feat(network): lean base pools (#2641)
Some checks are pending
Build / Build (MacOS 15) (push) Waiting to run
Build / Build (MacOS 26) (push) Waiting to run
Build / Build (AlmaLinux 10) (push) Waiting to run
Build / Build (Debian 12) (push) Waiting to run
Build / Build (Debian 13) (push) Waiting to run
Build / Build (Fedora 44) (push) Waiting to run
Build / Build (CentOS 10 Stream) (push) Waiting to run
Build / Build (CentOS 9 Stream) (push) Waiting to run
Build / Build (Ubuntu 26) (push) Waiting to run
Build / Build (Ubuntu 22) (push) Waiting to run
Build / Build (Ubuntu 24) (push) Waiting to run
Some checks are pending
Build / Build (MacOS 15) (push) Waiting to run
Build / Build (MacOS 26) (push) Waiting to run
Build / Build (AlmaLinux 10) (push) Waiting to run
Build / Build (Debian 12) (push) Waiting to run
Build / Build (Debian 13) (push) Waiting to run
Build / Build (Fedora 44) (push) Waiting to run
Build / Build (CentOS 10 Stream) (push) Waiting to run
Build / Build (CentOS 9 Stream) (push) Waiting to run
Build / Build (Ubuntu 26) (push) Waiting to run
Build / Build (Ubuntu 22) (push) Waiting to run
Build / Build (Ubuntu 24) (push) Waiting to run
**Follow-on to #2639 (merged). References a local IORingGroup `1.0.13-preview.11` pack until 1.0.13 (modernuo/IORingGroup#15) is published; do not merge before that switch.** ## Summary Consumes IORingGroup's lean base pools (modernuo/IORingGroup#15): both network pools now start with one slab, grow a slab at a time with the population, and trim idle slabs back after quiet periods. - Fixes the transport's send-pool cap: previously only 1024 of the 4096 connections could get a send buffer; connection 1025 was closed at accept. - Network memory at boot drops from about 96 MB to about 10 MB at the defaults; a full 4096 logged-in connections is about 1.25 GB of base buffers plus the growth budget. - New settings: `network.initialBufferSlabs` (default 1; slabs of each pool held from boot and the trim floor) and `network.maxBufferSlabs` (default 128; divides the connection maximum into slabs, 32 connections per slab). Both are coerced with a warning; the same value feeds the ring table and the manager so they cannot drift. - The Debug-only maintenance line includes base-pool capacity and releases. - `dev-docs/server-requirements.md` rewrites the network memory story and adds the two settings. ## Pre-auth buffers Every connection starts on the transport's platform-minimum buffers (4 KB receive, 4 KB send) instead of the base pools. It is promoted to full-size buffers (64 KB receive, `network.sendBufferSize` send) when the game server verifies its account — the point where `NetState.Account` is assigned in the `GameServer_AwaitingGameServerLogin` or `GameServer_LoggedIn` state, so a verdict that lands after the parser has moved on still promotes. Nothing ever moves back. The login-server pass stays on the small buffers for its whole lifetime. Before credentials verify, nothing promotes: the 4 KB send ring is the entire pre-auth send budget, and a connection that overruns it is dropped as exhausted, exactly as the receive side drops a packet header declaring more bytes than the receive buffer can hold (new guard in `HandlePacket`; it also closes the old 65535-byte edge on 64 KB buffers). The stock login sequence sends under 2 KB. After credentials verify, the send path promotes on demand if it ever needs to (unbudgeted, outside the memory ceiling and the shrink bookkeeping — promotion is not growth), and the oversize-packet guard waits on a pending receive promotion or retries a stalled one once for a verified account before disconnecting. A completion that fills the receive buffer arms no receive, so `HandleReceive` now calls `RingSocket.ResumeReceive()` after the parse loop — at 4 KB a burst of small packets fills the buffer in one completion. Net effect: a flood of unauthenticated connections tops out at about 32 MB across the full 4096-connection cap where the platform minimum is 4 KB (the transport's retained slabs and the base pools used by logged-in players are separate), and never allocates a base-pool slab. Platform note: on Windows Server 2012 R2 / 2016 the transport's legacy mapping path floors at 64 KB: the pre-auth receive pool is off there (its base is 64 KB), while the pre-auth send buffer starts at 64 KB under the 256 KB base. The server logs the effective sizes at startup. ## Testing Server.Tests (905) and UOContent.Tests green on the preview pack (one pre-existing `FamiliarAITests` failure from #2644 reproduces on `main`, tracked separately). New tests cover both coercions, that the ring's registration table equals `RequiredRegisteredBuffers` for the configured values, promotion on game-server auth and on a late account, no promotion on the login server, pre-credential overrun ending in exhaustion, post-credential on-demand promotion, the oversize-packet guard through loopback (error, wait, retry with an account, and a promotion made pending mid-parse), and receiving again after a burst fills the initial buffer.
This commit is contained in:
parent
9d9e672a09
commit
24bcfee554
6 changed files with 682 additions and 22 deletions
|
|
@ -29,18 +29,26 @@ namespace Server.Network;
|
|||
public partial class NetState
|
||||
{
|
||||
// Buffer sizes
|
||||
private const int RecvBufferSize = 1024 * 64; // 64KB recv buffers
|
||||
internal const int RecvBufferSize = 1024 * 64; // 64KB recv buffers
|
||||
private const int DefaultSendBufferSize = 1024 * 256; // 256KB send buffers
|
||||
private const int MinSendBufferSize = 1024 * 64; // Platform allocation granularity
|
||||
private const int DefaultMaxSendBufferSize = 1024 * 1024 * 2; // 2 MB
|
||||
private const long DefaultSendBufferGrowthBudget = 1024L * 1024 * 256; // 256 MB
|
||||
private const int DefaultMemoryCeilingPercent = 80;
|
||||
private const int DefaultInitialBufferSlabs = 1; // one slab of each base pool warm at boot
|
||||
internal const int DefaultMaxBufferSlabs = 128; // 32 connections per slab at MaxConnections
|
||||
|
||||
// Transport ceiling; larger values overflow its tier enumeration
|
||||
private const int TransportMaxSendBufferSize = 1024 * 1024 * 256; // 256 MB
|
||||
private const int MaxConnections = 4096; // Max concurrent connections
|
||||
|
||||
internal static int SendBufferSize { get; private set; }
|
||||
internal static int MaxSendBufferSize { get; private set; }
|
||||
|
||||
// Pre-auth buffers are the smallest the platform can map; a platform whose floor is not below
|
||||
// the base size (the Windows legacy mapping path) starts sockets on base
|
||||
internal static int InitialRecvBufferSize { get; private set; }
|
||||
internal static int InitialSendBufferSize { get; private set; }
|
||||
private static long _sendBufferGrowthBudget;
|
||||
private static int _memoryCeilingPercent;
|
||||
|
||||
|
|
@ -51,6 +59,7 @@ public partial class NetState
|
|||
private static long _lastTierCapacityBytes;
|
||||
private static int _lastTierInUse;
|
||||
private static int _lastTierRetainFloor;
|
||||
private static long _lastBaseCapacityBytes;
|
||||
|
||||
private static readonly Queue<NetState> _disposed = [];
|
||||
private static readonly TimeSpan ConnectingSocketIdleLimit = TimeSpan.FromMilliseconds(5000); // 5 seconds
|
||||
|
|
@ -141,23 +150,46 @@ public partial class NetState
|
|||
|
||||
// Per-connection send buffer: the lever for "send buffer exhausted" disconnects, and the
|
||||
// per-connection memory ceiling.
|
||||
var sendBufferSize = GetSendBufferSize();
|
||||
MaxSendBufferSize = GetPowerOfTwoSetting("network.sendBufferMaxSize", DefaultMaxSendBufferSize, sendBufferSize);
|
||||
SendBufferSize = GetSendBufferSize();
|
||||
MaxSendBufferSize = GetPowerOfTwoSetting("network.sendBufferMaxSize", DefaultMaxSendBufferSize, SendBufferSize);
|
||||
_sendBufferGrowthBudget = CoerceSendBufferGrowthBudget(
|
||||
ServerConfiguration.GetOrUpdateSetting("network.sendBufferGrowthBudget", DefaultSendBufferGrowthBudget),
|
||||
sendBufferSize,
|
||||
SendBufferSize,
|
||||
MaxSendBufferSize
|
||||
);
|
||||
// 0 disables the ceiling
|
||||
_memoryCeilingPercent = Math.Clamp(ServerConfiguration.GetOrUpdateSetting("network.memoryCeilingPercent", DefaultMemoryCeilingPercent), 0, 100);
|
||||
_availableMemoryBytes = GC.GetGCMemoryInfo().TotalAvailableMemoryBytes;
|
||||
|
||||
const int maxBufferSlabs = 32;
|
||||
// Divides MaxConnections into base-pool slabs; both pools still reach MaxConnections, so
|
||||
// this sets slab granularity, not a connection or memory ceiling.
|
||||
var maxBufferSlabs = CoerceMaxBufferSlabs(
|
||||
ServerConfiguration.GetOrUpdateSetting("network.maxBufferSlabs", DefaultMaxBufferSlabs)
|
||||
);
|
||||
|
||||
// Slabs of each base pool held from boot; the pools grow and trim from here
|
||||
var initialBufferSlabs = CoerceInitialBufferSlabs(
|
||||
ServerConfiguration.GetOrUpdateSetting("network.initialBufferSlabs", DefaultInitialBufferSlabs),
|
||||
BasePoolSlabCount(maxBufferSlabs)
|
||||
);
|
||||
|
||||
// Until the game server verifies credentials a connection holds the smallest buffers the
|
||||
// platform can map; a flood can fill these pools but never reaches the base pools. This is a
|
||||
// request: the manager raises it to the platform floor and turns the pool off if that leaves
|
||||
// no room below the base size (the Windows legacy mapping path floors at 64 KiB, which can
|
||||
// equal RecvBufferSize). The effective sizes are read back below once the manager knows them.
|
||||
var initialBufferSize = IORingBuffer.MinimumSize;
|
||||
|
||||
// Both calls take the same slab count; the manager throws if the table is smaller. Sized by
|
||||
// the request, not the effective size the manager may coerce down to - conservative, never small.
|
||||
var ring = IORingGroup.Create(
|
||||
queueSize: MaxConnections * 2,
|
||||
maxConnections: MaxConnections,
|
||||
maxOutstandingSends: maxOutstandingSends,
|
||||
maxRegisteredBuffers: RingSocketManager.RequiredRegisteredBuffers(MaxConnections, sendBufferSize, MaxSendBufferSize, _sendBufferGrowthBudget, maxBufferSlabs)
|
||||
maxRegisteredBuffers: RingSocketManager.RequiredRegisteredBuffers(
|
||||
MaxConnections, SendBufferSize, MaxSendBufferSize, _sendBufferGrowthBudget, maxBufferSlabs,
|
||||
initialBufferSize, initialBufferSize
|
||||
)
|
||||
);
|
||||
|
||||
// Create socket manager which handles buffer pools and socket lifecycle
|
||||
|
|
@ -165,13 +197,28 @@ public partial class NetState
|
|||
ring,
|
||||
maxSockets: MaxConnections,
|
||||
recvBufferSize: RecvBufferSize,
|
||||
sendBufferSize: sendBufferSize,
|
||||
initialBufferSlabs: 8,
|
||||
sendBufferSize: SendBufferSize,
|
||||
initialBufferSlabs: initialBufferSlabs,
|
||||
maxBufferSlabs: maxBufferSlabs,
|
||||
maxSendBufferSize: MaxSendBufferSize,
|
||||
sendBufferGrowthBudget: _sendBufferGrowthBudget
|
||||
sendBufferGrowthBudget: _sendBufferGrowthBudget,
|
||||
initialRecvBufferSize: initialBufferSize,
|
||||
initialSendBufferSize: initialBufferSize
|
||||
);
|
||||
|
||||
InitialRecvBufferSize = _socketManager.InitialRecvBufferSize;
|
||||
InitialSendBufferSize = _socketManager.InitialSendBufferSize;
|
||||
|
||||
if (InitialRecvBufferSize == 0 || InitialSendBufferSize == 0)
|
||||
{
|
||||
logger.Information(
|
||||
"Pre-auth buffers off where the platform floor ({Minimum} bytes) leaves no room below the base size (recv {Recv}, send {Send})",
|
||||
IORingBuffer.MinimumSize,
|
||||
RecvBufferSize,
|
||||
SendBufferSize
|
||||
);
|
||||
}
|
||||
|
||||
_maintenanceTimer = Timer.DelayCall(TimeSpan.FromMinutes(1), TimeSpan.FromMinutes(1), MaintainSendBuffers);
|
||||
}
|
||||
|
||||
|
|
@ -193,27 +240,79 @@ public partial class NetState
|
|||
var stats = _socketManager.Maintain();
|
||||
var changed = stats.TierCapacityBytes != _lastTierCapacityBytes ||
|
||||
stats.TierInUse != _lastTierInUse ||
|
||||
stats.TierRetainFloor != _lastTierRetainFloor;
|
||||
stats.TierRetainFloor != _lastTierRetainFloor ||
|
||||
stats.BaseCapacityBytes != _lastBaseCapacityBytes;
|
||||
_lastTierCapacityBytes = stats.TierCapacityBytes;
|
||||
_lastTierInUse = stats.TierInUse;
|
||||
_lastTierRetainFloor = stats.TierRetainFloor;
|
||||
_lastBaseCapacityBytes = stats.BaseCapacityBytes;
|
||||
|
||||
// Quiet unless something moved
|
||||
if (changed || stats.BuffersReleased > 0 || stats.GrowthRefusals > 0 || capRefusals > 0 || ceilingRefusals > 0)
|
||||
if (changed || stats.BuffersReleased > 0 || stats.BaseBuffersReleased > 0 || stats.GrowthRefusals > 0 ||
|
||||
capRefusals > 0 || ceilingRefusals > 0)
|
||||
{
|
||||
logger.Debug(
|
||||
"Send buffer tiers: {Capacity} bytes of tier capacity, {InUse} buffers in use, floor {Floor} buffers, released {Released}, refused: budget {BudgetRefusals}, at max {CapRefusals}, ceiling {CeilingRefusals}",
|
||||
"Send buffer tiers: {Capacity} bytes of tier capacity, {InUse} buffers in use, floor {Floor} buffers, released {Released}, refused: budget {BudgetRefusals}, at max {CapRefusals}, ceiling {CeilingRefusals}, base {BaseCapacity} bytes, released {BaseReleased}",
|
||||
stats.TierCapacityBytes,
|
||||
stats.TierInUse,
|
||||
stats.TierRetainFloor,
|
||||
stats.BuffersReleased,
|
||||
stats.GrowthRefusals,
|
||||
capRefusals,
|
||||
ceilingRefusals
|
||||
ceilingRefusals,
|
||||
stats.BaseCapacityBytes,
|
||||
stats.BaseBuffersReleased
|
||||
);
|
||||
}
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Slabs each base pool is divided into, after the transport applies its minimum slab size.
|
||||
/// </summary>
|
||||
internal static int BasePoolSlabCount(int maxBufferSlabs) =>
|
||||
RingSocketManager.BasePoolSlabCount(MaxConnections, maxBufferSlabs);
|
||||
|
||||
/// <summary>
|
||||
/// Clamps the base-pool slab divisor. Fewer than one slab is meaningless, and more slabs than
|
||||
/// connections cannot make a slab any smaller.
|
||||
/// </summary>
|
||||
internal static int CoerceMaxBufferSlabs(int configured)
|
||||
{
|
||||
var slabs = Math.Clamp(configured, 1, MaxConnections);
|
||||
|
||||
if (slabs != configured)
|
||||
{
|
||||
logger.Warning(
|
||||
"network.maxBufferSlabs {Configured} is outside 1..{Maximum}; using {Adjusted}",
|
||||
configured,
|
||||
MaxConnections,
|
||||
slabs
|
||||
);
|
||||
}
|
||||
|
||||
return slabs;
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Clamps the slabs of each base pool held from boot; a pool never holds more slabs than it has.
|
||||
/// </summary>
|
||||
internal static int CoerceInitialBufferSlabs(int configured, int slabCount)
|
||||
{
|
||||
var slabs = Math.Clamp(configured, 1, slabCount);
|
||||
|
||||
if (slabs != configured)
|
||||
{
|
||||
logger.Warning(
|
||||
"network.initialBufferSlabs {Configured} is outside 1..{Maximum}; using {Adjusted}",
|
||||
configured,
|
||||
slabCount,
|
||||
slabs
|
||||
);
|
||||
}
|
||||
|
||||
return slabs;
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Reads the configured send buffer size, coerced to a power of two of at least the platform
|
||||
/// allocation granularity. IORingBuffer requires this and would otherwise throw at socket
|
||||
|
|
|
|||
|
|
@ -227,7 +227,39 @@ public partial class NetState : IComparable<NetState>, IValueLinkListNode<NetSta
|
|||
public IAccount Account
|
||||
{
|
||||
get => _account;
|
||||
set => _account = value;
|
||||
set
|
||||
{
|
||||
_account = value;
|
||||
|
||||
// 0x91 credentials just verified: the character list and the world-entry burst follow.
|
||||
// The login-server pass stays on the initial buffers; it never sends more than a server list.
|
||||
// An off-loop password check can land its verdict after HandleReceive already flipped the
|
||||
// state to LoggedIn, so both states promote; both calls below are no-ops once applied.
|
||||
if (value != null && _protocolState is ProtocolState.GameServer_AwaitingGameServerLogin or ProtocolState.GameServer_LoggedIn)
|
||||
{
|
||||
PromoteBuffers();
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
private void PromoteBuffers()
|
||||
{
|
||||
if (_socket == null)
|
||||
{
|
||||
return;
|
||||
}
|
||||
|
||||
if (!_socketManager.TryPromoteSendBuffer(_socket) && _socket.SendBuffer.PhysicalSize < SendBufferSize)
|
||||
{
|
||||
// The send path promotes on demand and disconnects if that fails too
|
||||
logger.Debug("{NetState}: send buffer promotion deferred to the send path", this);
|
||||
}
|
||||
|
||||
if (!_socketManager.TryPromoteRecvBuffer(_socket) && _socket.RecvBuffer.PhysicalSize < RecvBufferSize)
|
||||
{
|
||||
// The oversize-packet guard in HandlePacket gets one more try where the small buffer matters
|
||||
logger.Debug("{NetState}: recv buffer promotion deferred", this);
|
||||
}
|
||||
}
|
||||
|
||||
public string Assistant { get; set; }
|
||||
|
|
@ -525,6 +557,19 @@ public partial class NetState : IComparable<NetState>, IValueLinkListNode<NetSta
|
|||
return false;
|
||||
}
|
||||
|
||||
if (_socket.SendBuffer.PhysicalSize < SendBufferSize)
|
||||
{
|
||||
// Promotion is unbudgeted and is not growth, so neither the ceiling nor the shrink
|
||||
// bookkeeping applies. Before credentials verify nothing promotes: the 4 KiB ring is the
|
||||
// whole pre-auth send budget, and a connection that exceeds it is dropped as exhausted.
|
||||
if (_account == null)
|
||||
{
|
||||
return false;
|
||||
}
|
||||
|
||||
return _socketManager.TryPromoteSendBuffer(_socket);
|
||||
}
|
||||
|
||||
if (!UnderMemoryCeiling())
|
||||
{
|
||||
_ceilingRefusals++;
|
||||
|
|
@ -1084,6 +1129,13 @@ public partial class NetState : IComparable<NetState>, IValueLinkListNode<NetSta
|
|||
break;
|
||||
}
|
||||
}
|
||||
|
||||
// A completion that fills the buffer arms no receive; whatever the loop consumed is free
|
||||
// space again. No-op while a receive is armed or the buffer is still full.
|
||||
if (_running)
|
||||
{
|
||||
_socket.ResumeReceive();
|
||||
}
|
||||
}
|
||||
catch (Exception ex)
|
||||
{
|
||||
|
|
@ -1137,6 +1189,22 @@ public partial class NetState : IComparable<NetState>, IValueLinkListNode<NetSta
|
|||
}
|
||||
}
|
||||
|
||||
// Can never complete: the buffer holds PhysicalSize - 1 bytes (one slot tells full from
|
||||
// empty) and a recv arms only into free space
|
||||
if (packetLength >= _socket.RecvBuffer.PhysicalSize)
|
||||
{
|
||||
// A verified account whose promotion could not be applied earlier gets one more try here,
|
||||
// where the small buffer actually matters; the swap lands before the next completion's event
|
||||
if (_account != null && _socket.RecvBuffer.PhysicalSize < RecvBufferSize &&
|
||||
_socketManager.TryPromoteRecvBuffer(_socket))
|
||||
{
|
||||
return ParserState.AwaitingPartialPacket;
|
||||
}
|
||||
|
||||
LogInfo($"Received packet 0x{packetId:X2} declaring {packetLength} bytes, more than the receive buffer holds.");
|
||||
return ParserState.Error;
|
||||
}
|
||||
|
||||
// Not enough data, let's wait for more to come in
|
||||
if (length < packetLength)
|
||||
{
|
||||
|
|
|
|||
|
|
@ -34,7 +34,7 @@
|
|||
</Target>
|
||||
<ItemGroup>
|
||||
<ProjectReference Include="..\Logger\Logger.csproj" />
|
||||
<PackageReference Include="IORingGroup" Version="1.0.12" />
|
||||
<PackageReference Include="IORingGroup" Version="1.0.13" />
|
||||
<PackageReference Include="CommunityToolkit.HighPerformance" Version="8.4.2" />
|
||||
<PackageReference Include="LibDeflate.Bindings" Version="1.0.4" />
|
||||
<PackageReference Include="System.IO.Hashing" Version="10.0.12" />
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue