| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
A series of correctness and API hygiene fixes to the intrusive refcount primitives in `zenbase`, culminating in the removal of `RefPtr<T>` in favour of a single unified `Ref<T>` smart pointer.
The changes are motivated by two pieces of latent UB sitting under every `Ref<T>` / `TRefCounted<T>` in the codebase, plus a handful of API footguns on the smart-pointer side (silent raw-pointer decay, missing converting moves, unconstrained conversions from unrelated types).
## Correctness fixes
- **Strict-aliasing UB in atomic helpers** — `AtomicIncrement`/`Decrement`/`Add` took a `volatile uint32_t&` and reinterpret-cast it to `std::atomic<T>*`. The object was never constructed as a `std::atomic`, so the access was type-punning UB. Fixed by changing `m_RefCount` to `std::atomic<uint32_t>` directly in `RefCounted`, `TRefCounted<T>` and `IoBufferCore`. The helpers (and `zenbase/atomic.h`) are later removed entirely — the three callers now invoke `fetch_add`/`fetch_sub` directly.
- **const_cast of non-mutable member** — `AddRef()` / `Release()` are `const` but mutated `m_RefCount` via `const_cast`. Since `m_RefCount` wasn't `mutable`, writing through the cast was UB for any `const`-qualified holder (e.g. a `static const` refcounted singleton). Fixed by marking `m_RefCount` `mutable` and dropping the `const_cast` in `AddRef`/`Release`.
- **Public non-virtual `TRefCounted` destructor** — allowed `delete basePtr;` to slice past the CRTP `DeleteThis()` contract. Moved to `protected`.
## Memory-ordering cleanup
- `AddRef` weakened from seq_cst to **relaxed** (a thread can only take a new reference via one it already holds; nothing needs to synchronize).
- `Release` weakened from seq_cst to **acq_rel** (sufficient to order prior writes before the destructor, and make the decrement visible to observers).
- Diagnostic `RefCount()` / `GetRefCount()` reads made **relaxed** and spelled out as explicit `.load()` — the returned value is stale the moment it's observed, so stronger ordering gives no guarantee.
- No-op on x86 (`lock xadd` either way), but removes a full barrier on every `Ref<T>` copy on ARM64 (Apple silicon / Windows-on-ARM).
## `RefPtr` / `Ref` unification
Before this branch, `RefPtr<T>` and `Ref<T>` were subtly different in ways that made the safer of the two (`Ref`) harder to use and the looser one (`RefPtr`) dangerous:
- `RefPtr::operator T*()` was implicit — `delete refPtr;` compiled silently (double-delete), and the raw pointer could outlive the temporary `RefPtr` it was extracted from. Made `explicit`, then removed entirely once call sites were migrated to `.Get()`.
- `RefPtr(T*)` was implicit while `RefPtr(RefPtr<Derived>&&)` was `explicit` — exactly the opposite of the safety intent. Reversed.
- `RefPtr`'s converting move was unconstrained (any `RefPtr<U>` with an implicitly-convertible `U*` satisfied it, including `void*` and multiple-inheritance base offsets). Added a `DerivedFrom<U, T>` constraint matching `Ref<T>`.
- `Ref<T>` was missing a converting move ctor / move-assignment from `Ref<Derived>` — upcasts of rvalues were going through `AddRef`+`Release` instead of a pointer steal. Added.
- `Release()` and the non-move smart-pointer ops were not `noexcept`, despite being so in practice. Marked `noexcept` throughout.
After all of the above, the two types were functionally identical. The final commit deletes `RefPtr` and rewrites the ~10 consumer files to use `Ref`.
|
| |
|
|
| |
* move session service to zenserver base class and make it available in all zenserver modes
* fix deadlock in sessionsclient shutdown
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
### Security: Input validation & path safety
- **Reject local file references by default** in package parsing — only allow when explicitly opted in by the service (`ParseFlags::kAllowLocalReferences`) and validated by an `ILocalRefPolicy` (fail-closed: no policy = rejected)
- **`DataRootLocalRefPolicy`** restricts local ref paths to the server's data root via canonical path prefix matching
- **Validate attachment hashes** in compute HTTP handlers — decompresses and re-hashes each attachment at ingestion time to reject tampered payloads
- **Path traversal validation** for worker descriptions (`pathvalidation.h`) — rejects absolute paths, `..` components, Windows reserved device names, and invalid filename characters
- **Harden CbPackage parsing** against corrupt inputs — overflow-safe attachment count, bounds checks on local ref offset/size, graceful failure instead of `ZEN_ASSERT` for untrusted data
- **Harden legacy package parser** — reject zero-size binary fields, missing mappers, and optionally validate resolved attachment hashes
- **Bounds check in `CbPackageReader::MarshalLocalChunkReference`** — detect when `MakeFromFile` silently clamps offset+size to file size
### Reliability: Lock consolidation & bug fixes
- **Consolidate three action map locks into one** (`m_ActionMapLock`) — eliminates deadlock risk from multi-lock ordering, simplifies state transitions, and fixes a race where newly enqueued actions were briefly invisible to `GetActionResult`/`FindActionResult`
- **Fix infinite loop in `BaseRunnerGroup::SubmitActions`** when actions exceed total runner capacity — cap round-robin at `TotalCapacity` and default unassigned results to "No capacity"
- **Fix `MakeSafeAbsolutePathInPlace` for UNC paths** — `\server\share` now correctly becomes `\?\UNC\server\share` instead of `\?\server\share`
- **Fix `max_retries=0`** — previously fell through to the default of 3; now correctly means "no retries"
### New: ManagedProcessRunner
- Cross-platform process runner backed by `SubprocessManager` — uses async exit callbacks instead of polling, delegates CPU/memory metrics to the manager's built-in sampler
- `ProcessGroup` (JobObject on Windows, process group on POSIX) for bulk cancellation on shutdown
- `--managed` flag on `zen exec inproc` to select this runner
- Refactored monitor thread lifecycle — `StartMonitorThread()` now called from derived constructors to avoid calling virtual functions from base constructor
### Process management
- **Suppress crash dialogs** via `JOB_OBJECT_UILIMIT_ERRORMODE` + `SEM_NOGPFAULTERRORBOX` in both `WindowsProcessRunner` and `JobObject::Initialize` — prevents WER/Dr. Watson modal dialogs from blocking the monitor thread
- **CREATE_SUSPENDED → AssignProcessToJobObject → ResumeThread** pattern in `WindowsProcessRunner` — ensures job object assignment before process execution
- **Move stdout/stderr callbacks to `Spawn()` parameters** in `SubprocessManager` — prevents race where early output could be missed before callback installation
- Consistent PID logging across all runner types
### Test infrastructure
- **`zentest-appstub`**: Added `Fail` (configurable exit code) and `Crash` (abort / nullptr deref) test functions
- **Compute integration tests**: exit code handling, auto-retry exhaustion, manual reschedule after failure, mixed success/failure queues, crash handling (abort + nullptr), crash auto-retry, immediate query visibility after enqueue
- **Package format tests**: truncated header, bad magic, attachment count overflow, truncated data, local ref rejection/acceptance, policy enforcement (inside/outside root, traversal, no-policy fail-closed)
- **Legacy package parser tests**: empty input, zero-size binary, hash resolution with/without mapper, hash mismatch detection
- **UNC path tests** for `MakeSafeAbsolutePath`
### Misc
- ANSI color helper macros (`ZEN_RED`, `ZEN_BRIGHT_WHITE`, etc.) and `ZEN_BOLD`/`ZEN_DIM`/etc.
- Generic `fmt::formatter` for types with free `ToString` functions
- Compute dashboard: truncated hash display with monospace font and hover for full value
- Renamed `usonpackage_forcelink` → `cbpackage_forcelink`
- Compute enabled by default in xmake config (releases still explicitly disable)
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
## Summary
This PR adds a session management service, several new dashboard pages, and a number of infrastructure improvements.
### Sessions Service
- `SessionsServiceClient` in `zenutil` announces sessions to a remote zenserver with a 15s heartbeat (POST/PUT/DELETE lifecycle)
- Storage server registers itself with its own local sessions service on startup
- Session mode attribute coupled to server mode (Compute, Proxy, Hub, etc.)
- Ended sessions tracked with `ended_at` timestamp; status filtering (Active/Ended/All)
- `--sessions-url` config option for remote session announcement
- In-process log sink (`InProcSessionLogSink`) forwards server log output to the server's own session, visible in the dashboard
### Session Log Viewer
- POST/GET endpoints for session logs (`/sessions/{id}/log`) supporting raw text and structured JSON/CbObject with batch `entries` array
- In-memory log storage per session (capped at 10k entries) with cursor-based pagination for efficient incremental fetching
- Log panel in the sessions dashboard with incremental DOM updates, auto-scroll (Follow toggle), newest-first toggle, text filter, and log-level coloring
- Auto-selects the server's own session on page load
### TCP Log Streaming
- `LogStreamListener` and `TcpLogStreamSink` for log delivery over TCP
- Sequence numbers on each message with drop detection and synthetic "dropped" notice on gaps
- Gathered buffer writes to reduce syscall overhead when flushing batches
- Tests covering basic delivery, multi-line splitting, drop detection, and sequencing
### New Dashboard Pages
- **Sessions**: master-detail layout with selectable rows, metadata panel, live WebSocket updates, paging, abbreviated date formatting, and "this" pill for the local session
- **Object Store**: summary stats tiles and bucket table with click-to-expand inline object listing (`GET /obj/`)
- **Storage**: per-volume disk usage breakdown (`GET /admin/storage`), Garbage Collection status section (next-run countdown, last-run stats), and GC History table with paginated rows and expandable detail panels
- **Network**: overview tiles, per-service request table, proxy connections, and live WebSocket updates; distinct client IPs and session counts via HyperLogLog
### Documentation Page
- In-dashboard Docs page with sidebar navigation, markdown rendering (via `marked`), Mermaid diagram support (theme-aware), collapsible sections, text filtering with highlighting, and cross-document linking
- New user-facing docs: `overview.md` (with architecture and per-mode diagrams), `sessions.md`, `cache.md`, `projects.md`; updated `compute.md`
- Dev docs moved to `docs/dev/`
### Infrastructure & Bug Fixes
- **Deflate compression** for the embedded frontend zip (~3.4MB → ~950KB); zlib inflate support added to `ZipFs` with cached decompressed buffers
- **Local IP addresses**: `GetLocalIpAddresses()` (Windows via `GetAdaptersAddresses`, Linux/Mac via `getifaddrs`); surfaced in `/status/status`, `/health/info`, and the dashboard banner
- **Dashboard nav**: unified into `zen-nav` web component with `MutationObserver` for dynamically added links, CSS `::part()` to merge banner/nav border radii, and prefix-based active link detection
- Stats broadcast refactored from manual JSON string concatenation to `CbObjectWriter`; `CbObject`-to-JS conversion improved for `TimeSpan`, `DateTime`, and large integers
- Stats WebSocket boilerplate consolidated into `ZenPage.connect_stats_ws()`
|
| |
|
|
|
|
|
|
|
|
| |
- **Frontend dashboard overhaul**: Unified compute/main dashboards into a single shared UI. Added new pages for cache, projects, metrics, sessions, info (build/runtime config, system stats). Added live-update via WebSockets with pause control, sortable detail tables, themed styling. Refactored compute/hub/orchestrator pages into modular JS.
- **HTTP server fixes and stats**: Fixed http.sys local-only fallback when default port is in use, implemented root endpoint redirect for http.sys, fixed Linux/Mac port reuse. Added /stats endpoint exposing HTTP server metrics (bytes transferred, request rates). Added WebSocket stats tracking.
- **OTEL/diagnostics hardening**: Improved OTLP HTTP exporter with better error handling and resilience. Extended diagnostics services configuration.
- **Session management**: Added new sessions service with HTTP endpoints for registering, updating, querying, and removing sessions. Includes session log file support. This is still WIP.
- **CLI subcommand support**: Added support for commands with subcommands in the zen CLI tool, with improved command dispatch.
- **Misc**: Exposed CPU usage/hostname to frontend, fixed JS compact binary float32/float64 decoding, limited projects displayed on front page to 25 sorted by last access, added vscode:// link support.
Also contains some fixes from TSAN analysis.
|
| |
|
|
|
|
|
|
|
|
| |
- Added local process runners for Linux/Wine, Mac with some sandboxing support
- Horde & Nomad provisioning for development and testing
- Client session queues with lifecycle management (active/draining/cancelled), automatic retry with configurable limits, and manual reschedule API
- Improved web UI for orchestrator, compute, and hub dashboards with WebSocket push updates
- Some security hardening
- Improved scalability and `zen exec` command
Still experimental - compute support is disabled by default
|
| |
|
|
|
|
|
|
|
| |
this change adds the `zencompute` component, which can be used to distribute work dispatched from UE using the DDB (Derived Data Build) APIs via zenserver
this change also adds a distinct zenserver compute mode (`zenserver compute`) which is intended to be used for leaf compute nodes
to exercise the compute functionality without directly involving UE, a `zen exec` subcommand is also added, which can be used to feed replays through the system
all new functionality is considered *experimental* and disabled by default at this time, behind the `zencompute` option in xmake config
|
| |
|
|
|
|
|
| |
* add system metrics output to top command
* removed unnecessary xmake directives
* file system API/comment tweaks
* fixed out-of-range access in httpserver test
* updated ZenServer base API to allow customization by mode
|
| |
|
| |
this breaks out the configuration logic to allow multiple applications to share common configuration and initialization logic whilst customizing chosen aspects of the process
|
|
|
* move all storage-related services into storage tree
* move config into config/
* also move admin service into storage since it mostly has storage related functionality
* header consolidation
|