Skip to content

Comments

keylimectl: A replacement for keylime_tenant in rust#1068

Draft
ansasaki wants to merge 41 commits intokeylime:masterfrom
ansasaki:keylimectl
Draft

keylimectl: A replacement for keylime_tenant in rust#1068
ansasaki wants to merge 41 commits intokeylime:masterfrom
ansasaki:keylimectl

Conversation

@ansasaki
Copy link
Contributor

@ansasaki ansasaki commented Aug 4, 2025

Disclaimer: this is an AI generated rewrite. We should be careful reviewing it.

Adds a modern Rust replacement for keylime_tenant with full API compatibility and improved usability.

Features

  • Agent Management: add, remove, update, status, reactivate commands
  • Policy Management: runtime and measured boot policy CRUD operations
  • Resource Listing: agents, policies with detailed/basic views
  • Multi-format Output: JSON, table, YAML with configurable verbosity
  • Robust Error Handling: typed errors with context and retry logic
  • TLS Support: mutual authentication with certificate validation
  • Configuration: file-based config with CLI overrides

Implementation

  • 8,512 lines of documented Rust code
  • 158 comprehensive unit tests (100% pass rate)
  • 0 clippy warnings, full type safety
  • Modular architecture with proper abstractions
  • IPv6 support and exponential backoff retry

Usage

keylimectl agent add <uuid> --ip 192.168.1.100 --port 9002
keylimectl policy create web-policy --file policy.json
keylimectl list agents --detailed

Replaces Python keylime_tenant while maintaining backward compatibility.

@ansasaki ansasaki marked this pull request as draft August 4, 2025 12:04
@ansasaki ansasaki force-pushed the keylimectl branch 2 times, most recently from b55ee8e to da44cbc Compare August 4, 2025 15:43
@codecov
Copy link

codecov bot commented Aug 5, 2025

Codecov Report

❌ Patch coverage is 0% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 42.18%. Comparing base (bec5d94) to head (d1ff80a).

Files with missing lines Patch % Lines
keylime-push-model-agent/src/attestation.rs 0.00% 1 Missing ⚠️
Additional details and impacted files
Flag Coverage Δ
e2e-testsuite 42.18% <0.00%> (-16.10%) ⬇️
upstream-unit-tests 42.18% <0.00%> (-16.10%) ⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines Coverage Δ
keylime-push-model-agent/src/struct_filler.rs 0.00% <ø> (-25.44%) ⬇️
keylime-push-model-agent/src/attestation.rs 0.00% <0.00%> (-44.76%) ⬇️

... and 53 files with indirect coverage changes

🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

@ansasaki ansasaki force-pushed the keylimectl branch 3 times, most recently from 7cd11bf to 9609be7 Compare August 7, 2025 10:19
@ansasaki ansasaki mentioned this pull request Aug 27, 2025
30 tasks
@ansasaki ansasaki mentioned this pull request Sep 24, 2025
30 tasks
@ansasaki ansasaki force-pushed the keylimectl branch 6 times, most recently from f15c57b to d1ff80a Compare October 6, 2025 17:19
@sarroutbi sarroutbi mentioned this pull request Nov 26, 2025
36 tasks
@ansasaki ansasaki force-pushed the keylimectl branch 9 times, most recently from 9b7d3d7 to f1bf332 Compare February 23, 2026 16:49
ansasaki and others added 23 commits February 24, 2026 17:49
Add --timeout CLI flag, no-argument behavior with config summary and
dynamic clap help, interactive configuration wizard (dialoguer), config
file path tracking, formalized search paths, and integration tests.

Key changes:
- --timeout global flag overrides client.timeout
- Running without subcommand shows config summary + clap help
- `configure` subcommand with interactive wizard (wizard feature flag)
  and --non-interactive mode for scripted configuration
- Config struct tracks loaded_from path, has_config_file() helper
- .keylimectl/config.toml added as highest-priority auto-discovery path
- 348 tests (340 unit + 8 integration), clippy clean across all feature
  combinations

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add the `keylimectl info` command with subcommands for inspecting
configuration, server status, agents, and TLS certificates. Info
commands work even with incomplete configuration (warn on validation
instead of exiting), making them useful for troubleshooting.

Subcommands:
- `info` (no subcommand): local diagnostics with version, features,
  config file search results, effective config with per-field source
  annotations (cli/env_var/config_file/default), KEYLIME_* env vars
- `info verifier`: connect to verifier, report API version, agent count
- `info registrar`: connect to registrar, report API version, agent count
- `info agent <ID>`: combined verifier+registrar view with summary
- `info tls`: validate cert files, check expiration, verify cert/key pair

Infrastructure:
- CliOverrides struct tracks which CLI args were provided for source
  annotation in the info command
- Config::config_search_paths() exposes file search paths
- RegistrarClient::api_version() public getter added
- Info command gets its own match arm in main() with validation bypass

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…with new CLI subcommands

Convert commands/policy.rs into a directory module (commands/policy/)
with the existing CRUD operations moved to crud.rs. Add new CLI
subcommands for local policy tools:

- policy generate runtime/measured-boot/tpm
- policy sign (DSSE signing)
- policy verify-signature
- policy validate
- policy convert (legacy format conversion)
- verify evidence (top-level command for one-shot attestation)

Add new error types: PolicyGenerationError, DsseError, EvidenceError.
Create policy_tools module skeleton for shared policy logic.

All new subcommands return "not yet implemented" stubs. Existing CRUD
commands and all 372 unit tests continue to pass unchanged.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…ypes

Add Rust types for policy schemas that match the Python implementation,
enabling serialization compatibility between Python and Rust codebases.

- RuntimePolicy: v1 schema with digests, excludes, keyrings, IMA config
- MeasuredBootPolicy: UEFI Secure Boot reference state types
- TpmPolicy: PCR mask and expected values with helper methods

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…tion

Add IMA measurement list parsing, flat-text and JSON allowlist parsing,
exclude list parsing, and file digest calculation for local runtime
policy generation via `keylimectl policy generate runtime`.

- ima_parser: parse IMA logs (ima, ima-ng, ima-sig, ima-buf templates)
- ima_parser: parse flat-text and JSON allowlists, exclude lists
- digest: calculate file digests using OpenSSL (sha1/256/384/512/sm3)
- generate: wire Runtime subcommand to parse inputs and build policy

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add filesystem tree scanning for rootfs digest calculation and policy
merge utilities. Wire --rootfs and --skip-path CLI args to the runtime
policy generate command using tokio::spawn_blocking for CPU-bound work.

- filesystem: recursive directory walk with skip paths and symlink exclusion
- merge: union of digests, excludes, keyrings, and ima-buf entries
- runtime_policy: add deduplication to add_digest/add_keyring/add_ima_buf

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…ation

Add Dead Simple Signing Envelope (DSSE) support for policy signing and
verification with ECDSA P-256 and X.509 certificate backends. Implements
PAE (Pre-Authentication Encoding), envelope sign/verify protocol, key
generation/loading, and self-signed certificate creation.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…oot, and TPM policies

Add structural and content validation for all three policy types with
auto-detection. Validates digest formats, required fields, PCR mask
consistency, and schema compatibility. Supports optional DSSE signature
verification during validation via --signature-key.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add conversion from JSON and flat-text allowlists to v1 runtime policy
format. Supports auto-detection of input format, exclude list merging,
and verification key injection. Adds in-memory parsing helpers for
JSON and flat-text allowlists to ima_parser.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add verify evidence command that posts TPM or TEE attestation evidence
to the verifier's /verify/evidence endpoint. Reads quote, AK, EK files
as base64, sends with nonce and policies, parses verification results
including failure details.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add TPM policy generation from PCR values file with index filtering
and mask calculation. Add measured boot policy generation from UEFI
event logs using the shared crate's UefiLogHandler, extracting S-CRTM,
platform firmware, and Secure Boot variable measurements.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add 23 integration tests in tests/policy_tools.rs covering:
- Help output for all new subcommands (generate, sign, validate, convert)
- Runtime policy generation from IMA logs, allowlists, with excludes
- TPM policy generation from PCR values files
- Policy validation for runtime and TPM policy types
- DSSE signing (ECDSA and X.509 backends) and verification
- Legacy allowlist conversion (flat-text and JSON formats)
- End-to-end pipeline: generate -> validate -> sign -> verify

Fix bugs found during integration testing:
- Remove duplicate stdout output (commands called output.success()
  AND main.rs dispatcher also called it, producing double JSON)
- Remove default_value on --ima-measurement-list to make it truly
  optional (previously always read /sys/kernel/security/ima even
  when only --allowlist was specified)
- Add PolicyAction::is_local_only() and match arm in main() so
  local-only policy commands (generate, sign, verify-signature,
  validate, convert) bypass strict TLS config validation

Update ROADMAP.md to mark Phase 6 items as complete.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add policy_tools/privilege module with helpers for detecting permission
errors and suggesting sudo retries. Add PrivilegeRequired error variant
to PolicyGenerationError for privileged operations like TPM access,
initramfs reading, and boot event log parsing.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
… measured boot event log

Add EV_EFI_VARIABLE_AUTHORITY and EV_EFI_PLATFORM_FIRMWARE_BLOB2 event
types to the shared UEFI log handler. Create uefi_event_data module for
parsing UEFI_VARIABLE_DATA and EV_IPL event data structures. Extract
boot chain entries (shim/grub/kernel from PCR 4), kernel command line
(PCR 8), initrd/vmlinuz digests (PCR 9), MOK digests (MokList/MokListX),
and vendor_db from EV_EFI_VARIABLE_AUTHORITY events. Add
vmlinuz_plain_sha256 field to KernelEntry for non-SecureBoot systems.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…ature flag

Add tpm-local feature flag (aliases dep:tss-esapi). Implement
generate_from_tpm() that opens the local TPM via TCTI, reads PCR
values for requested indices and hash algorithm, and builds a TpmPolicy.
On permission errors accessing /dev/tpmrm0, suggest running with sudo.
Pass hash_alg through from CLI to TPM generation. Without the feature
flag, --from-tpm produces a clear error naming the required feature.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add ability to extract and hash files from initramfs/initrd images for
runtime policy generation. Supports gzip, zstd, xz, and bzip2
compression formats with CPIO new-ASCII archive parsing.

Key components:
- Compression detection via magic bytes with automatic decompression
- Early microcode CPIO archive detection and skipping
- In-memory CPIO parsing that hashes files without extracting to disk
- Privilege detection for /boot directory access with sudo suggestion

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…eature flag

Add --local-rpm-repo and --remote-rpm-repo options for runtime policy
generation. Local repos are scanned for RPM files and their headers
are parsed for file digests. Remote repos use filelists-ext.xml as a
fast path, falling back to downloading individual RPMs.

Key components:
- RPM header parsing via pure-Rust rpm crate (no librpm-devel needed)
- repomd.xml and filelists-ext.xml parsing via quick-xml
- Automatic decompression of metadata files (gzip, xz, zstd, bzip2)
- Feature-gated: rebuild with --features rpm-repo to enable
- Replace xz2 with liblzma to avoid native library linking conflict

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
… update ROADMAP

Add 5 new integration tests for Phase 6b features:
- Runtime help shows --ramdisk-dir, --local-rpm-repo, --remote-rpm-repo
- TPM help shows --from-tpm
- Nonexistent ramdisk dir fails with error
- Empty ramdisk dir succeeds with no initrd digests

Update ROADMAP.md to mark all deferred Phase 6 items as complete:
initramfs extraction, RPM repository analysis, measured boot kernel/
bootloader/MOK extraction, and TPM PCR reading.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Simplify the CLI flag name for querying the registrar directly.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Signed-off-by: Anderson Toshiyuki Sasaki <ansasaki@redhat.com>
…ion polling

The previous implementation matched operational_state against string
values, but the verifier returns it as an integer. Instead, use the
attestation_status field ("PENDING", "PASS", "FAIL") which the verifier
computes for both push and pull mode agents.

On failure, report the operational state, severity level, and last
event ID so the user understands why attestation failed. On timeout,
use the correct error type to avoid the misleading "Failed to list
verifier" message.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…ment

The accept_tpm_signing_algs field was set to ["rsa", "ecdsa"], which
are encryption algorithm names, not signing algorithm names. The
verifier rejected quotes signed with rsassa because it was not in
the accepted list. Use the correct signing algorithm names matching
the Python tenant defaults: ["ecschnorr", "rsassa"].

Also align accept_tpm_hash_algs with tenant defaults by including
sha512 and sha384, and dropping sha1.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…gent

The agent expects the payload as base64-encoded AES-256-GCM ciphertext
(base64(iv || ciphertext || tag), encrypted with key K). Previously,
the raw file content was sent without encryption or encoding, causing
"Invalid base64 encoding in payload" errors.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…d retry

Two issues caused --verify to always fail:

1. HMAC encoding mismatch: the agent returns HMAC as hex (matching
   Python's do_hmac hexdigest), but we compared against base64-encoded
   HMAC. Changed to hex::encode.

2. No retry: the agent needs V from the verifier before it can compute
   K = U XOR V. The Python tenant retries in a loop; we did a single
   attempt. Added exponential backoff retry (up to 12 attempts).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add animated progress spinners using indicatif for long-running
operations (attestation polling, key derivation retry) and optional
color output via console. Spinners auto-detect TTY on stderr and fall
back to plain text when piped. Colors apply to stderr only, keeping
stdout clean for machine consumption.

New --color flag (auto|always|never) controls color output. The
OutputHandler now supports start_wait() which returns an RAII
WaitHandle for polling loops with live status updates.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add the agent's self-signed mTLS certificate (from the registrar
database) as a trusted root CA when connecting to agents in pull mode.
This allows verifying the agent's TLS certificate without disabling
certificate verification globally, matching the Python tenant behavior.

Also change accept_invalid_hostnames default from true to false, since
certificates should have proper SANs set.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Signed-off-by: Anderson Toshiyuki Sasaki <ansasaki@redhat.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants