[GH-ISSUE #3502] [TE][CXI]: Split transfers that straddle BufferDesc chunk boundaries #898

New issue

Open

opened 2026-10-29 21:20:41 +00:00 by bsakaguchi · 1 comment

bsakaguchi commented

2026-10-29 21:20:41 +00:00

Owner

Originally created by @staryxchen on GitHub (Aug 18, 2026).
Original GitHub issue: https://github.com/kvcache-ai/Mooncake/issues/3502

Parent: #3500

Problem

CxiTransport::registerLocalMemory already auto-splits buffers larger than max_mr_size into multiple BufferDescs.

CxiTransport::submitTransferTask still posts one slice per request. A transfer that crosses a chunk seam fails selectDevice.

Until this is fixed, Store must keep splitting CXI segments in GetTransportRegistrationLimit().

Proposed work

Same as EFA / RDMA: cap each posted slice at source and target BufferDesc remainder, then bind keys per sub-slice.

Test plan

Needs CXI / Slingshot hardware. Suggested coverage matching rdma_large_mr_test.cpp (destination seam, source seam, past-first-chunk WRITE).

I do not have CXI hardware to implement or verify this.

Originally created by @staryxchen on GitHub (Aug 18, 2026). Original GitHub issue: https://github.com/kvcache-ai/Mooncake/issues/3502 Parent: #3500 ## Problem `CxiTransport::registerLocalMemory` already auto-splits buffers larger than `max_mr_size` into multiple `BufferDesc`s. `CxiTransport::submitTransferTask` still posts **one slice per request**. A transfer that crosses a chunk seam fails `selectDevice`. Until this is fixed, Store must keep splitting CXI segments in `GetTransportRegistrationLimit()`. ## Proposed work Same as EFA / RDMA: cap each posted slice at source and target `BufferDesc` remainder, then bind keys per sub-slice. ## Test plan Needs CXI / Slingshot hardware. Suggested coverage matching `rdma_large_mr_test.cpp` (destination seam, source seam, past-first-chunk WRITE). I do not have CXI hardware to implement or verify this.

bsakaguchi added the

Transfer Engine

feature request

labels

2026-10-29 21:20:41 +00:00

bsakaguchi commented

2026-10-29 21:20:42 +00:00

Author

Owner

@github-actions[bot] commented on GitHub (Aug 18, 2026):

Thanks for opening this issue, @staryxchen!

Field	Value
Issue	#3502
GitHub user ID	`151037142`
Reporter	@staryxchen

A maintainer will triage this when possible. To help us respond faster, please include:

Mooncake version or commit SHA
Environment (OS, CUDA/driver, RDMA stack if relevant)
Steps to reproduce and expected vs. actual behavior

Useful links: Documentation · Contributing guide

This message was posted automatically by the issue bot.

@github-actions[bot] commented on GitHub (Aug 18, 2026): Thanks for opening this issue, @staryxchen! | Field | Value | |-------|-------| | **Issue** | #3502 | | **GitHub user ID** | `151037142` | | **Reporter** | @staryxchen | A maintainer will triage this when possible. To help us respond faster, please include: - Mooncake version or commit SHA - Environment (OS, CUDA/driver, RDMA stack if relevant) - Steps to reproduce and expected vs. actual behavior Useful links: [Documentation](https://kvcache-ai.github.io/Mooncake/) · [Contributing guide](https://github.com/kvcache-ai/Mooncake/blob/main/CONTRIBUTING.md) > This message was posted automatically by the issue bot.