Skip to content

fix(tekton): require the create event's ref_type on every create trigger - #5569

Merged
ti-chi-bot[bot] merged 1 commit into
mainfrom
fix/create-event-ref-type
Nov 13, 2026
Merged

ti-chi-bot[bot] merged 1 commit into
mainfrom
fix/create-event-ref-type

Conversation

@wuhuizuo

Copy link
Copy Markdown
Contributor

Problem

GitHub delivers branch creation and tag creation as the same create event.
body.ref carries the bare short name in both cases; body.ref_type
("tag" / "branch") is the only discriminator. Of the 26 create-event triggers
only tidbcloud/serverless/tag-create.yaml tested it.

So every release trigger fired on a branch whose name happens to match the
tag regex — v9.0.0, v26.3.1, and ^v.+ for tiflow-operator. And
bindings/github-tag-create.yaml passes body.ref as both git-ref and
git-revision, with the pipeline's default +refs/heads/*:refs/heads/*
refspec, so the branch was fetched and built exactly like the tag: with
gcp-classic-build-params — the official …/hub registry and live publisher
URLs, under the image-releaser ServiceAccount.

Someone who can push a branch to an allow-listed product repository (a much wider
population than those who may cut release tags) could therefore drive the release
build path. The mirror case exists too: the branch-creation triggers
(OWNERS-sync, README placeholder bump) also fire on tag creation.

Change

Add a first CEL interceptor deciding the event kind, before any name or
repository filter runs:

  • 23 tag-creation triggers → body.ref_type == 'tag'
  • 3 branch-creation triggers → body.ref_type == 'branch'

Verification

  • New structural test: every create-event trigger must carry the matching
    ref_type predicate, so a trigger added later cannot silently omit it.
  • New behavioural test: the same tag-shaped ref (v1.2.3-alpha.0) passes the
    pingcap/tiproxy release trigger with ref_type: tag and is refused with
    ref_type: branch; the tikv/tikv branch trigger behaves the other way round.
  • All 26 changed files parse, deno test -A tekton/tests/ green (39 passed).

GitHub delivers branch creation and tag creation as the same `create` event.
`body.ref` carries the bare name in both cases, so `body.ref_type` is the only
discriminator — but only `tidbcloud/serverless/tag-create.yaml` tested it.

Every release trigger therefore fired on a *branch* whose name happens to match
the tag regex (`v9.0.0`, `v26.3.1`, `^v.+` for tiflow-operator), and
`bindings/github-tag-create.yaml` passes `body.ref` as both `git-ref` and
`git-revision`, so the branch was fetched and built exactly like the tag: with
`gcp-classic-build-params`, i.e. the official `hub` registry and live publisher
URLs. Branch push rights are a much wider population than tag-cut rights. The
mirror case is the branch-creation triggers, which also fire on tag creation.

Add a first interceptor to all 23 tag-creation triggers (`body.ref_type ==
'tag'`) and all 3 branch-creation triggers (`body.ref_type == 'branch'`), so the
event kind is decided before any name or repository filter runs.

Verified:
- `tekton/tests/trigger-security.test.ts` gains a structural check that every
  create-event trigger carries the matching `ref_type` predicate (so a new
  trigger cannot be added without it) and a behavioural pair: a tag-shaped `ref`
  with `ref_type: branch` no longer passes the tiproxy release trigger, and a
  tag creation no longer passes the tikv branch trigger;
- all 26 changed files parse, and `deno test -A tekton/tests/` is green
  (39 passed).
@wuhuizuo

Copy link
Copy Markdown
Contributor Author

/approve

1 similar comment
@wuhuizuo

Copy link
Copy Markdown
Contributor Author

/approve

@ti-chi-bot ti-chi-bot Bot added the approved label Nov 13, 2026
@ti-chi-bot

ti-chi-bot Bot commented Nov 13, 2026

Copy link
Copy Markdown

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: wuhuizuo

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Details Needs approval from an approver in each of these files:

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

1 similar comment
@ti-chi-bot

ti-chi-bot Bot commented Nov 13, 2026

Copy link
Copy Markdown

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: wuhuizuo

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Details Needs approval from an approver in each of these files:

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

@ti-chi-bot ti-chi-bot Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I have already done a preliminary review for you, and I hope to help you do a better job.

Summary

This pull request addresses a critical security issue where GitHub create events for both branches and tags could inadvertently trigger incorrect pipelines due to the lack of a discriminator for ref_type. The fix introduces ref_type checks (tag or branch) for all 26 create-event triggers to ensure only the appropriate pipelines are executed. The approach is robust, adding explicit CEL interceptors to enforce ref_type validation and including structural and behavioral tests to guarantee future compliance. The PR is well-structured, but a few areas could benefit from improvement in maintainability and edge case handling.


Critical Issues

  • Test Coverage for Edge Cases:
    • File: tekton/tests/trigger-security.test.ts, Lines 35-66.
    • Issue: The behavioral test covers only a single tag-shaped branch and a single branch-shaped tag. Other potential edge cases, such as malformed or unexpected ref_type values (e.g., null, empty string, invalid strings) are not tested.
    • Suggestion: Extend the behavioral test to include cases for invalid or missing ref_type values. For example:
      Deno.test("invalid ref_type values do not trigger any pipelines", () => {
        const tagTrigger = load<NamedTrigger>(
          `${triggerDirectory}pingcap/tiproxy/git-create-tag.yaml`,
        );
        const invalidEvent = payload("pingcap/tiproxy", "create");
        invalidEvent.ref_type = "invalid"; // Unexpected value
        assertEquals(filter(tagTrigger, invalidEvent), false);
      
        invalidEvent.ref_type = ""; // Empty string
        assertEquals(filter(tagTrigger, invalidEvent), false);
      
        delete invalidEvent.ref_type; // Missing value
        assertEquals(filter(tagTrigger, invalidEvent), false);
      });

Code Improvements

  • Centralize ref_type Validation Logic:

    • Files: Multiple YAML files (e.g., git-create-tag.yaml, git-create-branch.yaml).
    • Issue: The body.ref_type filter logic (body.ref_type == 'tag' or body.ref_type == 'branch') is duplicated across all triggers, increasing maintenance overhead and risk of future inconsistencies.
    • Suggestion: Abstract this common filter logic into a reusable CEL interceptor template or shared function. For instance:
      interceptors:
        - name: validate-ref-type
          ref: { name: cel }
          params:
            - name: filter
              value: "body.ref_type == '{{ expected_ref_type }}'"
      Then, use this template in individual trigger files with parameter substitution for expected_ref_type.
  • Error Handling for Missing ref_type:

    • Files: Multiple YAML files (e.g., git-create-tag.yaml, git-create-branch.yaml).
    • Issue: The CEL filter assumes body.ref_type is always present, but there’s no safeguard for cases where it might be missing (e.g., due to malformed payload).
    • Suggestion: Update CEL filters to explicitly check for the presence of body.ref_type before comparison:
      value: "has(body.ref_type) && body.ref_type == 'tag'"

Best Practices

  • Improve Test Descriptions:

    • File: tekton/tests/trigger-security.test.ts, Lines 35-66.
    • Issue: Test descriptions like "a branch named like a release tag does not drive a release build" could be more concise and specific.
    • Suggestion: Use descriptions that focus on the key behavior being validated, e.g., "validate ref_type prevents incorrect pipeline execution".
  • Reduce Code Duplication in Tests:

    • File: tekton/tests/trigger-security.test.ts, Lines 35-66.
    • Issue: The test code for loading triggers and applying filters is repeated. This can lead to harder-to-maintain tests.
    • Suggestion: Abstract common logic into reusable helper functions:
      function validateTrigger(triggerPath: string, payload: object, expected: boolean): void {
        const trigger = load<NamedTrigger>(triggerPath);
        assertEquals(filter(trigger, payload), expected);
      }
  • Documentation for New Structural Test:

    • File: tekton/tests/trigger-security.test.ts.
    • Issue: The purpose of the structural test checking all create-event triggers is not documented in the code itself, though it’s described in the PR description.
    • Suggestion: Add a brief comment explaining the goal of the test, e.g.:
      // Ensure all create-event triggers include a ref_type filter to prevent incorrect pipeline executions.

Summary of Recommendations

Critical Issues

  1. Add edge case tests for invalid or missing ref_type.

Code Improvements

  1. Centralize ref_type filter logic across YAML files to reduce duplication.
  2. Enhance CEL filters to handle missing body.ref_type explicitly.

Best Practices

  1. Use more concise and specific test descriptions.
  2. Refactor test code to reduce duplication using helper functions.
  3. Document the purpose of the structural tests to improve maintainability.

These changes will improve the robustness, maintainability, and test coverage of the PR.

@ti-chi-bot ti-chi-bot Bot added the size/L label Nov 13, 2026
ti-chi-bot Bot pushed a commit that referenced this pull request Nov 13, 2026
…5570)

## Problem

`pull-verify-tekton-yaml` reads `.apiVersion`, `.kind` and
`.metadata.name` once
per file. A file holding **two** documents (`_/git-create-branch.yaml`
holds two
`Trigger`s) makes `yq` print `Trigger\n---\nTrigger`, which the kind
allowlist
rejects:

```
unknown or missing kind: 'Trigger
---
Trigger'
```

So the check fails on a valid file the moment that file is touched —
which is
what happened in #5569.

Two kinds of file it could never accept either:

- `kustomization.yaml` — not a Tekton manifest, yet it was flagged for
  `Kustomization` + empty `metadata.name`;
- `StepAction` — a Tekton resource missing from the allowlist, so every
  `tekton/v1/step-actions/*` change failed.

## Change

Evaluate **every document** and report the ones that are not a named
Tekton
resource; skip `kustomization.yaml`; add `StepAction` to the allowlist.
A file's
document count no longer matters.

## Verification

- Ran the check against **every** Tekton manifest in the repository
  (`PULL_BASE_SHA=<empty tree>`): 119 files, exit 0.
- Negative case: a fixture whose first document is a valid `Trigger` and
whose
  second is `kind: NotATektonThing` reports
`invalid document: apiVersion="v1" kind="NotATektonThing" name="bad"`
and exits 1
  (and the valid document produces no noise).
@wuhuizuo

Copy link
Copy Markdown
Contributor Author

/retest pull-verify-tekton-yaml

@wuhuizuo

Copy link
Copy Markdown
Contributor Author

/retest

@ti-chi-bot
ti-chi-bot Bot merged commit 311fbe4 into main Nov 13, 2026
8 checks passed
@ti-chi-bot
ti-chi-bot Bot deleted the fix/create-event-ref-type branch November 13, 2026 03:43
ti-chi-bot Bot pushed a commit that referenced this pull request Nov 13, 2026
…ry (#5591)

## Why

Since #5474 removed the `ce-paramGithubRepo` allowlist bypass, the
devbuild filters have also required `ce-paramProduct` to equal
`body.repository.name`. That equality is neither a security boundary nor
a routing input:

- the repository allowlist is body-only, keyed on
`body.repository.full_name`; no product header widens it;
- every non-Serverless route derives `component` from the allowlisted
repository (`repository.name`, plus the `tidbcloud/*` and
`pingcap/ticse` overrides), so the header cannot change what is built
either.

What it did do was reject legitimate builds. tibuild labels a component
that lives in a repository with another name, and `githubRepo` may be
passed explicitly:

| product | repository |
| --- | --- |
| `ticdc-newarch` | `pingcap/ticdc` |
| `br`, `dumpling`, `tidb-lightning` | `pingcap/tidb` |
| `drainer`, `pump` | `pingcap/tidb-binlog` |
| `dm` | `pingcap/tiflow` |

A rejected event creates no `PipelineRun` and the listener logs nothing,
so the devbuild stays `PENDING` forever. Five of them are stuck right
now (every devbuild since 2026-11-06 whose product does not equal its
repository name; all 120 that do match have real statuses):

| devbuild | created (UTC) | product | repository | ref | kind |
| --- | --- | --- | --- | --- | --- |
| 3030009 | 2026-11-13 09:35 | `ticdc-newarch` | `pingcap/ticdc` |
`tag/v8.5.4-20261113-5691a52` | hotfix |
| 2970069 | 2026-11-11 08:11 | `br` | `pingcap/tidb` |
`tag/v8.5.4-20261111-4d32ecb` | hotfix |
| 3030012 | 2026-11-13 10:53 | `ticdc` | `pingcap/tiflow` |
`branch/master` | devbuild |
| 2970059 | 2026-11-10 09:55 | `ticdc` | `pingcap/tiflow` | `pull/12954`
| devbuild |
| 2970060 | 2026-11-10 09:56 | `ticdc` | `pingcap/tiflow` | `pull/12954`
| devbuild |

The last `ticdc-newarch` hotfix that actually ran is 2880034
(2026-11-03), i.e. before the merge.

## Change

Drop the product/repository equality from all nine `fake-github-*`
routes (`branch-push`, `pr`, `tag-create` × plain, `-darwin`,
`-single-platform`).

The Serverless routes keep it: there the product *is* the component
(`extensions.build.serverless ? extensions.build.product : ...`), so it
still has to name the Serverless repository.

`tekton/tests/trigger-security.test.ts` now asserts the replacement
invariant — whatever the header says, the route and the component come
from the repository:

```
ACCEPT  pingcap/ticdc    product=ticdc-newarch  component=ticdc
ACCEPT  pingcap/tidb     product=br             component=tidb
ACCEPT  pingcap/tiflow   product=ticdc          component=tiflow
REJECT  attacker/evil    product=tidb           component=-
```

Note the third line: with the product decoupled, `ticdc` on
`pingcap/tiflow` builds the **tiflow** component. Nothing about the
request's product label influences the build, which is the point.

## Security

Unchanged and unaffected:

- the allowlist is still `body.repository.full_name in [...]`, and
`trigger-security.test.ts` still asserts that no route reads
`ce-paramGithubRepo`, keys the allowlist on `repository.name`, or binds
`repository.clone_url`;
- a forged event for a non-allowlisted repository is still rejected with
arbitrary product headers;
- a Serverless product still cannot route a non-Serverless repository
(that branch takes the component from the header, so it keeps the
equality).

The `fake-github-*` triggers are only selected by `el-tibuild`, which
verifies tibuild's CloudEvent HMAC (PingCAP-QE/ee-ops#2619).

## Verification

```
kustomize build tekton/v1/triggers/triggers/env-gcp   -> OK
deno test --frozen --lock=tekton/tests/deno.lock --allow-read --allow-write \
  --allow-env --allow-run=git,sh,deno tekton/tests/*.test.ts
  -> ok | 43 passed (302 steps) | 0 failed
```

## Second commit (please review separately)

`deno test tekton/tests/*.test.ts` is red on `main`, independent of this
change — I reproduced it on a clean `origin/main` worktree (3f2c3d0).
The second commit repairs the TiCSE formal-route assertions:

- #5533 added `pingcap/ticse` to `cseRepositories`, so the tag and
branch-creation assertions ran with a `branch-push` payload; those
triggers require `body.ref_type` (#5569) and threw `field not found:
ref_type`. They now get payloads that carry it.
- #5588 raised `tidbcloud/cloud-storage-engine` to `1h` on the
branch-push route only, while the test asserted a single `30m` for both
routes. It now asserts each route's own value.

If you would rather land that as its own PR, say so and I will split it
out.

## Follow-ups (not in this PR)

- **tibuild (PingCAP-QE/ee-apps)**: a devbuild whose CloudEvents
produced no `PipelineRun` stays `PENDING` forever — 80 such records are
in the database, the oldest from 2026-08-10. A watchdog (`PENDING` +
`triggersEventIDs` non-empty + `pipelines` empty after N minutes →
error) would have turned this incident into a visible failure instead of
a stalled queue.
- **`git-create-tag-build-ng.yaml`**: the tag route still carries the
old `30m` budget for `tidbcloud/cloud-storage-engine` while the
branch-push route has `1h` after #5588. Looks unintended; worth a
one-line follow-up.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

Status: ✅ Done

Development

Successfully merging this pull request may close these issues.

1 participant