Add support for multi-process port sharing with CIBIR.

docs/Settings.md

@@ -169,7 +169,7 @@ These parameters are accessed by calling [GetParam](./api/GetParam.md) or [SetPa
 |-------------------------------------------|---------------------------|-----------|-----------------------------------------------------------|
 | `QUIC_PARAM_LISTENER_LOCAL_ADDRESS`<br> 0 | QUIC_ADDR                 | Get-only  | Get the full address tuple the server is listening on.    |
 | `QUIC_PARAM_LISTENER_STATS`<br> 1         | QUIC_LISTENER_STATISTICS  | Get-only  | Get statistics specific to this Listener instance.        |
-| `QUIC_PARAM_LISTENER_CIBIR_ID`<br> 2      | uint8_t[]                 | Both      | The CIBIR well-known idenfitier.                          |
+| `QUIC_PARAM_LISTENER_CIBIR_ID`<br> 2      | uint8_t[]                 | Both      | Sets a CIBIR (CID-Based Identification and Routing) well-known identifier. CIBIR does 2 things when set: 1. XDP will now steer packets to the correct process/listener by matching the CIBIR prefix within the packet QUIC Connection ID. 2. In the case of a port collision when reserving OS UDP/TCP sockets, MsQuic will continue with initializing the datapath. If XDP is not available/enabled, then no traffic will flow for the listener that experiences a collision. |
 | `QUIC_PARAM_DOS_MODE_EVENTS`<br> 2        | BOOLEAN                   | Both      | The Listener opted in for DoS Mode event.                 |
 | `QUIC_PARAM_LISTENER_PARTITION_INDEX`<br> (preview) | uint16_t           | Both      | The partition to use for listener callback events and incoming connections. |
 

src/generated/linux/datapath_winuser.c.clog.h

@@ -159,6 +159,26 @@ tracepoint(CLOG_DATAPATH_WINUSER_C, DatapathTestSetIpv6TrafficClassFailed , arg2
 
 
 
+/*----------------------------------------------------------
+// Decoder Ring for DatapathCibirSkipNoXdp
+// [data][%p] CIBIR configured, skipping OS socket reservation but XDP not %s
+// QuicTraceLogWarning(
+                                DatapathCibirSkipNoXdp,
+                                "[data][%p] CIBIR configured, skipping OS socket reservation but XDP not %s",
+                                Socket,
+                                !XdpAvailable ? "available" : "enabled");
+// arg2 = arg2 = Socket = arg2
+// arg3 = arg3 = !XdpAvailable ? "available" : "enabled" = arg3
+----------------------------------------------------------*/
+#ifndef _clog_4_ARGS_TRACE_DatapathCibirSkipNoXdp
+#define _clog_4_ARGS_TRACE_DatapathCibirSkipNoXdp(uniqueId, encoded_arg_string, arg2, arg3)\
+tracepoint(CLOG_DATAPATH_WINUSER_C, DatapathCibirSkipNoXdp , arg2, arg3);\
+
+#endif
+
+
+
+
 /*----------------------------------------------------------
 // Decoder Ring for DatapathRecvEmpty
 // [data][%p] Dropping datagram with empty payload.

src/generated/linux/datapath_winuser.c.clog.h.lttng.h

@@ -134,6 +134,29 @@ TRACEPOINT_EVENT(CLOG_DATAPATH_WINUSER_C, DatapathTestSetIpv6TrafficClassFailed,
 
 
 
+/*----------------------------------------------------------
+// Decoder Ring for DatapathCibirSkipNoXdp
+// [data][%p] CIBIR configured, skipping OS socket reservation but XDP not %s
+// QuicTraceLogWarning(
+                                DatapathCibirSkipNoXdp,
+                                "[data][%p] CIBIR configured, skipping OS socket reservation but XDP not %s",
+                                Socket,
+                                !XdpAvailable ? "available" : "enabled");
+// arg2 = arg2 = Socket = arg2
+// arg3 = arg3 = !XdpAvailable ? "available" : "enabled" = arg3
+----------------------------------------------------------*/
+TRACEPOINT_EVENT(CLOG_DATAPATH_WINUSER_C, DatapathCibirSkipNoXdp,
+    TP_ARGS(
+        const void *, arg2,
+        const char *, arg3), 
+    TP_FIELDS(
+        ctf_integer_hex(uint64_t, arg2, (uint64_t)arg2)
+        ctf_string(arg3, arg3)
+    )
+)
+
+
+
 /*----------------------------------------------------------
 // Decoder Ring for DatapathRecvEmpty
 // [data][%p] Dropping datagram with empty payload.

src/manifest/clog.sidecar

@@ -2443,6 +2443,22 @@
       ],
       "macroName": "QuicTraceLogConnVerbose"
     },
+    "DatapathCibirSkipNoXdp": {
+      "ModuleProperites": {},
+      "TraceString": "[data][%p] CIBIR configured, skipping OS socket reservation but XDP not %s",
+      "UniqueId": "DatapathCibirSkipNoXdp",
+      "splitArgs": [
+        {
+          "DefinationEncoding": "p",
+          "MacroVariableName": "arg2"
+        },
+        {
+          "DefinationEncoding": "s",
+          "MacroVariableName": "arg3"
+        }
+      ],
+      "macroName": "QuicTraceLogWarning"
+    },
     "DatapathCreated": {
       "ModuleProperites": {},
       "TraceString": "[data][%p] Created, local=%!ADDR!, remote=%!ADDR!",
@@ -14022,6 +14038,11 @@
         "TraceID": "DatagramSendStateChanged",
         "EncodingString": "[conn][%p] Indicating DATAGRAM_SEND_STATE_CHANGED to %u"
       },
+      {
+        "UniquenessHash": "8343563a-502d-077c-44f6-6dfd65e927f5",
+        "TraceID": "DatapathCibirSkipNoXdp",
+        "EncodingString": "[data][%p] CIBIR configured, skipping OS socket reservation but XDP not %s"
+      },
       {
         "UniquenessHash": "1cde4174-4172-15b7-b59b-dcbb6410b43a",
         "TraceID": "DatapathCreated",

src/platform/datapath_raw_socket.c

@@ -890,7 +890,7 @@ CxPlatTryAddSocket(
     // binding an auxiliary (dual stack) socket.
     //
 
-    if (Socket->ReserveAuxTcpSock) {
+    if (Socket->ReserveAuxTcpSock && !Socket->SkipCreatingOsSockets) {
         Socket->AuxSocket =
             socket(
                 AF_INET6,
@@ -929,6 +929,14 @@ CxPlatTryAddSocket(
         }
 
         if (Socket->CibirIdLength) {
+            //
+            // Setting SO_REUSEADDR does NOT robustly allow
+            // multiple processes to share the same port on
+            // Windows (WinSock).
+            // This code was added primarily for Linux XDP,
+            // where Linux sockets DO actually allow robust
+            // port sharing...
+            //
             Option = TRUE;
             Result =
                 setsockopt(
@@ -960,7 +968,7 @@ CxPlatTryAddSocket(
 
     CxPlatRwLockAcquireExclusive(&Pool->Lock);
 
-    if (Socket->ReserveAuxTcpSock) {
+    if (Socket->ReserveAuxTcpSock && !Socket->SkipCreatingOsSockets) {
         QUIC_ADDR_STR LocalAddressString = {0};
         QuicAddrToString(&MappedAddress, &LocalAddressString);
         QuicTraceLogVerbose(

src/platform/datapath_winuser.c

@@ -1273,7 +1273,7 @@ SocketCreateUdp(
 
     if (Socket->ReserveAuxTcpSock && !IsServerSocket) {
         //
-        // Client will skip normal socket settings to use AuxSocket in raw socket.
+        // QTIP clients will skip normal UDP socket reservation to use AuxSocket (TCP socket reservation) in raw socket.
         //
         goto Skip;
     }
@@ -1724,6 +1724,27 @@ SocketCreateUdp(
                         Socket,
                         WsaError,
                         "SIO_ACQUIRE_PORT_RESERVATION");
+
+                    if (Config->CibirIdLength > 0 && IsServerSocket) {
+                        //
+                        // CIBIR is configured. A port collision here is actually
+                        // expected if multiple server processes were configured to share
+                        // the same UDP and/or TCP port.
+                        //
+                        CxPlatRefInitializeEx(&Socket->RefCount, 1);
+                        Socket->SkipCreatingOsSockets = TRUE;
+                        BOOLEAN XdpAvailable = Datapath->RawDataPath != NULL;
+                        BOOLEAN XdpEnabled = Config->Flags & CXPLAT_SOCKET_FLAG_XDP;
+                        if (!XdpAvailable || !XdpEnabled) {
+                            QuicTraceLogWarning(
+                                DatapathCibirSkipNoXdp,
+                                "[data][%p] CIBIR configured, skipping OS socket reservation but XDP not %s",
+                                Socket,
+                                !XdpAvailable ? "available" : "enabled");
+                        }
+                        goto Skip;
+                    }
+
                     Status = HRESULT_FROM_WIN32(WsaError);
                     goto Error;
                 }
@@ -1831,6 +1852,25 @@ SocketCreateUdp(
 
 Skip:
 
+    if (Socket->SkipCreatingOsSockets) {
+        //
+        // Clean up all partially-initialized per-proc sockets since
+        // we're skipping OS socket creation (XDP-only via CIBIR).
+        //
+        for (uint16_t i = 0; i < SocketCount; i++) {
+            CXPLAT_SOCKET_PROC* Proc = &Socket->PerProcSockets[i];
+            if (Proc->Socket != INVALID_SOCKET) {
+                closesocket(Proc->Socket);
+                Proc->Socket = INVALID_SOCKET;
+            }
+            if (Proc->DatapathProc) {
+                CxPlatProcessorContextRelease(Proc->DatapathProc);
+                Proc->DatapathProc = NULL;
+            }
+            CxPlatRundownUninitialize(&Proc->RundownRef);
+        }
+    }
+
     if (Config->RemoteAddress != NULL) {
         Socket->RemoteAddress = *Config->RemoteAddress;
     } else {
@@ -1843,7 +1883,7 @@ SocketCreateUdp(
     //
     *NewSocket = Socket;
 
-    if (!Socket->ReserveAuxTcpSock) {
+    if (!Socket->ReserveAuxTcpSock && !Socket->SkipCreatingOsSockets) {
         for (uint16_t i = 0; i < SocketCount; i++) {
             CxPlatDataPathStartReceiveAsync(&Socket->PerProcSockets[i]);
             Socket->PerProcSockets[i].IoStarted = TRUE;
@@ -2380,8 +2420,12 @@ SocketDelete(
     CXPLAT_DBG_ASSERT(!Socket->Uninitialized);
     Socket->Uninitialized = TRUE;
 
-    if (Socket->ReserveAuxTcpSock && Socket->HasFixedRemoteAddress) {
-        // QTIP did not initialize PerProcSockets only for Client sockets.
+    if ((Socket->ReserveAuxTcpSock && Socket->HasFixedRemoteAddress) ||
+        Socket->SkipCreatingOsSockets) {
+        //
+        // QTIP client or CIBIR skip: PerProcSockets were not initialized
+        // (or already cleaned up). Just release the socket directly.
+        //
         CxPlatSocketRelease(Socket);
     } else {
         const uint16_t SocketCount =

src/platform/platform_internal.h

@@ -212,6 +212,8 @@ typedef struct CXPLAT_SOCKET {
     uint8_t HasFixedRemoteAddress : 1;
     uint8_t RawSocketAvailable : 1;
 
+    uint8_t SkipCreatingOsSockets : 1;
+
     CXPLAT_RUNDOWN_REF Rundown[0]; // Per-proc
 
 } CXPLAT_SOCKET;
@@ -595,6 +597,8 @@ typedef struct CXPLAT_SOCKET {
 
     uint8_t RawSocketAvailable : 1;
 
+    uint8_t SkipCreatingOsSockets : 1;
+
     //
     // Per-processor socket contexts.
     //
@@ -900,6 +904,8 @@ typedef struct CXPLAT_SOCKET {
 
     uint8_t RawSocketAvailable : 1;
 
+    uint8_t SkipCreatingOsSockets : 1;
+
     //
     // Set of socket contexts one per proc.
     //

src/platform/unittest/DataPathTest.cpp

@@ -531,6 +531,7 @@ struct CxPlatSocket {
     CXPLAT_SOCKET* Socket {nullptr};
     QUIC_STATUS InitStatus {QUIC_STATUS_INVALID_STATE};
     CXPLAT_ROUTE Route {0};
+    uint8_t CibirIdLength {0};
     CxPlatSocket() { }
     CxPlatSocket(
         _In_ CxPlatDataPath& Datapath,
@@ -571,6 +572,7 @@ struct CxPlatSocket {
         UdpConfig.Flags = InternalFlags;
         UdpConfig.InterfaceIndex = 0;
         UdpConfig.CallbackContext = CallbackContext;
+        UdpConfig.CibirIdLength = CibirIdLength;
         InitStatus =
             CxPlatSocketCreateUdp(
                 Datapath,
@@ -848,6 +850,32 @@ TEST_P(DataPathTest, UdpData)
     ASSERT_TRUE(CxPlatEventWaitWithTimeout(RecvContext.ClientCompletion, 2000));
 }
 
+TEST_P(DataPathTest, UdpDataShareCibirUdpPort) {
+    UdpRecvContext RecvContext;
+    CxPlatDataPath Datapath(&UdpRecvCallbacks);
+    VERIFY_QUIC_SUCCESS(Datapath.GetInitStatus());
+    auto unspecAddress = GetNewUnspecAddr();
+    CxPlatSocket Server1;
+    //
+    // Set CibirIdLength to some non-zero value to trigger the cibir path in the datapath code,
+    // which should allow multiple cxplat sockets to be created on the same udp port so long as
+    // cibir is configured. 
+    //
+    Server1.CibirIdLength = 8;
+    Server1.CreateUdp(Datapath, &unspecAddress.SockAddr, nullptr, &RecvContext);
+    while (Server1.GetInitStatus() == QUIC_STATUS_ADDRESS_IN_USE) {
+        unspecAddress.SockAddr.Ipv4.sin_port = GetNextPort();
+        Server1.CreateUdp(Datapath, &unspecAddress.SockAddr, nullptr, &RecvContext);
+    }
+    VERIFY_QUIC_SUCCESS(Server1.GetInitStatus());
+    ASSERT_NE(nullptr, Server1.Socket);
+    CxPlatSocket Server2;
+    Server2.CibirIdLength = 8;
+    Server2.CreateUdp(Datapath, &unspecAddress.SockAddr, nullptr, &RecvContext);
+    VERIFY_QUIC_SUCCESS(Server2.GetInitStatus());
+    ASSERT_NE(nullptr, Server2.Socket);
+}
+
 TEST_P(DataPathTest, UdpDataPolling)
 {
     QUIC_GLOBAL_EXECUTION_CONFIG Config = { QUIC_GLOBAL_EXECUTION_CONFIG_FLAG_NONE, UINT32_MAX, 0 };