Conversation
The nightly-oci job has never passed. Both failure modes are TCG-timing artifacts, not functional bugs — the OCI smoke runs clean end-to-end under KVM (login as root, unshare namespaces, chroot, exec the entrypoint, and emit oci-uts-isolated). Mode 1 (the common one): machine-speed login keystrokes garble under slow TCG — a byte-reordering race in the console line discipline turns `root` into e.g. `orot` — so getty prints "Login incorrect" while the harness blindly waits out the full 900s prompt timeout for a `narf> ` that never comes. run-interactive's Stage 0 now retries the credential exchange (up to 8 attempts), resyncing on getty's re-prompt and fast-failing on "Login incorrect" instead of waiting for "Welcome"; it also types the login a bit slower (25ms vs 5ms per byte) to reduce the garble in the first place. Even a lost newline recovers, since the next attempt's bytes complete the stuck line. Mode 2 (rarer): a flaky TCG-only #UD right after the shell prompt that QEMU exits on. That fails fast (stdout EOF), so the workflow now passes --retries 3 — a cheap re-boot clears the flake while a genuine break still fails every attempt. Validated: the smoke passes under KVM and twice under TCG (the CI configuration: no balloon, 2048 MB, 900s timeouts) with the reworked login. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The nightly-oci job has never passed. Both failure modes are TCG-timing artifacts, not functional bugs — the OCI smoke runs clean end-to-end under KVM (login as root, unshare namespaces, chroot, exec the entrypoint, and emit oci-uts-isolated).
Mode 1 (the common one): machine-speed login keystrokes garble under slow TCG — a byte-reordering race in the console line discipline turns
rootinto e.g.orot— so getty prints "Login incorrect" while the harness blindly waits out the full 900s prompt timeout for anarf>that never comes. run-interactive's Stage 0 now retries the credential exchange (up to 8 attempts), resyncing on getty's re-prompt and fast-failing on "Login incorrect" instead of waiting for "Welcome"; it also types the login a bit slower (25ms vs 5ms per byte) to reduce the garble in the first place. Even a lost newline recovers, since the next attempt's bytes complete the stuck line.Mode 2 (rarer): a flaky TCG-only #UD right after the shell prompt that QEMU exits on. That fails fast (stdout EOF), so the workflow now passes --retries 3 — a cheap re-boot clears the flake while a genuine break still fails every attempt.
Validated: the smoke passes under KVM and twice under TCG (the CI configuration: no balloon, 2048 MB, 900s timeouts) with the reworked login.