[Bug]: RDMA Device Misidentification in Container Environment

[uniqueni]

Bug Report

Environment:

• Hardware: H20 machine with 4 physical RDMA NICs
• Container setup: 2 GPUs requesting 2 virtual RDMA devices

Issue: RDMA device discovery incorrectly identifies devices in the container environment.

Reproduction:

1. Deploy container with 2 GPUs on H20 machine (4 physical RDMA NICs available)
2. Request 2 virtual RDMA devices for the container
3. Observe incorrect device identification and GID index lookup

Expected: Virtual RDMA devices should be correctly mapped and identified Actual: Device discovery fails to properly recognize the virtual RDMA devices

Before submitting...

• [ ] Ensure you searched for relevant issues and read the [documentation]

[uniqueni]

@stmatengss I will fix this bug because I have the environment

[stmatengss]

@stmatengss I will fix this bug because I have the environment

Thx!

[uniqueni]

fixed https://github.com/kvcache-ai/Mooncake/pull/1077