
• Mooncake PG (mooncake-pg)
• Integration (mooncake-integration)
• P2P Store (mooncake-p2p-store)
• Python Wheel (mooncake-wheel)
• Common (mooncake-common)
• Mooncake RL (mooncake-rl)
• CI/CD
• Docs
• Other

Type of Change

• Bug fix
• New feature
• Refactor
• Breaking change
• Documentation update
• Performance improvement
• Other

How Has This Been Tested?

Test commands:
cmake .. -DBUILD_UNIT_TESTS=ON && cmake --build . -j$(nproc)
ctest -R config_test --output-on-failure

Test results:

• Unit tests pass
• Integration tests pass (if applicable)
• Manual testing done (describe below)

config_test gains 6 IbPortEnvTest cases: unset default, valid override, and rejection of 0 / 256 / -1 / non-numeric. All pass.

Verified the tests are not inert: with val >= 0 restored, ZeroIsRejected and NonNumericIsRejected fail; with the fix they pass.

Full ctest: 28/31 pass. The 3 failures (tcp_transport_test, transfer_metadata_test, memory_location_test) are pre-existing environment issues in the build container — no etcd metadata server, and no CAP_SYS_NICE for mbind —
and are unrelated to this change.

Checklist

• I have performed a self-review of my own code
• I have formatted my code using ./scripts/code_format.sh
• I have run pre-commit run --all-files and all hooks pass
• I have updated the documentation (if applicable)
• I have added tests to prove my changes are effective
• For changes >500 LOC: I have filed an RFC issue

pre-commit was run on the two touched files only (all hooks pass); --all-files was not run because cmake-format rewrites unrelated CMakeLists.txt files. Docs need no change — MC_IB_PORT is documented as "default value 1" and
never advertised 0 as valid.

AI Assistance Disclosure

• No AI tools were used
• AI tools were used (specify below)

[staryxchen]

Good catch! LGTM.