-
Couldn't load subscription status.
- Fork 7.3k
gh attestation verify policy enforcement refactor
#9848
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
Signed-off-by: Meredith Lancaster <[email protected]>
Signed-off-by: Meredith Lancaster <[email protected]>
Signed-off-by: Meredith Lancaster <[email protected]>
Signed-off-by: Meredith Lancaster <[email protected]>
Signed-off-by: Meredith Lancaster <[email protected]>
Signed-off-by: Meredith Lancaster <[email protected]>
Signed-off-by: Meredith Lancaster <[email protected]>
Signed-off-by: Meredith Lancaster <[email protected]>
Signed-off-by: Meredith Lancaster <[email protected]>
Signed-off-by: Meredith Lancaster <[email protected]>
Signed-off-by: Meredith Lancaster <[email protected]>
gh attestation verify policy enforcement refactor
Signed-off-by: Meredith Lancaster <[email protected]>
Signed-off-by: Meredith Lancaster <[email protected]>
Signed-off-by: Meredith Lancaster <[email protected]>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I left a ton of comments but I am liking where this is heading!
| } | ||
|
|
||
| type EnforcementCriteria struct { | ||
| Extensions Extensions |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why not literally embed the CertificateSummary? & store the SANRegex separately. That way we get every field by default.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
| Extensions Extensions | |
| Certificate certificate.Summary |
or w/e
| Extensions Extensions | ||
| PredicateType string | ||
| Artifact artifact.DigestedArtifact | ||
| OIDCIssuer string |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
On the topic of embedding the CertificateSummary, since we have to compare the Issuer every time, we can move this inside the Certificate
| return fmt.Errorf("expected Issuer to be %s, got %s -- if you have a custom OIDC issuer policy for your enterprise, use the --cert-oidc-issuer flag with your expected issuer", want, certIssuer) | ||
| } else { | ||
| return fmt.Errorf("expected Issuer to be %s, got %s", want, certIssuer) | ||
| if c.OIDCIssuer != "" { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Since we don't check it elsewhere, c.OIDCIssuer being blank is an error condition 🤔 .
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Like, if it is == "" then we should throw an error kind of thing. What are you verifying?
pkg/cmd/attestation/verify/policy.go
Outdated
| // if tenant is provided, select the appropriate default based on the tenant | ||
| // otherwise, use the provided OIDCIssuer | ||
| if opts.Tenant != "" { | ||
| c.OIDCIssuer = fmt.Sprintf(verification.GitHubTenantOIDCIssuer, opts.Tenant) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is not a clean mapping of the logic. If you have a tenant you may still specify a custom OIDC issuer, contrast with:
if issuer != GitHubOIDCIssuer {
want = issuer
} else {
if tenant != "" {
want = fmt.Sprintf(GitHubTenantOIDCIssuer, tenant)
} else {
want = GitHubOIDCIssuer
}
}
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
(if no tests are failing because of this then we should add one)
pkg/cmd/attestation/verify/policy.go
Outdated
|
|
||
| return verify.SubjectAlternativeNameMatcher{}, nil | ||
| if opts.DenySelfHostedRunner { | ||
| c.Extensions.RunnerEnvironment = GitHubRunner |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
let's preserve the comment re: the empty string matching for AnyRunner
pkg/cmd/attestation/verify/policy.go
Outdated
|
|
||
| func buildSANMatcher(opts *Options) (verify.SubjectAlternativeNameMatcher, error) { | ||
| func newEnforcementCriteria(opts *Options, a artifact.DigestedArtifact) (verification.EnforcementCriteria, error) { | ||
| c := verification.EnforcementCriteria{ |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The Enforcement criteria maps Options to strings we're verifying. What does having artifact in here buy us?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
like, the contents of the opts will never modify the DigestedArtifact, right? so why is it EC's responsibility to be a container for it?
pkg/cmd/attestation/verify/policy.go
Outdated
|
|
||
| func buildVerifyPolicy(opts *Options, a artifact.DigestedArtifact) (verify.PolicyBuilder, error) { | ||
| artifactDigestPolicyOption, err := verification.BuildDigestPolicyOption(a) | ||
| func SigstorePolicy(c verification.EnforcementCriteria) (verify.PolicyBuilder, error) { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
so here I'd say either
NewSigstoreVerifyPolicy or buildSigstoreVerifyPolicy,
and
- i feel like it should take the artifact as a param, so we can liberate the enforcement criteria from having to care about the artifact
pkg/cmd/attestation/verify/verify.go
Outdated
| policy, err := buildVerifyPolicy(opts, *artifact) | ||
| opts.Logger.VerbosePrintf("Verifying attestations with predicate type: %s\n", opts.PredicateType) | ||
|
|
||
| ec, err := newEnforcementCriteria(opts, *artifact) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
imho:
- this should happen before we even load the bundles. The PredicateType in FilterAttestations should be coming from the
ec, not fromopts - like, even before we digest the artifact. nothing in the EC depends on the artifact or a network call, right?
- and then right after we call newEnforcementCriteria we should call
err := Valid()or w/e.
here I am thinking that ec.Validate() is separate from newEC only in so far that it might be easier to check for invalid states separately from setting them the flags.
for example, if the OIDCIssuer/Certificate.Issuer is empty on the Valid() call then that's an error! but it might not be obvious where to stick that check in the new func.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I ended up creating a separate EnforcementCriteria#Valid method for checking that all required fields are set. It felt a little too crowded doing it in the constructor.
Signed-off-by: Meredith Lancaster <[email protected]>
Signed-off-by: Meredith Lancaster <[email protected]>
Co-authored-by: Phill MV <[email protected]>
Signed-off-by: Meredith Lancaster <[email protected]>
Signed-off-by: Meredith Lancaster <[email protected]>
Signed-off-by: Meredith Lancaster <[email protected]>
Signed-off-by: Meredith Lancaster <[email protected]>
Signed-off-by: Meredith Lancaster <[email protected]>
Signed-off-by: Meredith Lancaster <[email protected]>
| func newEnforcementCriteria(opts *Options) (verification.EnforcementCriteria, error) { | ||
| c := verification.EnforcementCriteria{} | ||
|
|
||
| // Set SANRegex using either the opts.SignerRepo or opts.SignerWorkflow values |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This section of setting SAN and/or SANRegex is using logic originally from
cli/pkg/cmd/attestation/verify/policy.go
Lines 28 to 44 in 30066b0
| func buildSANMatcher(opts *Options) (verify.SubjectAlternativeNameMatcher, error) { | |
| if opts.SignerRepo != "" { | |
| signedRepoRegex := expandToGitHubURL(opts.Tenant, opts.SignerRepo) | |
| return verify.NewSANMatcher("", signedRepoRegex) | |
| } else if opts.SignerWorkflow != "" { | |
| validatedWorkflowRegex, err := validateSignerWorkflow(opts) | |
| if err != nil { | |
| return verify.SubjectAlternativeNameMatcher{}, err | |
| } | |
| return verify.NewSANMatcher("", validatedWorkflowRegex) | |
| } else if opts.SAN != "" || opts.SANRegex != "" { | |
| return verify.NewSANMatcher(opts.SAN, opts.SANRegex) | |
| } | |
| return verify.SubjectAlternativeNameMatcher{}, nil | |
| } |
| c.Certificate.RunnerEnvironment = "" | ||
| } | ||
|
|
||
| // If the Repo option is provided, set the SourceRepositoryURI extension |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
A lot of the logic originally in https://github.com/cli/cli/blob/30066b0042d0c5928d959e288144300cb28196c9/pkg/cmd/attestation/verification/extensions.go has been moved to here.
| return fmt.Errorf("expected Issuer to be %s, got %s -- if you have a custom OIDC issuer policy for your enterprise, use the --cert-oidc-issuer flag with your expected issuer", want, certIssuer) | ||
| } else { | ||
| return fmt.Errorf("expected Issuer to be %s, got %s", want, certIssuer) | ||
| if criteria.Certificate.Issuer != "" { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
imho skip the empty check here,: criteria.Certificate.Issuer == "" is always going to be an invalid state.
we now validate this, so it's less risky but, i feel like,
a) this should never generate an error so it's safe to always check, and
b) if this does ever generate an error that'll tell us about a bug somewhere else
🤔
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I lean towards not checking for the empty string because of (b) since it exposes a bug and feels less risky now that we have a validation function for the struct.
Signed-off-by: Meredith Lancaster <[email protected]>
Signed-off-by: Meredith Lancaster <[email protected]>
Signed-off-by: Meredith Lancaster <[email protected]>
Signed-off-by: Meredith Lancaster <[email protected]>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This looks good to me! I think we landed in a nice clean place that centralizes the logic & makes enforcement decisions easier to reason about & expose to the user (in a future PR).
I'm only not approving because I haven't had the time/mental energy rn to read thru the tests 😄.
Co-authored-by: Phill MV <[email protected]>
Signed-off-by: Meredith Lancaster <[email protected]>
Signed-off-by: Meredith Lancaster <[email protected]>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
lgtm
This MR contains the following updates: | Package | Update | Change | |---|---|---| | [cli/cli](https://github.com/cli/cli) | minor | `v2.59.0` -> `v2.62.0` | MR created with the help of [el-capitano/tools/renovate-bot](https://gitlab.com/el-capitano/tools/renovate-bot). **Proposed changes to behavior should be submitted there as MRs.** --- ### Release Notes <details> <summary>cli/cli (cli/cli)</summary> ### [`v2.62.0`](https://github.com/cli/cli/releases/tag/v2.62.0): GitHub CLI 2.62.0 [Compare Source](cli/cli@v2.61.0...v2.62.0) #### What's Changed - Update monotonic verification logic and testing by [@​malancas](https://github.com/malancas) in cli/cli#9856 - Check extension for latest version when executed by [@​andyfeller](https://github.com/andyfeller) in cli/cli#9866 - Shorten extension release checking from 3s to 1s by [@​andyfeller](https://github.com/andyfeller) in cli/cli#9914 - Mention GitHub CLI team on discussion issues by [@​andyfeller](https://github.com/andyfeller) in cli/cli#9920 **Full Changelog**: cli/cli@v2.61.0...v2.62.0 #### Security - A security vulnerability has been identified in GitHub CLI that could allow remote code execution (RCE) when users connect to a malicious Codespace SSH server and use the `gh codespace ssh` or `gh codespace logs` commands. For more information, see GHSA-p2h2-3vg9-4p87 #### GitHub CLI notifies users about latest extension upgrades Similar to the notification of latest `gh` releases, the `v2.62.0` version of GitHub CLI will notify users about latest extension upgrades when the extension is used: ```shell $ gh ado2gh ... A new release of ado2gh is available: 1.7.0 → 1.8.0 To upgrade, run: gh extension upgrade ado2gh --force https://github.com/github/gh-ado2gh ``` ##### Why does this matter? This removes a common pain point of extension authors as they have had to reverse engineer and implement a similar mechanism within their extensions directly. With this quality of life improvement, there are 2 big benefits: 1. Extension authors will hopefully see increased adoption of newer releases while having lower bar to maintaining their extensions. 2. GitHub CLI users will have greater awareness of new features, bug fixes, and security fixes to the extensions used. ##### What do you need to do? Extension authors should review their extensions and consider removing any custom logic previously implemented to notify users of new releases. ### [`v2.61.0`](https://github.com/cli/cli/releases/tag/v2.61.0): GitHub CLI 2.61.0 [Compare Source](cli/cli@v2.60.1...v2.61.0) #### Ensure users understand consequences before making repository visibility changes In `v2.61.0`, `gh repo edit` command has been enhanced to inform users about [consequences of changing visibility](https://gh.io/setting-repository-visibility) and ensure users are intentional before making irreversible changes: 1. Interactive `gh repo edit` visibility change requires confirmation when changing from `public`, `private`, or `internal` 2. Non-interactive `gh repo edit --visibility` change requires new `--accept-visibility-change-consequences` flag to confirm 3. New content to inform users of consequences - Incorporate [GitHub Docs content](https://gh.io/setting-repository-visibility) into help usage and interactive `gh repo edit` experience - Expanded help usage to call out most concerning consequences - Display repository star and watcher counts to understand impact before confirming #### What's Changed - Add acceptance test for `project` command by [@​jtmcg](https://github.com/jtmcg) in cli/cli#9816 - Add comprehensive testscript for `gh ruleset` by [@​andyfeller](https://github.com/andyfeller) in cli/cli#9815 - Add comprehensive testscript for gh ext commandset by [@​andyfeller](https://github.com/andyfeller) in cli/cli#9810 - Require visibility confirmation in `gh repo edit` by [@​andyfeller](https://github.com/andyfeller) in cli/cli#9845 - Clean up skipped online tests for `gh attestation verify` by [@​malancas](https://github.com/malancas) in cli/cli#9838 - `gh attestation verify` should only verify provenance attestations by default by [@​malancas](https://github.com/malancas) in cli/cli#9825 - Set `dnf5` commands as default by [@​its-miroma](https://github.com/its-miroma) in cli/cli#9844 - Fix verbiage for deleting workflow runs by [@​akx](https://github.com/akx) in cli/cli#9876 - Bump github.com/creack/pty from 1.1.23 to 1.1.24 by [@​dependabot](https://github.com/dependabot) in cli/cli#9862 - `gh attestation verify` policy enforcement refactor by [@​malancas](https://github.com/malancas) in cli/cli#9848 - Simplify Sigstore verification result handling in `gh attestation verify` by [@​malancas](https://github.com/malancas) in cli/cli#9877 - Print empty array for `gh cache list` when `--json` is provided by [@​williammartin](https://github.com/williammartin) in cli/cli#9883 - Bump actions/attest-build-provenance from 1.4.3 to 1.4.4 by [@​dependabot](https://github.com/dependabot) in cli/cli#9884 - Create the automatic key when specified with -i by [@​cmbrose](https://github.com/cmbrose) in cli/cli#9881 - fix: `gh pr create -w` ignore template flag by [@​nilvng](https://github.com/nilvng) in cli/cli#9863 #### New Contributors - [@​akx](https://github.com/akx) made their first contribution in cli/cli#9876 - [@​nilvng](https://github.com/nilvng) made their first contribution in cli/cli#9863 **Full Changelog**: cli/cli@v2.60.1...v2.61.0 ### [`v2.60.1`](https://github.com/cli/cli/releases/tag/v2.60.1): GitHub CLI 2.60.1 [Compare Source](cli/cli@v2.60.0...v2.60.1) This is a small patch release to fix installing `gh` via `go install` which was broken with v2.60.0. #### What's Changed - Update testscript to use hard fork by [@​williammartin](https://github.com/williammartin) in cli/cli#9821 **Full Changelog**: cli/cli@v2.60.0...v2.60.1 ### [`v2.60.0`](https://github.com/cli/cli/releases/tag/v2.60.0): GitHub CLI 2.60.0 [Compare Source](cli/cli@v2.59.0...v2.60.0) #### What's Changed - Add ArchivedAt field by [@​tsukasaI](https://github.com/tsukasaI) in cli/cli#9790 - Include startedAt, completedAt in run steps data by [@​andyfeller](https://github.com/andyfeller) in cli/cli#9774 - Adjust environment help for host and tokens by [@​williammartin](https://github.com/williammartin) in cli/cli#9809 - Add handling of empty titles for Issues and MRs by [@​jtmcg](https://github.com/jtmcg) in cli/cli#9701 - `LiveSigstoreVerifier.Verify` should error if no attestations are present by [@​phillmv](https://github.com/phillmv) in cli/cli#9742 - `gh at verify` retries fetching attestations if it receives a 5xx by [@​phillmv](https://github.com/phillmv) in cli/cli#9797 - Prevent local extension installations with invalid names and conflicts with core commands and other extensions by [@​BagToad](https://github.com/BagToad) in cli/cli#9794 - Rewrite a sentence in CONTRIBUTING.md by [@​muzimuzhi](https://github.com/muzimuzhi) in cli/cli#9772 - Use new GitHub preview terms in `working-with-us.md` by [@​BagToad](https://github.com/BagToad) in cli/cli#9800 - Use new GitHub previews terminology in attestation commands' help docs by [@​BagToad](https://github.com/BagToad) in cli/cli#9799 - Clarify in README that `gh` is supported on GitHub Enterprise Cloud by [@​BagToad](https://github.com/BagToad) in cli/cli#9805 - build(deps): bump github.com/gabriel-vasile/mimetype from 1.4.5 to 1.4.6 by [@​dependabot](https://github.com/dependabot) in cli/cli#9752 ##### Acceptance Test Changes - Add acceptance tests for `workflow`, `run`, and `cache` commands by [@​BagToad](https://github.com/BagToad) in cli/cli#9766 - Add basic `api` acceptance tests by [@​BagToad](https://github.com/BagToad) in cli/cli#9770 - Add acceptance tests for `release` commands by [@​BagToad](https://github.com/BagToad) in cli/cli#9771 - Add acceptance tests for `org` and `ssh-key` commands by [@​BagToad](https://github.com/BagToad) in cli/cli#9812 - Add acceptance tests for `gh auth` commands by [@​jtmcg](https://github.com/jtmcg) in cli/cli#9787 - Add acceptance tests for `repo` commands by [@​jtmcg](https://github.com/jtmcg) in cli/cli#9783 - Add acceptance tests for `search` command by [@​BagToad](https://github.com/BagToad) in cli/cli#9786 - Add acceptance tests for `variable` commands by [@​andyfeller](https://github.com/andyfeller) in cli/cli#978 - Add testscripts for gpg-key and label commands by [@​williammartin](https://github.com/williammartin) in cli/cli#9811 - Use forked testscript for token redaction by [@​williammartin](https://github.com/williammartin) in cli/cli#9804 - Add acceptance tests for `secret` commands by [@​andyfeller](https://github.com/andyfeller) in cli/cli#9782 - Note token redaction in Acceptance test README by [@​williammartin](https://github.com/williammartin) in cli/cli#9813 #### New Contributors - [@​tsukasaI](https://github.com/tsukasaI) made their first contribution in cli/cli#9790 **Full Changelog**: cli/cli@v2.59.0...v2.60.0 </details> --- ### Configuration 📅 **Schedule**: Branch creation - At any time (no schedule defined), Automerge - At any time (no schedule defined). 🚦 **Automerge**: Disabled by config. Please merge this manually once you are satisfied. ♻ **Rebasing**: Whenever MR becomes conflicted, or you tick the rebase/retry checkbox. 🔕 **Ignore**: Close this MR and you won't be reminded about this update again. --- - [ ] <!-- rebase-check -->If you want to rebase/retry this MR, check this box --- This MR has been generated by [Renovate Bot](https://github.com/renovatebot/renovate). <!--renovate-debug:eyJjcmVhdGVkSW5WZXIiOiIzNy40NDAuNyIsInVwZGF0ZWRJblZlciI6IjM3LjQ0MC43IiwidGFyZ2V0QnJhbmNoIjoibWFpbiIsImxhYmVscyI6WyJSZW5vdmF0ZSBCb3QiXX0=-->
This is an initial refactor of how
gh attestation verifyhandles policy enforcement. It consolidates the logic used to evaluate policy options into a single struct.cc #9850