feat: add TimeSlave process and libTSClient shm IPC channel by gordon9901 · Pull Request #8 · eclipse-score/inc_time

gordon9901 · 2026-03-27T03:18:31Z

feat: introduce TimeSlave process and libTSClient IPC channel

Summary

This PR splits the gPTP protocol stack out of TimeDaemon into a new standalone process — TimeSlave — and introduces a lightweight shared-memory IPC library (libTSClient) so that TimeDaemon can consume PtpTimeInfo snapshots from TimeSlave without running the raw-socket protocol itself.

Changes

score/TimeSlave/ (new)
A self-contained gPTP slave process (score::ts::TimeSlave) that:

Runs GptpEngine (existing RxThread + PdelayThread logic) to participate in IEEE 802.1AS protocol.
Includes platform implementations for both Linux and QNX (raw_socket, network_identity, phc_adjuster).
Exposes instrumentation (probe) and recording (recorder) capabilities.
Publishes GptpIpcData snapshots to shared memory via GptpIpcPublisher.

score/libTSClient/ (new)
A process-boundary IPC library providing:

GptpIpcData / GptpIpcStatus / GptpIpcSyncFupData / GptpIpcPDelayData
(gptp_ipc_data.h, namespace score::ts) — IPC-layer data types conveying PTP time, status, and measurement fields.
GptpIpcRegion
(gptp_ipc_channel.h) — 64-byte cache-line-aligned shared memory layout with a seqlock (seq / seq_confirm) for lock-free single-writer / multi-reader access.
Default channel name: kGptpIpcName = "/gptp_ptp_info" (namespace score::ts::details).
GptpIpcPublisher — creates and writes to the POSIX shared memory segment (used by TimeSlave).
GptpIpcReceiver — opens and reads from the segment with seqlock retry (used by TimeDaemon via GPTPRealMachine).
gptp_ipc.h — convenience header aggregating all of the above.

Both publisher and receiver reside in namespace score::ts::details.

score/TimeDaemon/code/ptp_machine/real/ (new)

details::ShmPTPEngine
(real/details/shm_ptp_engine.h, namespace score::td::details) — a PTPEngine-compatible adapter backed by GptpIpcReceiver; converts GptpIpcData (libTSClient) to PtpTimeInfo (TimeDaemon).
GPTPRealMachine — type alias for PTPMachine<details::ShmPTPEngine>, slotting cleanly into the existing TimeDaemon pipeline.
CreateGPTPRealMachine(name, ipc_name) — factory function for convenient construction; ipc_name defaults to kGptpIpcName.

score/TimeDaemon/code/common/data_types/BUILD, score/TimeDaemon/code/data_types/BUILD (modified)
Widened Bazel visibility from //score/TimeDaemon:__subpackages__ to //score:__subpackages__ so that TimeSlave and libTSClient can reuse shared data types (PtpTimeInfo, etc.) without duplication.

score/TimeDaemon/code/ptp_machine/BUILD (modified)
Added //score/TimeDaemon/code/ptp_machine/real:unit_test_suite to the aggregated test target.

Architecture

TimeSlave (new process)
  └── GptpEngine (RxThread + PdelayThread)
        └── GptpIpcPublisher  ──[POSIX shm / seqlock]──►  GptpIpcReceiver
                                                                 └── details::ShmPTPEngine
                                                                       └── GPTPRealMachine
                                                                             └── TimeDaemon pipeline

Ticket:eclipse-score/score#2691

github-actions · 2026-03-27T03:19:15Z

License Check Results

🚀 The license check job ran with the Bazel command:

bazel run //:license-check

Status: ⚠️ Needs Review

Click to expand output

[License Check Output]
2026/04/28 07:25:41 Downloading https://releases.bazel.build/8.3.0/release/bazel-8.3.0-linux-x86_64...
Extracting Bazel installation...
Starting local Bazel server (8.3.0) and connecting to it...
INFO: Invocation ID: 8bf3ab65-33fd-4535-a8f3-15d3c9a25d29
Computing main repo mapping: 
Computing main repo mapping: 
Computing main repo mapping: 
Computing main repo mapping: 
Computing main repo mapping: 
Computing main repo mapping: 
Computing main repo mapping: 
Computing main repo mapping: 
WARNING: For repository 'aspect_rules_lint', the root module requires module version aspect_rules_lint@1.5.3, but got aspect_rules_lint@2.0.0 in the resolved dependency graph. Please update the version in your MODULE.bazel or set --check_direct_dependencies=off
Computing main repo mapping: 
Computing main repo mapping: 
Computing main repo mapping: 
Loading: 
Loading: 2 packages loaded
Loading: 2 packages loaded
    currently loading: 
Loading: 2 packages loaded
    currently loading: 
Loading: 2 packages loaded
    currently loading: 
Loading: 2 packages loaded
    currently loading: 
Analyzing: target //:license-check (3 packages loaded, 0 targets configured)
Analyzing: target //:license-check (3 packages loaded, 0 targets configured)

Analyzing: target //:license-check (52 packages loaded, 9 targets configured)

Analyzing: target //:license-check (86 packages loaded, 9 targets configured)

Analyzing: target //:license-check (88 packages loaded, 9 targets configured)

Analyzing: target //:license-check (104 packages loaded, 62 targets configured)

Analyzing: target //:license-check (150 packages loaded, 2657 targets configured)

Analyzing: target //:license-check (158 packages loaded, 6558 targets configured)

Analyzing: target //:license-check (169 packages loaded, 8047 targets configured)

Analyzing: target //:license-check (169 packages loaded, 8049 targets configured)

Analyzing: target //:license-check (169 packages loaded, 8049 targets configured)

Analyzing: target //:license-check (169 packages loaded, 8049 targets configured)

Analyzing: target //:license-check (173 packages loaded, 10058 targets configured)

INFO: Analyzed target //:license-check (174 packages loaded, 10184 targets configured).
[7 / 16] Creating runfiles tree bazel-out/k8-opt-exec-ST-d57f47055a04/bin/external/score_tooling+/dash/tool/formatters/dash_format_converter.runfiles [for tool]; 0s local
[9 / 16] checking cached actions
INFO: From Generating Dash formatted dependency file ...:
INFO: Successfully converted 2 packages from Cargo.lock to bazel-out/k8-fastbuild/bin/formatted.txt
[13 / 16] JavaToolchainCompileClasses external/rules_java+/toolchains/platformclasspath_classes; 0s disk-cache, processwrapper-sandbox
[15 / 16] [Prepa] Building license.check.license_check.jar ()
INFO: Found 1 target...
Target //:license.check.license_check up-to-date:
  bazel-bin/license.check.license_check
  bazel-bin/license.check.license_check.jar
INFO: Elapsed time: 37.455s, Critical Path: 2.38s
INFO: 16 processes: 12 internal, 3 processwrapper-sandbox, 1 worker.
INFO: Build completed successfully, 16 total actions
INFO: Running command line: bazel-bin/license.check.license_check ./formatted.txt <args omitted>
usage: org.eclipse.dash.licenses.cli.Main [-batch <int>] [-cd <url>]
       [-confidence <int>] [-ef <url>] [-excludeSources <sources>] [-help] [-lic
       <url>] [-project <shortname>] [-repo <url>] [-review] [-summary <file>]
       [-timeout <seconds>] [-token <token>]

github-actions · 2026-04-01T06:47:25Z

The created documentation from the pull request is available at: docu-html

BjoernAtBosch

Some comments for now

BjoernAtBosch · 2026-04-24T09:30:14Z

+    const TmvT t3c = TmvT{t3.ns + c1.ns + c2.ns};
+    // t4 = BPF_T_BINTIME (PHC) receive timestamp of Pdelay_Resp (main BPF fd)
+    const TmvT t4 = resp_.recvHardwareTS;
+
+    const std::int64_t delay = ((t2.ns - t1.ns) + (t4.ns - t3c.ns)) / 2LL;
+
+    if (delay < 0)
+        return;
+
+    PDelayResult r{};
+    r.path_delay_ns = delay;
+    r.valid = true;
+


Shouldn't we move all kind of evaluation to the TimeDaemon?
That component will be later ASIL certifiable and should do all kind of timestamp validations ...
TimeSlave cannot be certified (i.e. is QM only). It could induce errors due to interference.

Agreed. GptpIpcPDelayData now carries all raw timestamps (t1/t2/t3/t4), so TimeDaemon can recompute and validate independently.

The current checks in TimeSlave are kept as QM best-effort guards to avoid writing invalid data to shared memory. I suggest discussing the ASIL-B validation migration to TimeDaemon in a meeting first; once agreed, we track it as a separate issue, outside this PR. Does that work for you?

Cool. Yes, valid remark. Accepted

BjoernAtBosch · 2026-04-24T09:37:05Z

+     * @param buf      Buffer large enough to hold existing payload plus header.
+     * @param buf_len  In/out: payload length → frame length after prepend.
+     * @param src_mac  Source MAC address (should be the port's own MAC).
+     * @return true on success, false if the buffer would overflow.


Param buf_capacity is not described

Fixed. Added @param buf_capacity description in the Doxygen comments of frame_codec.h.

BjoernAtBosch · 2026-04-24T16:24:50Z

+    std::lock_guard<std::mutex> lk(mutex_);
+    resp_ = msg;
+}
+
+void PeerDelayMeasurer::OnResponseFollowUp(const PTPMessage& msg)
+{
+
+    std::lock_guard<std::mutex> lk(mutex_);
+    resp_fup_ = msg;
+    ComputeAndStoreUnlocked();


In those two functions the received messages are just set without checking if the contained sequence id is the expected one. Can we rely on the assumption, that messages of e.g. an earlier request could be received after the next request?

Fixed. Added early filtering in OnResponse() and OnResponseFollowUp(): messages with non-matching sequence IDs (vs req_.ptpHdr.sequenceId) are ignored and do not update state. The existing 3-way seq ID consistency check in ComputeAndStoreUnlocked() is kept as a second line of defense.

BjoernAtBosch · 2026-04-24T16:26:12Z

+
+int PeerDelayMeasurer::SendRequest(IRawSocket& socket)
+{
+    PTPMessage req{};


I guess we don't have recognition, if either no responses or multiple responses to a single request are received (to detect non-time-aware switches)

Agreed. This is now handled as follows: a resp_count_ counter has been introduced and is reset to 0 together with req_ under lock in SendRequest(). OnResponse() increments it only for matching sequence IDs. If resp_count_ > 1, ComputeAndStoreUnlocked() returns early without producing a result, covering the IEEE 802.1AS non-time-aware switch detection case. The no-response case is implicitly handled by the existing sequence ID consistency check (no computation triggered, last valid result retained).

Yes, I think that's fine for now. I guess later we will need a more detailed concept how to deal with/behave in that kind of scenarios.

BjoernAtBosch

LGTM

BjoernAtBosch · 2026-04-28T16:38:13Z

+
+int PeerDelayMeasurer::SendRequest(IRawSocket& socket)
+{
+    PTPMessage req{};


Yes, I think that's fine for now. I guess later we will need a more detailed concept how to deal with/behave in that kind of scenarios.

BjoernAtBosch · 2026-04-28T16:39:49Z

+    const TmvT t3c = TmvT{t3.ns + c1.ns + c2.ns};
+    // t4 = BPF_T_BINTIME (PHC) receive timestamp of Pdelay_Resp (main BPF fd)
+    const TmvT t4 = resp_.recvHardwareTS;
+
+    const std::int64_t delay = ((t2.ns - t1.ns) + (t4.ns - t3c.ns)) / 2LL;
+
+    if (delay < 0)
+        return;
+
+    PDelayResult r{};
+    r.path_delay_ns = delay;
+    r.valid = true;
+


Cool. Yes, valid remark. Accepted

feat: add TimeSlave process and libTSClient shm IPC channel

e1ece4a

gordon9901 had a problem deploying to workflow-approval March 27, 2026 03:18 — with GitHub Actions Failure

Update the doc folder, solve the check problem

8729c59

gordon9901 had a problem deploying to workflow-approval March 27, 2026 05:56 — with GitHub Actions Failure

fix bazel build failed

85ea119

gordon9901 had a problem deploying to workflow-approval March 27, 2026 06:28 — with GitHub Actions Failure

Add docs content

56648fb

gordon9901 had a problem deploying to workflow-approval March 27, 2026 06:50 — with GitHub Actions Failure

gordon9901 requested a review from lavrovvalera March 27, 2026 07:05

code review issue resolved

81e30fe

gordon9901 had a problem deploying to workflow-approval April 1, 2026 05:29 — with GitHub Actions Failure

[ECARX][TimeSlave]Add document content

c1de3cd

gordon9901 had a problem deploying to workflow-approval April 1, 2026 06:41 — with GitHub Actions Failure

lavrovvalera reviewed Apr 7, 2026

View reviewed changes

Incorporate review feedback

648de4f

gordon9901 requested a deployment to workflow-approval April 10, 2026 06:29 — with GitHub Actions Waiting

[ECARX][TimeSlave]fix bazel build failed

40ede65

gordon9901 requested a deployment to workflow-approval April 10, 2026 07:01 — with GitHub Actions Waiting

[ECARX][TimeSlave]Fix the PHC clock adjust issue

77f241c

gordon9901 requested a deployment to workflow-approval April 20, 2026 09:08 — with GitHub Actions Waiting

lavrovvalera reviewed Apr 20, 2026

View reviewed changes

[ECARX][TimeSlave]Revise the comments raised by the reviewer

8c014e9

gordon9901 requested a deployment to workflow-approval April 21, 2026 06:21 — with GitHub Actions Waiting

[ECARX][TimeSlave]Delete the local test demo

201059d

gordon9901 requested a deployment to workflow-approval April 21, 2026 06:30 — with GitHub Actions Waiting

gordon9901 requested a review from lavrovvalera April 21, 2026 09:47

lavrovvalera approved these changes Apr 24, 2026

View reviewed changes

BjoernAtBosch reviewed Apr 24, 2026

View reviewed changes

gordon9901 requested a deployment to workflow-approval April 28, 2026 05:59 — with GitHub Actions Waiting

gordon9901 force-pushed the ecarx_gptp branch from 517ff25 to 392fd30 Compare April 28, 2026 06:08

gordon9901 requested a deployment to workflow-approval April 28, 2026 06:08 — with GitHub Actions Waiting

gordon9901 force-pushed the ecarx_gptp branch from 392fd30 to 6fdd308 Compare April 28, 2026 06:53

gordon9901 requested a deployment to workflow-approval April 28, 2026 06:53 — with GitHub Actions Waiting

[ECARX][TimeSlave] Reject stale/duplicate PDelayResp messages

5c09177

gordon9901 force-pushed the ecarx_gptp branch from 6fdd308 to 5c09177 Compare April 28, 2026 07:25

gordon9901 requested a deployment to workflow-approval April 28, 2026 07:25 — with GitHub Actions Waiting

BjoernAtBosch approved these changes Apr 28, 2026

View reviewed changes

Conversation

gordon9901 commented Mar 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

feat: introduce TimeSlave process and libTSClient IPC channel

Summary

Changes

Architecture

Uh oh!

github-actions Bot commented Mar 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

License Check Results

Uh oh!

github-actions Bot commented Apr 1, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

BjoernAtBosch left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

BjoernAtBosch left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

gordon9901 commented Mar 27, 2026 •

edited

Loading

github-actions Bot commented Mar 27, 2026 •

edited

Loading