Files
SenMeshNet/docs/research/sftp-shared-channel-cancellation.md
zhaolei 2f9f8da48e
Some checks failed
AI automation / Route event (push) Has been cancelled
AI automation / Hand reopened issue to maintainers (push) Has been cancelled
AI automation / Clean source issue state (push) Has been cancelled
AI automation / Reconcile handoffs (push) Has been cancelled
AI automation / Classify issue (push) Has been cancelled
AI automation / Claude Code smoke (push) Has been cancelled
AI automation / Review issue follow-up (push) Has been cancelled
AI automation / Publish issue follow-up (push) Has been cancelled
AI automation / Implement with Claude Code (push) Has been cancelled
AI automation / Publish implement PR (push) Has been cancelled
AI automation / Continue queued issue comments (push) Has been cancelled
AI automation / Codex review loop (push) Has been cancelled
AI automation / Publish Codex fix (push) Has been cancelled
AI automation / Clear Codex dispatch marker (push) Has been cancelled
AI automation / Own PR re-request Codex (push) Has been cancelled
AI automation / External PR re-request Codex (push) Has been cancelled
AI automation / Poll Codex reaction / retry (push) Has been cancelled
Initial commit: NetMesh smart ops platform with 3D topology, inspection reports, RCA and AI ops assistant
2026-10-15 13:53:04 +08:00

1.2 KiB

Shared SSH connection cancellation audit

Cancelling a transfer while its isolated SFTP channel is opening calls the bounded-open abort path. Previously that path ended/destroyed the physical SSH transport. Production pooled SFTP sessions normally share a terminal SSH connection, so cancelling one transfer can disconnect terminals, browsing and other transfers. The same teardown happened on the channel-open deadline.

The real startTransfer/cancelTransfer path reproduces physical end/destroy calls; its old test double lacked those methods and could not detect the problem.

The repair rejects the abandoned caller promptly, retains the shared transport, closes a channel arriving after abandonment, and blocks further channel-open allocation on that transport while any abandoned request remains unresolved. This bounds retry accumulation without queuing controls behind a transfer. Healthy parallel opens remain allowed; callback settlement or transport closure releases the abandoned-open bookkeeping. Existing active channels remain usable.

This separately reproduced mechanism is relevant to the connection-loss themes in #2973 and #2832, but does not establish the cause of their VPN/jump-host reports.