Files
DorkNet/DorkNet.Server/Services/PlayerPresenceService.cs
T
Alexa 323c17bedb Refactor room data blob handling and permissions
- Updated the RoomsController to utilize synthetic CDN keys for data blobs.
- Introduced methods for determining current or synthetic data blob names.
- Adjusted room role permissions to align with new game role definitions.
- Enhanced the SanitizeController with a new endpoint for purity checks.
- Modified Match and PlayerPresence services to handle deferred presence commits.
- Improved RoomDataBlobService to support new game role and tool tag settings.
- Added support for boot dorm grace period in PlayerPresenceService.
- Updated protobuf definitions to reflect new room role and game role structures.

You are now able to go into dorm and the rec center!
2026-06-24 12:02:13 +03:00

503 lines
19 KiB
C#

using System.Collections.Concurrent;
using System.Text.Json;
using DorkNet.Server.Controllers.Match;
using Microsoft.Extensions.Logging;
using StackExchange.Redis;
namespace DorkNet.Server.Services;
/// <summary>
/// Cross-replica tracker for which RoomInstance each authenticated
/// player most recently went to via /goto. The presence heartbeat
/// reads this back so the client's local state matches what the
/// server says, and is the only passive path that extends the room
/// TTL.
///
/// Why this exists: every ~5s the client POSTs /player/heartbeat,
/// expecting a PlayerPresence response that mirrors its own cached
/// presence. If the server returns roomInstance:null but the client
/// knows it's in DormRoom, RecNet.Matchmaking.OnPresenceHeartbeatResponse
/// logs "presence heartbeat response indicates local presence is
/// out-of-sync" repeatedly, the client gives up after a few rounds,
/// and shows the "Error: -2 / We lost the connection to Rec Room"
/// UI even though no network connection actually dropped.
///
/// **Multi-replica**: when <see cref="IConnectionMultiplexer"/> is
/// available, presence lives in Redis as <c>presence:{playerId}</c>
/// (JSON-serialised <see cref="RoomInstanceDto"/>). EXPIRE is set to
/// <see cref="PresenceTtl"/> so a player who silently drops
/// (uninstall, kill -9) ages out automatically; their next /goto
/// refreshes the value and heartbeat refreshes the TTL. Read-only
/// presence queries intentionally do not refresh it, otherwise admin
/// pages and friends-list polls keep stale rooms alive forever. When
/// no Redis is configured, falls back to a process-local
/// <see cref="ConcurrentDictionary"/>
/// — identical to pre-PR-2 behaviour and fine for single-instance
/// dev workflows.
///
/// State is intentionally non-persistent — a server restart drops
/// every presence record, and clients re-establish via their next
/// /goto call. Method signatures stay sync because the call sites
/// (heartbeat, /goto, room-state queries) are too numerous to pay
/// the cost of an async cascade. Sync StackExchange.Redis ops are
/// sub-millisecond on the same network at our QPS.
/// </summary>
public class PlayerPresenceService
{
private readonly ConcurrentDictionary<long, RoomInstanceDto> _local = new();
private readonly ConcurrentDictionary<long, DateTimeOffset> _localRoomExpires = new();
private readonly ConcurrentDictionary<long, DateTimeOffset> _localActiveExpires = new();
private readonly ConcurrentDictionary<long, DateTimeOffset> _localBootDormGraceExpires = new();
private readonly IConnectionMultiplexer? _redis;
private readonly ILogger<PlayerPresenceService>? _log;
/// <summary>
/// Presence must survive long room loads. The 2020 client can go
/// quiet for ~45 seconds while downloading/deserializing large
/// uploaded rooms; a 45s Redis TTL expires right before the next
/// heartbeat, making the server synthesize dorm fallback presence
/// and effectively bouncing the player back to dorm. Ten minutes
/// still ages out abandoned sessions, but does not punish slow
/// room loads or temporary network stalls.
/// </summary>
public static TimeSpan PresenceTtl { get; } = TimeSpan.FromMinutes(10);
/// <summary>
/// Short online-status marker refreshed by heartbeat/join activity.
/// This smooths over flaky SignalR notify sockets without letting a
/// stale room presence make someone look online for the full room TTL.
/// </summary>
public static TimeSpan ActivityTtl { get; } = TimeSpan.FromSeconds(90);
/// <summary>
/// Right after login the 2023 watch may boot into its cached offline
/// DormRoom before the live Photon join finishes. During that window
/// a null roomInstance is safer than returning stale/live presence:
/// the client treats the offline placeholder as invalid and accepts
/// null, but a live private dorm forces a resync.
/// </summary>
public static TimeSpan BootDormGraceTtl { get; } = TimeSpan.FromSeconds(45);
private static readonly JsonSerializerOptions JsonOpts = new()
{
// Match the wire format used by the matchmaking response so
// the round-trip is lossless. Properties are already tagged
// with [JsonPropertyName] for camelCase wire keys.
PropertyNamingPolicy = null,
};
public PlayerPresenceService(
IConnectionMultiplexer? redis = null,
ILogger<PlayerPresenceService>? log = null)
{
_redis = redis;
_log = log;
}
public void SetRoom(long playerId, RoomInstanceDto room)
{
if (_redis is { } mux)
{
try
{
var json = JsonSerializer.Serialize(room, JsonOpts);
mux.GetDatabase().StringSet(Key(playerId), json, PresenceTtl);
_log?.LogDebug("[presence] set redis player={PlayerId} room={Room} bytes={Bytes}",
playerId, room?.Name ?? "<null>", json.Length);
}
catch (Exception ex)
{
// Don't let a Redis blip take down the goto handler — log
// and fall through. Heartbeat will see no presence and the
// watch will resync via its next /goto.
_log?.LogWarning(ex, "[presence] redis SET failed for player={PlayerId}", playerId);
}
}
else
{
_local[playerId] = room;
_localRoomExpires[playerId] = DateTimeOffset.UtcNow.Add(PresenceTtl);
_log?.LogDebug("[presence] set local player={PlayerId} room={Room}",
playerId, room?.Name ?? "<null>");
}
}
public void TouchRoom(long playerId)
{
if (_redis is { } mux)
{
try
{
mux.GetDatabase().KeyExpire(Key(playerId), PresenceTtl);
}
catch (Exception ex)
{
_log?.LogWarning(ex, "[presence] redis EXPIRE failed for player={PlayerId}", playerId);
}
return;
}
if (_local.ContainsKey(playerId))
_localRoomExpires[playerId] = DateTimeOffset.UtcNow.Add(PresenceTtl);
}
public void MarkActive(long playerId)
{
if (_redis is { } mux)
{
try
{
mux.GetDatabase().StringSet(ActiveKey(playerId), "1", ActivityTtl);
mux.GetDatabase().SetAdd(ActiveSetKey, playerId);
}
catch (Exception ex)
{
_log?.LogWarning(ex, "[presence] redis active SET failed for player={PlayerId}", playerId);
}
return;
}
_localActiveExpires[playerId] = DateTimeOffset.UtcNow.Add(ActivityTtl);
}
public void BeginBootDormGrace(long playerId)
{
if (_redis is { } mux)
{
try
{
mux.GetDatabase().StringSet(BootDormGraceKey(playerId), "1", BootDormGraceTtl);
}
catch (Exception ex)
{
_log?.LogWarning(ex, "[presence] redis boot grace SET failed for player={PlayerId}", playerId);
}
return;
}
_localBootDormGraceExpires[playerId] = DateTimeOffset.UtcNow.Add(BootDormGraceTtl);
}
public bool IsBootDormGraceActive(long playerId)
{
if (_redis is { } mux)
{
try
{
return mux.GetDatabase().KeyExists(BootDormGraceKey(playerId));
}
catch (Exception ex)
{
_log?.LogWarning(ex, "[presence] redis boot grace GET failed for player={PlayerId}", playerId);
return false;
}
}
if (!_localBootDormGraceExpires.TryGetValue(playerId, out var expires)) return false;
if (expires > DateTimeOffset.UtcNow) return true;
_localBootDormGraceExpires.TryRemove(playerId, out _);
return false;
}
public void EndBootDormGrace(long playerId)
{
if (_redis is { } mux)
{
try
{
mux.GetDatabase().KeyDelete(BootDormGraceKey(playerId));
}
catch (Exception ex)
{
_log?.LogWarning(ex, "[presence] redis boot grace DEL failed for player={PlayerId}", playerId);
}
return;
}
_localBootDormGraceExpires.TryRemove(playerId, out _);
}
public IReadOnlyCollection<long> RecentlyActivePlayerIds()
{
if (_redis is { } mux)
{
try
{
var db = mux.GetDatabase();
var members = db.SetMembers(ActiveSetKey);
if (members.Length == 0) return Array.Empty<long>();
var activeIds = new List<long>(members.Length);
var stale = new List<RedisValue>();
foreach (var member in members)
{
if (!long.TryParse(member.ToString(), out var playerId))
{
stale.Add(member);
continue;
}
if (db.KeyExists(ActiveKey(playerId)))
activeIds.Add(playerId);
else
stale.Add(member);
}
if (stale.Count > 0)
db.SetRemove(ActiveSetKey, stale.ToArray());
return activeIds;
}
catch (Exception ex)
{
_log?.LogWarning(ex, "[presence] redis active scan failed");
return Array.Empty<long>();
}
}
var now = DateTimeOffset.UtcNow;
var active = new List<long>();
foreach (var kv in _localActiveExpires)
{
if (kv.Value > now)
{
active.Add(kv.Key);
}
else
{
_localActiveExpires.TryRemove(kv.Key, out _);
}
}
return active;
}
public bool IsRecentlyActive(long playerId)
{
if (_redis is { } mux)
{
try
{
return mux.GetDatabase().KeyExists(ActiveKey(playerId));
}
catch (Exception ex)
{
_log?.LogWarning(ex, "[presence] redis active GET failed for player={PlayerId}", playerId);
return false;
}
}
if (!_localActiveExpires.TryGetValue(playerId, out var expires)) return false;
if (expires > DateTimeOffset.UtcNow) return true;
_localActiveExpires.TryRemove(playerId, out _);
return false;
}
public RoomInstanceDto? GetRoom(long playerId)
{
if (_redis is { } mux)
{
try
{
var db = mux.GetDatabase();
var key = Key(playerId);
var v = db.StringGet(key);
if (!v.HasValue)
{
_log?.LogDebug("[presence] miss redis player={PlayerId}", playerId);
return null;
}
return JsonSerializer.Deserialize<RoomInstanceDto>(v.ToString(), JsonOpts);
}
catch (Exception ex)
{
_log?.LogWarning(ex, "[presence] redis GET failed for player={PlayerId}", playerId);
return null;
}
}
if (!_local.TryGetValue(playerId, out var room)) return null;
if (_localRoomExpires.TryGetValue(playerId, out var expires)
&& expires <= DateTimeOffset.UtcNow)
{
_local.TryRemove(playerId, out _);
_localRoomExpires.TryRemove(playerId, out _);
_localActiveExpires.TryRemove(playerId, out _);
return null;
}
return room;
}
public void Clear(long playerId)
{
if (_redis is { } mux)
{
var db = mux.GetDatabase();
db.KeyDelete(Key(playerId));
db.KeyDelete(ActiveKey(playerId));
db.KeyDelete(BootDormGraceKey(playerId));
db.SetRemove(ActiveSetKey, playerId);
}
else
{
_local.TryRemove(playerId, out _);
_localRoomExpires.TryRemove(playerId, out _);
_localActiveExpires.TryRemove(playerId, out _);
_localBootDormGraceExpires.TryRemove(playerId, out _);
}
}
/// <summary>List the currently-active instances of a given room,
/// grouped by PhotonRoomId, with the player ids currently presenced
/// in each. Drives the watch's RoomInstanceBrowser
/// (<c>GET match.*/room/{id}/instances</c>) so players can see
/// which instance their friends are in.
///
/// Redis caveat: without a secondary index per room we'd need a
/// SCAN over every presence key — too expensive to do on each
/// request. For now the Redis path returns empty and the local
/// in-process dictionary is the source of truth. Single-instance
/// deployments work; multi-replica Redis deployments need a
/// per-room set index to be added before the browser is reliable
/// there.</summary>
public IEnumerable<(RoomInstanceDto Room, List<long> PlayerIds)> EnumerateActiveInstances(long roomId)
{
if (_redis is not null) return Array.Empty<(RoomInstanceDto, List<long>)>();
var groups = new Dictionary<string, (RoomInstanceDto Room, List<long> PlayerIds)>(StringComparer.Ordinal);
foreach (var kv in _local)
{
var room = kv.Value;
if (room is null || room.RoomId != roomId) continue;
// Skip private instances — those are surfaced separately
// in the controller via PrivateInstanceService so we never
// leak a photonRoomId for an invite-only match into the
// public browser.
if (room.IsPrivate) continue;
var key = string.IsNullOrEmpty(room.PhotonRoomId)
? $"<{room.RoomInstanceId}>"
: room.PhotonRoomId;
if (!groups.TryGetValue(key, out var entry))
{
entry = (room, new List<long>());
groups[key] = entry;
}
entry.PlayerIds.Add(kv.Key);
}
return groups.Values;
}
private static string Key(long playerId) => $"presence:{playerId}";
private static string ActiveKey(long playerId) => $"presence-active:{playerId}";
private static string BootDormGraceKey(long playerId) => $"presence-boot-dorm-grace:{playerId}";
private const string ActiveSetKey = "presence-active";
// ── LoginLock single-session enforcement ─────────────────────────
// The 2020 watch generates a fresh GUID per session and sends it
// in the `LoginLock` form field on every match.* request. Real
// Rec Room rejects heartbeats whose lock doesn't match the
// most-recent /player/login — that's how a second login on the
// same account kicks the first session. We mirror that by storing
// the most-recent lock in Redis (or the in-process fallback)
// keyed by player id, with the same 45-s TTL as presence.
//
// SetLock OVERWRITES regardless of prior value — newest login
// wins. ValidateLock returns false when the caller's lock is stale
// or absent, which the heartbeat handler maps to a
// session-terminated response so the watch logs out.
private readonly ConcurrentDictionary<long, string> _localLocks = new();
public bool SetLock(long playerId, string loginLock)
{
if (string.IsNullOrWhiteSpace(loginLock)) return false;
var replacedActiveSession = false;
if (_redis is { } mux)
{
try
{
var db = mux.GetDatabase();
var key = LockKey(playerId);
var previous = db.StringGet(key);
replacedActiveSession = previous.HasValue &&
!string.Equals(previous.ToString(), loginLock, StringComparison.Ordinal);
db.StringSet(key, loginLock, PresenceTtl);
}
catch (Exception ex) { _log?.LogWarning(ex, "[lock] redis SET failed for player={PlayerId}", playerId); }
}
else
{
replacedActiveSession = _localLocks.TryGetValue(playerId, out var previous) &&
!string.Equals(previous, loginLock, StringComparison.Ordinal);
_localLocks[playerId] = loginLock;
}
if (replacedActiveSession)
_log?.LogWarning("[lock] concurrent login detected for player={PlayerId}; newest LoginLock wins", playerId);
return replacedActiveSession;
}
public bool ValidateLock(long playerId, string? loginLock)
{
if (string.IsNullOrWhiteSpace(loginLock)) return false;
if (_redis is { } mux)
{
try
{
var v = mux.GetDatabase().StringGet(LockKey(playerId));
if (!v.HasValue) return false;
// Refresh TTL on each successful validate so the
// active session keeps its lock alive across heartbeats.
if (string.Equals(v.ToString(), loginLock, StringComparison.Ordinal))
{
mux.GetDatabase().KeyExpire(LockKey(playerId), PresenceTtl);
return true;
}
return false;
}
catch (Exception ex)
{
_log?.LogWarning(ex, "[lock] redis GET failed for player={PlayerId}", playerId);
return true; // fail open — better double-session than mass kick on Redis blip
}
}
return _localLocks.TryGetValue(playerId, out var stored)
&& string.Equals(stored, loginLock, StringComparison.Ordinal);
}
private static string LockKey(long playerId) => $"loginlock:{playerId}";
// ── Photon region pings ──────────────────────────────────────────
// The 2020 watch reports its measured Photon-region latencies in a
// single POST after the matchmaking init handshake (see
// MatchPlayerController.PhotonRegionPings). We hold the latest
// reading per player in process memory only — region preferences
// change infrequently, lose-on-restart is fine, and pushing this
// through Redis would just bloat the key space for no gain.
private readonly ConcurrentDictionary<long, IReadOnlyDictionary<string, int>> _regionPings = new();
public void SetPhotonRegionPings(long playerId, IReadOnlyDictionary<string, int> pings)
{
if (pings.Count == 0) return;
_regionPings[playerId] = pings;
}
/// <summary>Returns the region code with the lowest reported ping,
/// or <c>null</c> when no pings were ever received for the
/// player (matchmaking falls back to its default region in that
/// case).</summary>
public string? GetPreferredPhotonRegion(long playerId)
{
if (!_regionPings.TryGetValue(playerId, out var pings) || pings.Count == 0)
return null;
string? best = null;
var bestMs = int.MaxValue;
foreach (var kv in pings)
{
if (kv.Value < bestMs)
{
bestMs = kv.Value;
best = kv.Key;
}
}
return best;
}
}