mirror of
https://github.com/rstrouse/ESPSomfy-RTS.git
synced 2026-10-31 22:02:17 +01:00
082d9776ed
- A websocket event too large for its buffer has been logged and not
sent since 8881289, not sent truncated. Retitle the entry and point at
MAX_SOCK_RESPONSE by name rather than a stale line number.
- List the web server task handlers that still change the shade model
loop() owns: /setPaired, /setRollingCode and restore.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
83 lines
6.6 KiB
Markdown
83 lines
6.6 KiB
Markdown
# Known Issues
|
||
|
||
## Concurrent mutation during chunked `/controller` response
|
||
|
||
**Status:** open
|
||
**Introduced:** 2026-04-19, alongside the chunked-response conversion of `/controller` (see [`ControllerChunker` in src/Web.cpp](../src/Web.cpp) and the original crash report about `cbuf::resize` aborting `async_tcp`).
|
||
|
||
### Summary
|
||
|
||
The chunked streaming of `/controller` lazily reads `somfy.rooms`, `somfy.shades`, `somfy.groups`, and `somfy.repeaters` across many `async_tcp` callback invocations as the client's TCP send window opens. If another task mutates those arrays mid-stream, the rendered JSON can be internally inconsistent.
|
||
|
||
### Why this is strictly worse than the pre-chunked code
|
||
|
||
The old `handleController` ran the serializers synchronously inside a single request-handler invocation, so the full JSON was built in memory *before* any bytes were sent. Window of exposure: a few milliseconds.
|
||
|
||
The new `ControllerChunker` reads state as the response drains. On a slow link or under backpressure, that window is hundreds of ms to seconds.
|
||
|
||
### Concrete failure modes
|
||
|
||
1. A shade is deleted mid-stream (e.g. via `/deleteShade`) — it may appear in the `shades` array but be missing from a later group's `linkedShades`, or vice versa.
|
||
2. `lastRollingCode` / position fields change between the `shades` pass and a later group's `linkedShades` pass — the client sees two values for the same shade in one document.
|
||
3. A group's `linkedShades` list is mutated while the chunker iterates inside `S_GROUPS` — an entry is skipped or emitted twice.
|
||
|
||
### Fix options (pick one later)
|
||
|
||
- **(a) Document and accept.** In practice users rarely mutate shades while the config UI is loading. Zero code change.
|
||
- **(b) FreeRTOS mutex around `somfy` reads/writes.** Acquire for the full duration of the chunked response and in every mutating path (RF RX, MQTT, web handlers). Cleanest but wide-reaching.
|
||
- **(c) Up-front snapshot.** At handler start, copy the subset of `somfy` state the response will serialize into the `ControllerChunker`. Defeats part of the memory benefit — a full snapshot of shades + groups is close in size to the old growing cbuf. Could be reduced by snapshotting only minimal fields (IDs, names, rolling codes) and reading the rest live.
|
||
|
||
### Related
|
||
|
||
- Same exposure exists in any other endpoint converted to chunked responses next (`/discovery`, `/shades`). Resolve this issue before expanding the pattern.
|
||
- Other handlers on the web server task still change the shade model that `loop()` owns: `/setPaired` sets `paired`, and `/setRollingCode` and a shades restore change rolling codes (see [Rolling codes written from two tasks](#rolling-codes-written-from-two-tasks)). Shade commands, set-My and unpair already go through the web command queue to `loop()`.
|
||
|
||
## Large websocket events are dropped
|
||
|
||
**Status:** open
|
||
**Location:** [`JsonSockEvent` in src/WResp.cpp](../src/WResp.cpp), buffer `g_response[MAX_SOCK_RESPONSE]` (2048 bytes) in [src/Sockets.cpp](../src/Sockets.cpp).
|
||
|
||
### Summary
|
||
|
||
Socket events are built into a fixed 2 KB static buffer. When a value doesn't fit, [`JsonSockEvent::_safecat`](../src/WResp.cpp) marks the event as overflowed, and `endEvent` logs an error and doesn't send it. Clients never see malformed JSON. The event is simply lost, and the client isn't told.
|
||
|
||
Unlike the `/controller` HTTP crash, this path does **not** abort — there is no growing cbuf and no `new[]` on the send path. Per-client frame allocations inside `AsyncWebSocket` are bounded by the 2 KB buffer size and have their own overflow guard (queue drop / client disconnect).
|
||
|
||
### Concrete failure modes
|
||
|
||
1. A single event serializing a fully-populated shade (~1.3–1.5 KB for a shade with all `SOMFY_MAX_LINKED_REMOTES` = 7 populated) gets close to the 2 KB limit. Any additional fields or long names push it over, and that shade's state update never reaches the clients.
|
||
2. Any event that loops over a collection (e.g. frequency-scan results, batch emits in `Somfy.cpp`) can exceed 2 KB depending on size, with no indication to the client beyond the ESP-side `ESP_LOGE` line.
|
||
|
||
### Fix options (pick one later)
|
||
|
||
- **(a) Fail loud.** Send a small error event in place of the dropped one so the client knows to reload the state.
|
||
- **(b) Split large events across frames.** Use the socket.io ack/chunk pattern to send an event in multiple frames when it wouldn't fit. Requires matching client-side reassembly.
|
||
- **(c) Raise `MAX_SOCK_RESPONSE`.** Cheapest, but just pushes the limit — doesn't eliminate the failure mode.
|
||
|
||
### Related
|
||
|
||
- Not the same code path as the `/controller` crash. Solve independently.
|
||
- Worth grepping for `JsonSockEvent` usages that iterate collections (see references in `Somfy.cpp`, `ESPNetwork.cpp`, `GitOTA.cpp`) to identify the most at-risk events.
|
||
|
||
## Rolling codes written from two tasks
|
||
|
||
**Status:** open
|
||
**Location:** `SomfyRemote::sendCommandAfter` and `setRollingCode` in [src/Somfy.cpp](../src/Somfy.cpp), the `/setRollingCode` handlers in [src/Web.cpp](../src/Web.cpp), and `ShadeConfigFile::readShadeRecord`/`readGroupRecord` in [src/ConfigFile.cpp](../src/ConfigFile.cpp).
|
||
|
||
### Summary
|
||
|
||
Sending a command reads the next rolling code, queues the frame, then commits the code (`peekNextRollingCode` → `submitTx` → `commitRollingCode`), all on `loop()`. `/setRollingCode` and a shades restore change the same remote's `lastRollingCode` and its `ShadeCodes` NVS key from the async web server task, with no lock against that sequence. Each NVS write is safe on its own, since every function opens its own handle (see [ADR-0001](adr/0001-local-nvs-handles-no-wrapper.md)), but the read-then-commit sequence is not.
|
||
|
||
### Concrete failure modes
|
||
|
||
1. `/setRollingCode` lands between the peek and the commit of a command. The command's commit overwrites the code just set, so the set has no effect.
|
||
2. A restore reads `ShadeCodes` while a command commits a newer code. The restored shade can come back with the older code, and the motor rejects its frames until the code passes the newer one.
|
||
|
||
### Fix options (pick one later)
|
||
|
||
- **(a) Run these handlers on `loop()`.** Queue `/setRollingCode` and restore through the web command queue like shade commands, so `loop()` is the only task that touches a remote's rolling code. Part of making `loop()` the only owner of the shade model.
|
||
- **(b) A rolling-code lock.** Hold a mutex across peek → queue → commit, and in `setRollingCode` and restore. Narrower, but it adds a lock to the command path.
|
||
|
||
### Related
|
||
|
||
- Same root cause as [Concurrent mutation during chunked `/controller` response](#concurrent-mutation-during-chunked-controller-response): the web server task changes the shade model that `loop()` owns.
|