[ovs-dev] [PATCH] datapath-windows: Support for OVS_KEY_ATTR_MPLS attribute

Sorin Vinturis svinturis at cloudbasesolutions.com
Wed Nov 11 20:11:03 UTC 2015


This patch adds OVS_KEY_ATTR_MPLS to the OVS flow mechanism.

Signed-off-by: Sorin Vinturis <svinturis at cloudbasesolutions.com>
---
 datapath-windows/ovsext/Actions.c      | 176 +++++++++++++++++++++++++++++++++
 datapath-windows/ovsext/DpInternal.h   |   7 ++
 datapath-windows/ovsext/Ethernet.h     |   2 +
 datapath-windows/ovsext/Flow.c         |  89 ++++++++++++++++-
 datapath-windows/ovsext/NetProto.h     |  33 +++++++
 datapath-windows/ovsext/PacketParser.c |  12 +--
 datapath-windows/ovsext/PacketParser.h |   7 ++
 7 files changed, 318 insertions(+), 8 deletions(-)

diff --git a/datapath-windows/ovsext/Actions.c b/datapath-windows/ovsext/Actions.c
index ce592b3..9ee1763 100644
--- a/datapath-windows/ovsext/Actions.c
+++ b/datapath-windows/ovsext/Actions.c
@@ -1074,6 +1074,142 @@ OvsPopVlanInPktBuf(OvsForwardingContext *ovsFwdCtx)
     return NDIS_STATUS_SUCCESS;
 }
 
+static __inline NDIS_STATUS
+OvsActionMplsPush(OvsForwardingContext *ovsFwdCtx,
+                  const struct ovs_action_push_mpls *mpls)
+{
+    NDIS_STATUS status;
+    PNET_BUFFER curNb = NULL;
+    PMDL curMdl = NULL;
+    PUINT8 bufferStart = NULL;
+    OVS_PACKET_HDR_INFO *layers = &ovsFwdCtx->layers;
+    EthHdr *ethHdr = NULL;
+    MPLSHdr *mplsHdr = NULL;
+    UINT32 mdlLen = 0, curMdlOffset = 0;
+    UINT32 packetLen = 0;
+    PNET_BUFFER_LIST newNbl;
+
+    curNb = NET_BUFFER_LIST_FIRST_NB(ovsFwdCtx->curNbl);
+    ASSERT(curNb->Next == NULL);
+    packetLen = NET_BUFFER_DATA_LENGTH(curNb);
+    curMdl = NET_BUFFER_CURRENT_MDL(curNb);
+    NdisQueryMdl(curMdl, &bufferStart, &mdlLen, LowPagePriority);
+    if (!bufferStart) {
+        ovsActionStats.noResource++;
+        return NDIS_STATUS_RESOURCES;
+    }
+    curMdlOffset = NET_BUFFER_CURRENT_MDL_OFFSET(curNb);
+    mdlLen -= curMdlOffset;
+    mdlLen -= NET_BUFFER_CURRENT_MDL_OFFSET(curNb);
+    ASSERT(mdlLen > 0);
+    /* Bail out if the L2 header is not in a contiguous buffer. */
+    if (MIN(packetLen, mdlLen) < sizeof *ethHdr) {
+        ASSERT(FALSE);
+        return NDIS_STATUS_FAILURE;
+    }
+    ASSERT((INT)mdlLen >= 0);
+
+    newNbl = OvsPartialCopyNBL(ovsFwdCtx->switchContext, ovsFwdCtx->curNbl,
+                               MPLS_HLEN, 0, TRUE /*copy NBL info*/);
+    if (!newNbl) {
+        ovsActionStats.noCopiedNbl++;
+        return NDIS_STATUS_RESOURCES;
+    }
+    OvsCompleteNBLForwardingCtx(ovsFwdCtx,
+                                L"Complete after partial copy.");
+
+    status = OvsInitForwardingCtx(ovsFwdCtx, ovsFwdCtx->switchContext,
+                                  newNbl, ovsFwdCtx->srcVportNo, 0,
+                                  NET_BUFFER_LIST_SWITCH_FORWARDING_DETAIL(newNbl),
+                                  NULL, &ovsFwdCtx->layers, FALSE);
+    if (status != NDIS_STATUS_SUCCESS) {
+        OvsCompleteNBLForwardingCtx(ovsFwdCtx,
+                                    L"OVS-Dropped due to resources");
+        return NDIS_STATUS_RESOURCES;
+    }
+
+    curNb = NET_BUFFER_LIST_FIRST_NB(ovsFwdCtx->curNbl);
+    ASSERT(curNb->Next == NULL);
+    curMdl = NET_BUFFER_CURRENT_MDL(curNb);
+    NdisQueryMdl(curMdl, &bufferStart, &mdlLen, LowPagePriority);
+    if (!curMdl) {
+        ovsActionStats.noResource++;
+        return NDIS_STATUS_RESOURCES;
+    }
+    curMdlOffset = NET_BUFFER_CURRENT_MDL_OFFSET(curNb);
+    mdlLen -= curMdlOffset;
+    ASSERT(mdlLen >= MPLS_HLEN);
+
+    ethHdr = (EthHdr *)(bufferStart + curMdlOffset);
+    ethHdr->Type = mpls->mpls_ethertype;
+
+    mplsHdr = (MPLSHdr *)(bufferStart + curMdlOffset + MPLS_HLEN);
+    mplsHdr->mpls_lse = mpls->mpls_lse;
+
+    layers->l3Offset += MPLS_HLEN;
+    layers->isIPv4 = 0;
+    layers->isIPv6 = 0;
+
+    NdisRetreatNetBufferDataStart(curNb, MPLS_HLEN, FALSE, NULL);
+
+    return NDIS_STATUS_SUCCESS;
+}
+
+static __inline NDIS_STATUS
+OvsActionMplsPop(OvsForwardingContext *ovsFwdCtx)
+{
+    PNET_BUFFER curNb;
+    PMDL curMdl;
+    PUINT8 bufferStart;
+    ULONG dataLength = sizeof (DL_EUI48) + sizeof (DL_EUI48);
+    UINT32 packetLen, mdlLen;
+    PNET_BUFFER_LIST newNbl;
+    NDIS_STATUS status;
+
+    PUINT8 tempBuffer[sizeof (DL_EUI48) + sizeof (DL_EUI48)];
+
+    newNbl = OvsPartialCopyNBL(ovsFwdCtx->switchContext, ovsFwdCtx->curNbl,
+                               0, 0, TRUE /* copy NBL info */);
+    if (!newNbl) {
+        ovsActionStats.noCopiedNbl++;
+        return NDIS_STATUS_RESOURCES;
+    }
+
+    /* Complete the original NBL and create a copy to modify. */
+    OvsCompleteNBLForwardingCtx(ovsFwdCtx, L"OVS-Dropped due to copy");
+
+    status = OvsInitForwardingCtx(ovsFwdCtx, ovsFwdCtx->switchContext,
+                                  newNbl, ovsFwdCtx->srcVportNo, 0,
+                                  NET_BUFFER_LIST_SWITCH_FORWARDING_DETAIL(newNbl),
+                                  NULL, &ovsFwdCtx->layers, FALSE);
+    if (status != NDIS_STATUS_SUCCESS) {
+        OvsCompleteNBLForwardingCtx(ovsFwdCtx,
+                                    L"Dropped due to resouces");
+        return NDIS_STATUS_RESOURCES;
+    }
+
+    curNb = NET_BUFFER_LIST_FIRST_NB(ovsFwdCtx->curNbl);
+    packetLen = NET_BUFFER_DATA_LENGTH(curNb);
+    ASSERT(curNb->Next == NULL);
+    curMdl = NET_BUFFER_CURRENT_MDL(curNb);
+    NdisQueryMdl(curMdl, &bufferStart, &mdlLen, LowPagePriority);
+    if (!bufferStart) {
+        return NDIS_STATUS_RESOURCES;
+    }
+    mdlLen -= NET_BUFFER_CURRENT_MDL_OFFSET(curNb);
+    /* Bail out if L2 + MPLS header is not contiguous in the first buffer. */
+    if (MIN(packetLen, mdlLen) < sizeof(EthHdr) + MPLS_HLEN) {
+        ASSERT(FALSE);
+        return NDIS_STATUS_FAILURE;
+    }
+    bufferStart += NET_BUFFER_CURRENT_MDL_OFFSET(curNb);
+    RtlCopyMemory(tempBuffer, bufferStart, dataLength);
+    RtlCopyMemory(bufferStart + MPLS_HLEN, tempBuffer, dataLength);
+    NdisAdvanceNetBufferDataStart(curNb, MPLS_HLEN, FALSE, NULL);
+
+    return NDIS_STATUS_SUCCESS;
+}
+
 /*
  * --------------------------------------------------------------------------
  * OvsTunnelAttrToIPv4TunnelKey --
@@ -1513,6 +1649,46 @@ OvsActionsExecute(POVS_SWITCH_CONTEXT switchContext,
             break;
         }
 
+        case OVS_ACTION_ATTR_PUSH_MPLS:
+        {
+            if (ovsFwdCtx.destPortsSizeOut > 0 || ovsFwdCtx.tunnelTxNic != NULL
+                || ovsFwdCtx.tunnelRxNic != NULL) {
+                status = OvsOutputBeforeSetAction(&ovsFwdCtx);
+                if (status != NDIS_STATUS_SUCCESS) {
+                    dropReason = L"OVS-adding destination failed";
+                    goto dropit;
+                }
+            }
+
+            status = OvsActionMplsPush(&ovsFwdCtx,
+                                       (struct ovs_action_push_mpls *)NlAttrGet
+                                       ((const PNL_ATTR)a));
+            if (status != NDIS_STATUS_SUCCESS) {
+                dropReason = L"OVS-set push MPLS failed";
+                goto dropit;
+            }
+            break;
+        }
+
+        case OVS_ACTION_ATTR_POP_MPLS:
+        {
+            if (ovsFwdCtx.destPortsSizeOut > 0 || ovsFwdCtx.tunnelTxNic != NULL
+                || ovsFwdCtx.tunnelRxNic != NULL) {
+                status = OvsOutputBeforeSetAction(&ovsFwdCtx);
+                if (status != NDIS_STATUS_SUCCESS) {
+                    dropReason = L"OVS-adding destination failed";
+                    goto dropit;
+                }
+            }
+
+            status = OvsActionMplsPop(&ovsFwdCtx);
+            if (status != NDIS_STATUS_SUCCESS) {
+                dropReason = L"OVS-set pop MPLS failed";
+                goto dropit;
+            }
+            break;
+        }
+
         case OVS_ACTION_ATTR_USERSPACE:
         {
             PNL_ATTR userdataAttr;
diff --git a/datapath-windows/ovsext/DpInternal.h b/datapath-windows/ovsext/DpInternal.h
index 8de48a2..c195494 100644
--- a/datapath-windows/ovsext/DpInternal.h
+++ b/datapath-windows/ovsext/DpInternal.h
@@ -20,6 +20,7 @@
 #include <netioapi.h>
 #define IFNAMSIZ IF_NAMESIZE
 #include "../ovsext/Netlink/Netlink.h"
+#include "NetProto.h"
 
 #define OVS_DP_NUMBER   ((uint32_t) 0)
 
@@ -149,6 +150,11 @@ typedef union OvsIPv4TunnelKey {
     uint64_t attr[NUM_PKT_ATTR_REQUIRED];
 } OvsIPv4TunnelKey;
 
+typedef struct MplsKey {
+    ovs_be32 top_lse;            /* MPLS topmost label stack entry. */
+    uint8    pad[4];
+} MplsKey; /* Size of 8 bytes. */
+
 typedef __declspec(align(8)) struct OvsFlowKey {
     OvsIPv4TunnelKey tunKey;     /* 24 bytes */
     L2Key l2;                    /* 24 bytes */
@@ -157,6 +163,7 @@ typedef __declspec(align(8)) struct OvsFlowKey {
         ArpKey arpKey;           /* size 24 */
         Ipv6Key ipv6Key;         /* size 48 */
         Icmp6Key icmp6Key;       /* size 72 */
+        MplsKey mplsKey;         /* size 8 */
     };
 } OvsFlowKey;
 
diff --git a/datapath-windows/ovsext/Ethernet.h b/datapath-windows/ovsext/Ethernet.h
index 22aa27c..1d69d47 100644
--- a/datapath-windows/ovsext/Ethernet.h
+++ b/datapath-windows/ovsext/Ethernet.h
@@ -66,6 +66,8 @@ typedef enum {
     ETH_TYPE_CDP         = 0x2000,
     ETH_TYPE_802_1PQ     = 0x8100, // not really a DIX type, but used as such
     ETH_TYPE_LLC         = 0xFFFF, // 0xFFFF is IANA reserved, used to mark LLC
+    ETH_TYPE_MPLS        = 0x8847,
+    ETH_TYPE_MPLS_MCAST  = 0x8848,
 } Eth_DixType;
 
 typedef enum {
diff --git a/datapath-windows/ovsext/Flow.c b/datapath-windows/ovsext/Flow.c
index b629c93..c989c14 100644
--- a/datapath-windows/ovsext/Flow.c
+++ b/datapath-windows/ovsext/Flow.c
@@ -80,6 +80,8 @@ static NTSTATUS _MapFlowIpv6KeyToNlKey(PNL_BUFFER nlBuf,
                                        Icmp6Key *ipv6FlowPutIcmpKey);
 static NTSTATUS _MapFlowArpKeyToNlKey(PNL_BUFFER nlBuf,
                                       ArpKey *arpFlowPutKey);
+static NTSTATUS _MapFlowMplsKeyToNlKey(PNL_BUFFER nlBuf,
+                                       MplsKey *mplsFlowPutKey);
 
 static NTSTATUS OvsDoDumpFlows(OvsFlowDumpInput *dumpInput,
                                OvsFlowDumpOutput *dumpOutput,
@@ -108,7 +110,7 @@ const NL_POLICY nlFlowPolicy[] = {
 
 /* For Parsing nested OVS_FLOW_ATTR_KEY attributes.
  * Some of the attributes like OVS_KEY_ATTR_RECIRC_ID
- * & OVS_KEY_ATTR_MPLS are not supported yet. */
+ * are not supported yet. */
 
 const NL_POLICY nlFlowKeyPolicy[] = {
     [OVS_KEY_ATTR_ENCAP] = {.type = NL_A_VAR_LEN, .optional = TRUE},
@@ -872,6 +874,13 @@ MapFlowKeyToNlKey(PNL_BUFFER nlBuf,
         break;
         }
 
+        case ETH_TYPE_MPLS:
+        case ETH_TYPE_MPLS_MCAST: {
+        MplsKey *mplsFlowPutKey = &(flowKey->mplsKey);
+        rc = _MapFlowMplsKeyToNlKey(nlBuf, mplsFlowPutKey);
+        break;
+        }
+
         default:
         break;
     }
@@ -1194,6 +1203,31 @@ done:
 
 /*
  *----------------------------------------------------------------------------
+ *  _MapFlowMplsKeyToNlKey --
+ *    Maps _MapFlowMplsKeyToNlKey to OVS_KEY_ATTR_MPLS attribute.
+ *----------------------------------------------------------------------------
+ */
+static NTSTATUS
+_MapFlowMplsKeyToNlKey(PNL_BUFFER nlBuf, MplsKey *mplsFlowPutKey)
+{
+    NTSTATUS rc = STATUS_SUCCESS;
+    struct ovs_key_mpls *mplsKey;
+
+    mplsKey = (struct ovs_key_mpls *)
+        NlMsgPutTailUnspecUninit(nlBuf, OVS_KEY_ATTR_MPLS, sizeof(*mplsKey));
+    if (!mplsKey) {
+        rc = STATUS_UNSUCCESSFUL;
+        goto done;
+    }
+
+    mplsKey->mpls_lse = mplsFlowPutKey->top_lse;
+
+done:
+    return rc;
+}
+
+/*
+ *----------------------------------------------------------------------------
  *  _MapNlToFlowPut --
  *    Maps input netlink message to OvsFlowPut.
  *----------------------------------------------------------------------------
@@ -1469,8 +1503,28 @@ _MapKeyAttrToFlowPut(PNL_ATTR *keyAttrs,
             arpFlowPutKey->pad[1] = 0;
             arpFlowPutKey->pad[2] = 0;
             destKey->l2.keyLen += OVS_ARP_KEY_SIZE;
-            break;
         }
+        break;
+    }
+    case ETH_TYPE_MPLS:
+    case ETH_TYPE_MPLS_MCAST: {
+
+        if (keyAttrs[OVS_KEY_ATTR_MPLS]) {
+            MplsKey *mplsFlowPutKey = &destKey->mplsKey;
+            const struct ovs_key_mpls *mplsKey;
+            UINT32 size = NlAttrGetSize(keyAttrs[OVS_KEY_ATTR_MPLS]);
+            UINT32 n = size / sizeof(struct ovs_key_mpls);
+
+            mplsKey = NlAttrGet(keyAttrs[OVS_KEY_ATTR_MPLS]);
+
+            mplsFlowPutKey->top_lse = mplsKey->mpls_lse;
+            mplsFlowPutKey->pad[0] = 0;
+            mplsFlowPutKey->pad[1] = 0;
+            mplsFlowPutKey->pad[2] = 0;
+            mplsFlowPutKey->pad[3] = 0;
+            destKey->l2.keyLen += (UINT16)n * sizeof(MplsKey);
+        }
+        break;
     }
     }
 }
@@ -1864,6 +1918,37 @@ OvsExtractFlow(const NET_BUFFER_LIST *packet,
                 memcpy(arpKey->arpTha, arp->arp_tha, ETH_ADDR_LENGTH);
             }
         }
+    } else if (flow->l2.dlType == htons(ETH_TYPE_MPLS) ||
+               flow->l2.dlType == htons(ETH_TYPE_MPLS_MCAST)) {
+        MPLSHdr mplsStorage;
+        const MPLSHdr *mpls;
+        MplsKey *mplsKey = &flow->mplsKey;
+        ((UINT64 *)mplsKey)[0] = 0;
+
+        /* In the presence of an MPLS label stack the end of the L2
+         * header and the beginning of the L3 header differ.
+         *
+         * A network packet may contain multiple MPLS labels, but we
+         * are only interested in the topmost label stack entry.
+         *
+         * Advance network header to the beginning of the L3 header.
+         * layers->l3Offset corresponds to the end of the L2 header.
+         */
+        for (UINT32 i = 0; i < FLOW_MAX_MPLS_LABELS; i++) {
+            mpls = OvsGetMpls(packet, layers->l3Offset, &mplsStorage);
+            if (mpls) {
+
+                /* Keep only the topmost MPLS label stack entry. */
+                if (i == 0) {
+                    mplsKey->top_lse = mpls->mpls_lse;
+                }
+
+                layers->l3Offset += sizeof(MPLSHdr);
+
+                if (mpls->mpls_lse & htonl(MPLS_LS_S_MASK))
+                    break;
+            }
+        }
     }
 
     return NDIS_STATUS_SUCCESS;
diff --git a/datapath-windows/ovsext/NetProto.h b/datapath-windows/ovsext/NetProto.h
index a364869..12e8bd9 100644
--- a/datapath-windows/ovsext/NetProto.h
+++ b/datapath-windows/ovsext/NetProto.h
@@ -366,4 +366,37 @@ typedef struct IPOpt {
 #define SOCKET_IPPROTO_UDP   17
 #define SOCKET_IPPROTO_GRE   47
 
+/* Reference: RFC 5462, RFC 3032
+ *
+ *  0                   1                   2                   3
+ *  0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+ * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ * |                Label                  | TC  |S|       TTL     |
+ * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *
+ *	Label:  Label Value, 20 bits
+ *	TC:     Traffic Class field, 3 bits
+ *	S:      Bottom of Stack, 1 bit
+ *	TTL:    Time to Live, 8 bits
+ */
+
+typedef struct MPLSHdr {
+    ovs_be32 mpls_lse;
+} MPLSHdr;
+
+/*
+ * MPLS definitions
+ */
+#define MPLS_HLEN               4
+#define FLOW_MAX_MPLS_LABELS    3
+#define MPLS_LS_LABEL_MASK      0xFFFFF000
+#define MPLS_LS_LABEL_SHIFT     12
+#define MPLS_LS_TC_MASK         0x00000E00
+#define MPLS_LS_TC_SHIFT        9
+#define MPLS_LS_S_MASK          0x00000100
+#define MPLS_LS_S_SHIFT         8
+#define MPLS_LS_TTL_MASK        0x000000FF
+#define MPLS_LS_TTL_SHIFT       0
+
+
 #endif /* __NET_PROTO_H_ */
diff --git a/datapath-windows/ovsext/PacketParser.c b/datapath-windows/ovsext/PacketParser.c
index e01be17..246c603 100644
--- a/datapath-windows/ovsext/PacketParser.c
+++ b/datapath-windows/ovsext/PacketParser.c
@@ -84,8 +84,8 @@ OvsGetPacketBytes(const NET_BUFFER_LIST *nbl,
 
 NDIS_STATUS
 OvsParseIPv6(const NET_BUFFER_LIST *packet,
-          OvsFlowKey *key,
-          POVS_PACKET_HDR_INFO layers)
+             OvsFlowKey *key,
+             POVS_PACKET_HDR_INFO layers)
 {
     UINT16 ofs = layers->l3Offset;
     IPv6Hdr ipv6HdrStorage;
@@ -178,8 +178,8 @@ OvsParseIPv6(const NET_BUFFER_LIST *packet,
 
 VOID
 OvsParseTcp(const NET_BUFFER_LIST *packet,
-         L4Key *flow,
-         POVS_PACKET_HDR_INFO layers)
+            L4Key *flow,
+            POVS_PACKET_HDR_INFO layers)
 {
     TCPHdr tcpStorage;
     const TCPHdr *tcp = OvsGetTcp(packet, layers->l4Offset, &tcpStorage);
@@ -193,8 +193,8 @@ OvsParseTcp(const NET_BUFFER_LIST *packet,
 
 VOID
 OvsParseUdp(const NET_BUFFER_LIST *packet,
-         L4Key *flow,
-         POVS_PACKET_HDR_INFO layers)
+            L4Key *flow,
+            POVS_PACKET_HDR_INFO layers)
 {
     UDPHdr udpStorage;
     const UDPHdr *udp = OvsGetUdp(packet, layers->l4Offset, &udpStorage);
diff --git a/datapath-windows/ovsext/PacketParser.h b/datapath-windows/ovsext/PacketParser.h
index 55d110f..96136b7 100644
--- a/datapath-windows/ovsext/PacketParser.h
+++ b/datapath-windows/ovsext/PacketParser.h
@@ -141,4 +141,11 @@ OvsGetIcmp(const NET_BUFFER_LIST *packet,
     return OvsGetPacketBytes(packet, sizeof *storage, ofs, storage);
 }
 
+static const MPLSHdr *
+OvsGetMpls(const NET_BUFFER_LIST *packet,
+           UINT32 ofs,
+           MPLSHdr *storage)
+{
+    return OvsGetPacketBytes(packet, sizeof *storage, ofs, storage);
+}
 #endif /* __PACKET_PARSER_H_ */
-- 
1.9.0.msysgit.0



More information about the dev mailing list