fix(bridge): a backoff that holds against a sidecar that refuses (D155)
All checks were successful
PR Checks / plugin-checks (pull_request) Successful in -1m38s

The reconnect backoff waited on `Wake`, which `Emit` signals on every
frame, so it lasted only until the next hook fired. That was invisible
while a failed connect was silent; against a sidecar that now refuses a
plugin naming another server (D155) the phase 18 walk measured some 250
redials a second, each printing two console lines and rebuilding the
boards. It now waits on `_stopping`, the handle that already exists so a
CONNECT is not abandoned by `Emit`; Unload still ends the wait at once.

The backoff also resets only after a link has held for 10 s
(StableLinkMs). A connection closed on arrival used to reset it to 500 ms
on every accept, so a refusal loop never backed off at all.

Walked: a plugin aimed at another server's sidecar is refused every 5 s
(the ceiling), not continuously; a stable link whose sidecar restarts is
back in 0.49 s, as before; no Unload stall on reload.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01E14m6SuuY6i1vASFeGDBeY
This commit is contained in:
2026-09-26 01:19:23 -05:00
parent 04491e70af
commit 3e928bb2f6

View File

@@ -65,6 +65,15 @@ namespace Oxide.Plugins
/// </summary>
private const int MaxBackoffMs = 5000;
/// <summary>
/// How long a link must stay up before its drop counts as an outage rather than a
/// refusal. The backoff resets only past it: a sidecar that closes every connection at
/// once — one refusing a plugin that names another server (D155), or one crash-looping —
/// would otherwise be redialled twice a second for ever, rebuilding the boards on the
/// main thread and printing two console lines each time.
/// </summary>
private const int StableLinkMs = 10000;
/// <summary>The longest line accepted from the sidecar. Over-long lines are discarded.</summary>
private const int MaxInboundLineBytes = 1024 * 1024;
@@ -570,6 +579,7 @@ namespace Oxide.Plugins
{
TcpClient client = null;
Thread reader = null;
System.Diagnostics.Stopwatch upFor = null;
int epoch = Interlocked.Increment(ref _epoch);
@@ -588,7 +598,7 @@ namespace Oxide.Plugins
_dead = false;
_connected = true;
_loggedConnectFailure = false; // so a LATER outage is reported again
backoffMs = 500;
upFor = System.Diagnostics.Stopwatch.StartNew();
Interlocked.Increment(ref _connects);
Puts($"connected to {host}:{port}");
@@ -660,14 +670,25 @@ namespace Oxide.Plugins
if (!_running)
break;
// A link that held is an outage: start again from the shortest wait. One that was
// closed on arrival keeps doubling (see StableLinkMs).
if (upFor != null && upFor.ElapsedMilliseconds >= StableLinkMs)
backoffMs = 500;
// Nothing is listening yet, or the sidecar restarted. Both are normal.
//
// `Wake`, not `Thread.Sleep`: Unload runs on the MAIN thread and joins this one, so
// an uninterruptible sleep here is a stall of up to MaxBackoffMs on every reload —
// Oxide reports it as `Calling 'Unload' took 1918ms`, and the server is frozen for
// that whole time. Unload sets `_running` false and signals `Wake`, so waiting on
// the handle turns that stall into an immediate return.
Wake.WaitOne(backoffMs);
// A handle, not `Thread.Sleep`: Unload runs on the MAIN thread and joins this one,
// so an uninterruptible sleep here is a stall of up to MaxBackoffMs on every reload
// — Oxide reports it as `Calling 'Unload' took 1918ms`, and the server is frozen for
// that whole time. Unload sets `_stopping`, so waiting on it turns that stall into
// an immediate return.
//
// `_stopping`, not `Wake`, for the reason `_stopping` gives: `Emit` signals `Wake`
// on every frame, so a backoff waited on it lasted until the next hook fired. That
// was invisible while a failed connect was silent; against a sidecar that refuses
// (D155) the phase 18 walk measured some 250 redials a second, each printing two
// console lines and rebuilding the boards.
_stopping.WaitOne(backoffMs);
backoffMs = Math.Min(backoffMs * 2, MaxBackoffMs);
}