[ovs-dev] [PATCH V6] datapath-windows: Improved offloading on STT tunnel

Paul Boca pboca at cloudbasesolutions.com
Thu May 12 09:22:04 UTC 2016


*Added OvsExtractLayers - populates only the layers field without unnecessary
memory operations for flow part
*If in STT header the flags are 0 then force packets checksums calculation
on receive.
*Ensure correct pseudo checksum is set for LSO both on send and receive.
Linux includes the segment length to TCP pseudo-checksum conforming to
RFC 793 but in case of LSO Windows expects this to be only on
Source IP Address, Destination IP Address, and Protocol.
*Fragment expiration on rx side of STT was set to 30 seconds, but the correct
timeout would be TTL of the packet

Signed-off-by: Paul-Daniel Boca <pboca at cloudbasesolutions.com>
Reviewed-by: Alin Gabriel Serdean <aserdean at cloudbasesolutions.com>
---
v2: Fixed a NULL pointer dereference.
      Removed some unused local variables and multiple initializations.
v3: Use LSO V2 in OvsDoEncapStt
      Fixed alignment and code style
      Use IpHdr TTL for fragment expiration on receive instead 30s
V4: Use stored MSS in STT header on rx for lsoInfo of encapsulated packet
      If STT_CSUM_VERIFIED flag is set then we don't have to extract
      layers on receive.
V5: If CSUM_VERIFIED or no flag is set in STT header then don't recompute
      checksums
V6: Add define for conversion of TTL to seconds
      Fixes LSO MSS on rx side
      Compute TCP checksum only once on TX side over STT header
---
 datapath-windows/ovsext/Flow.c         | 243 ++++++++++++++++++++++++++++-----
 datapath-windows/ovsext/Flow.h         |   2 +
 datapath-windows/ovsext/IpHelper.h     |   3 +-
 datapath-windows/ovsext/PacketParser.c |  97 +++++++------
 datapath-windows/ovsext/PacketParser.h |   8 +-
 datapath-windows/ovsext/Stt.c          | 127 +++++++++++++----
 datapath-windows/ovsext/Stt.h          |   1 -
 datapath-windows/ovsext/User.c         |  17 ++-
 8 files changed, 381 insertions(+), 117 deletions(-)

diff --git a/datapath-windows/ovsext/Flow.c b/datapath-windows/ovsext/Flow.c
index 1f23625..a49a60c 100644
--- a/datapath-windows/ovsext/Flow.c
+++ b/datapath-windows/ovsext/Flow.c
@@ -1566,7 +1566,8 @@ _MapKeyAttrToFlowPut(PNL_ATTR *keyAttrs,
 
                     ndKey = NlAttrGet(keyAttrs[OVS_KEY_ATTR_ND]);
                     RtlCopyMemory(&icmp6FlowPutKey->ndTarget,
-                                  ndKey->nd_target, sizeof (icmp6FlowPutKey->ndTarget));
+                                  ndKey->nd_target,
+                                  sizeof (icmp6FlowPutKey->ndTarget));
                     RtlCopyMemory(icmp6FlowPutKey->arpSha,
                                   ndKey->nd_sll, ETH_ADDR_LEN);
                     RtlCopyMemory(icmp6FlowPutKey->arpTha,
@@ -1596,8 +1597,10 @@ _MapKeyAttrToFlowPut(PNL_ATTR *keyAttrs,
             arpFlowPutKey->nwSrc = arpKey->arp_sip;
             arpFlowPutKey->nwDst = arpKey->arp_tip;
 
-            RtlCopyMemory(arpFlowPutKey->arpSha, arpKey->arp_sha, ETH_ADDR_LEN);
-            RtlCopyMemory(arpFlowPutKey->arpTha, arpKey->arp_tha, ETH_ADDR_LEN);
+            RtlCopyMemory(arpFlowPutKey->arpSha, arpKey->arp_sha,
+                          ETH_ADDR_LEN);
+            RtlCopyMemory(arpFlowPutKey->arpTha, arpKey->arp_tha,
+                          ETH_ADDR_LEN);
             /* Kernel datapath assumes 'arpFlowPutKey->nwProto' to be in host
              * order. */
             arpFlowPutKey->nwProto = (UINT8)ntohs((arpKey->arp_op));
@@ -1846,29 +1849,195 @@ OvsGetFlowMetadata(OvsFlowKey *key,
     return status;
 }
 
+
 /*
- *----------------------------------------------------------------------------
- * Initializes 'flow' members from 'packet', 'skb_priority', 'tun_id', and
- * 'ofp_in_port'.
- *
- * Initializes 'packet' header pointers as follows:
- *
- *    - packet->l2 to the start of the Ethernet header.
- *
- *    - packet->l3 to just past the Ethernet header, or just past the
- *      vlan_header if one is present, to the first byte of the payload of the
- *      Ethernet frame.
- *
- *    - packet->l4 to just past the IPv4 header, if one is present and has a
- *      correct length, and otherwise NULL.
- *
- *    - packet->l7 to just past the TCP, UDP, SCTP or ICMP header, if one is
- *      present and has a correct length, and otherwise NULL.
- *
- * Returns NDIS_STATUS_SUCCESS normally.  Fails only if packet data cannot be accessed
- * (e.g. if Pkt_CopyBytesOut() returns an error).
- *----------------------------------------------------------------------------
- */
+*----------------------------------------------------------------------------
+* Initializes 'layers' members from 'packet'
+*
+* Initializes 'layers' header pointers as follows:
+*
+*    - layers->l2 to the start of the Ethernet header.
+*
+*    - layers->l3 to just past the Ethernet header, or just past the
+*      vlan_header if one is present, to the first byte of the payload of the
+*      Ethernet frame.
+*
+*    - layers->l4 to just past the IPv4 header, if one is present and has a
+*      correct length, and otherwise NULL.
+*
+*    - layers->l7 to just past the TCP, UDP, SCTP or ICMP header, if one is
+*      present and has a correct length, and otherwise NULL.
+*
+*    - layers->isIPv4/isIPv6/isTcp/isUdp/isSctp based on the packet type
+*
+* Returns NDIS_STATUS_SUCCESS normally.
+* Fails only if packet data cannot be accessed.
+* (e.g. if OvsParseIPv6() returns an error).
+*----------------------------------------------------------------------------
+*/
+NDIS_STATUS
+OvsExtractLayers(const NET_BUFFER_LIST *packet,
+                 POVS_PACKET_HDR_INFO layers)
+{
+    struct Eth_Header *eth;
+    UINT8 offset = 0;
+    PVOID vlanTagValue;
+    ovs_be16 dlType;
+
+    layers->value = 0;
+
+    /* Link layer. */
+    eth = (Eth_Header *)GetStartAddrNBL((NET_BUFFER_LIST *)packet);
+
+    /*
+    * vlan_tci.
+    */
+    vlanTagValue = NET_BUFFER_LIST_INFO(packet, Ieee8021QNetBufferListInfo);
+    if (!vlanTagValue) {
+        if (eth->dix.typeNBO == ETH_TYPE_802_1PQ_NBO) {
+            offset = sizeof(Eth_802_1pq_Tag);
+        }
+
+        /*
+        * XXX Please note after this point, src mac and dst mac should
+        * not be accessed through eth
+        */
+        eth = (Eth_Header *)((UINT8 *)eth + offset);
+    }
+
+    /*
+    * dl_type.
+    *
+    * XXX assume that at least the first
+    * 12 bytes of received packets are mapped.  This code has the stronger
+    * assumption that at least the first 22 bytes of 'packet' is mapped (if my
+    * arithmetic is right).
+    */
+    if (ETH_TYPENOT8023(eth->dix.typeNBO)) {
+        dlType = eth->dix.typeNBO;
+        layers->l3Offset = ETH_HEADER_LEN_DIX + offset;
+    } else if (OvsPacketLenNBL(packet) >= ETH_HEADER_LEN_802_3 &&
+               eth->e802_3.llc.dsap == 0xaa &&
+               eth->e802_3.llc.ssap == 0xaa &&
+               eth->e802_3.llc.control == ETH_LLC_CONTROL_UFRAME &&
+               eth->e802_3.snap.snapOrg[0] == 0x00 &&
+               eth->e802_3.snap.snapOrg[1] == 0x00 &&
+               eth->e802_3.snap.snapOrg[2] == 0x00) {
+        dlType = eth->e802_3.snap.snapType.typeNBO;
+        layers->l3Offset = ETH_HEADER_LEN_802_3 + offset;
+    } else {
+        dlType = htons(OVSWIN_DL_TYPE_NONE);
+        layers->l3Offset = ETH_HEADER_LEN_DIX + offset;
+    }
+
+    /* Network layer. */
+    if (dlType == htons(ETH_TYPE_IPV4)) {
+        struct IPHdr ip_storage;
+        const struct IPHdr *nh;
+
+        layers->isIPv4 = 1;
+        nh = OvsGetIp(packet, layers->l3Offset, &ip_storage);
+        if (nh) {
+            layers->l4Offset = layers->l3Offset + nh->ihl * 4;
+
+            if (!(nh->frag_off & htons(IP_OFFSET))) {
+                if (nh->protocol == SOCKET_IPPROTO_TCP) {
+                    OvsParseTcp(packet, NULL, layers);
+                } else if (nh->protocol == SOCKET_IPPROTO_UDP) {
+                    OvsParseUdp(packet, NULL, layers);
+                } else if (nh->protocol == SOCKET_IPPROTO_SCTP) {
+                    OvsParseSctp(packet, NULL, layers);
+                } else if (nh->protocol == SOCKET_IPPROTO_ICMP) {
+                    ICMPHdr icmpStorage;
+                    const ICMPHdr *icmp;
+
+                    icmp = OvsGetIcmp(packet, layers->l4Offset, &icmpStorage);
+                    if (icmp) {
+                        layers->l7Offset = layers->l4Offset + sizeof *icmp;
+                    }
+                }
+            }
+        }
+    } else if (dlType == htons(ETH_TYPE_IPV6)) {
+        NDIS_STATUS status;
+        Ipv6Key ipv6Key;
+
+        status = OvsParseIPv6(packet, &ipv6Key, layers);
+        if (status != NDIS_STATUS_SUCCESS) {
+            return status;
+        }
+        layers->isIPv6 = 1;
+
+        if (ipv6Key.nwProto == SOCKET_IPPROTO_TCP) {
+            OvsParseTcp(packet, &(ipv6Key.l4), layers);
+        } else if (ipv6Key.nwProto == SOCKET_IPPROTO_UDP) {
+            OvsParseUdp(packet, &(ipv6Key.l4), layers);
+        } else if (ipv6Key.nwProto == SOCKET_IPPROTO_SCTP) {
+            OvsParseSctp(packet, &ipv6Key.l4, layers);
+        } else if (ipv6Key.nwProto == SOCKET_IPPROTO_ICMPV6) {
+            Icmp6Key icmp6Key;
+            OvsParseIcmpV6(packet, NULL, &icmp6Key, layers);
+        }
+    } else if (OvsEthertypeIsMpls(dlType)) {
+        MPLSHdr mplsStorage;
+        const MPLSHdr *mpls;
+
+        /*
+        * 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) {
+                break;
+            }
+
+            layers->l3Offset += MPLS_HLEN;
+            layers->l4Offset += MPLS_HLEN;
+
+            if (mpls->lse & htonl(MPLS_BOS_MASK)) {
+                /*
+                * Bottom of Stack bit is set, which means there are no
+                * remaining MPLS labels in the packet.
+                */
+                break;
+            }
+        }
+    }
+
+    return NDIS_STATUS_SUCCESS;
+}
+
+/*
+*----------------------------------------------------------------------------
+* Initializes 'flow' members from 'packet', 'skb_priority', 'tun_id', and
+* 'ofp_in_port'.
+*
+* Initializes 'packet' header pointers as follows:
+*
+*    - packet->l2 to the start of the Ethernet header.
+*
+*    - packet->l3 to just past the Ethernet header, or just past the
+*      vlan_header if one is present, to the first byte of the payload of the
+*      Ethernet frame.
+*
+*    - packet->l4 to just past the IPv4 header, if one is present and has a
+*      correct length, and otherwise NULL.
+*
+*    - packet->l7 to just past the TCP, UDP, SCTP or ICMP header, if one is
+*      present and has a correct length, and otherwise NULL.
+*
+* Returns NDIS_STATUS_SUCCESS normally.
+* Fails only if packet data cannot be accessed.
+* (e.g. if Pkt_CopyBytesOut() returns an error).
+*----------------------------------------------------------------------------
+*/
 NDIS_STATUS
 OvsExtractFlow(const NET_BUFFER_LIST *packet,
                UINT32 inPort,
@@ -1900,8 +2069,8 @@ OvsExtractFlow(const NET_BUFFER_LIST *packet,
 
     /* Link layer. */
     eth = (Eth_Header *)GetStartAddrNBL((NET_BUFFER_LIST *)packet);
-    memcpy(flow->l2.dlSrc, eth->src, ETH_ADDR_LENGTH);
-    memcpy(flow->l2.dlDst, eth->dst, ETH_ADDR_LENGTH);
+    RtlCopyMemory(flow->l2.dlSrc, eth->src, ETH_ADDR_LENGTH);
+    RtlCopyMemory(flow->l2.dlDst, eth->dst, ETH_ADDR_LENGTH);
 
     /*
      * vlan_tci.
@@ -1923,8 +2092,7 @@ OvsExtractFlow(const NET_BUFFER_LIST *packet,
             flow->l2.vlanTci = 0;
         }
         /*
-         * XXX
-         * Please note after this point, src mac and dst mac should
+         * XXX Please note after this point, src mac and dst mac should
          * not be accessed through eth
          */
         eth = (Eth_Header *)((UINT8 *)eth + offset);
@@ -1955,7 +2123,8 @@ OvsExtractFlow(const NET_BUFFER_LIST *packet,
         layers->l3Offset = ETH_HEADER_LEN_DIX + offset;
     }
 
-    flow->l2.keyLen = OVS_WIN_TUNNEL_KEY_SIZE + OVS_L2_KEY_SIZE - flow->l2.offset;
+    flow->l2.keyLen = OVS_WIN_TUNNEL_KEY_SIZE + OVS_L2_KEY_SIZE 
+                      - flow->l2.offset;
     /* Network layer. */
     if (flow->l2.dlType == htons(ETH_TYPE_IPV4)) {
         struct IPHdr ip_storage;
@@ -2012,9 +2181,9 @@ OvsExtractFlow(const NET_BUFFER_LIST *packet,
     } else if (flow->l2.dlType == htons(ETH_TYPE_IPV6)) {
         NDIS_STATUS status;
         flow->l2.keyLen += OVS_IPV6_KEY_SIZE;
-        status = OvsParseIPv6(packet, flow, layers);
+        status = OvsParseIPv6(packet, &flow->ipv6Key, layers);
         if (status != NDIS_STATUS_SUCCESS) {
-            memset(&flow->ipv6Key, 0, sizeof (Ipv6Key));
+            RtlZeroMemory(&flow->ipv6Key, sizeof (Ipv6Key));
             return status;
         }
         layers->isIPv6 = 1;
@@ -2029,7 +2198,7 @@ OvsExtractFlow(const NET_BUFFER_LIST *packet,
         } else if (flow->ipv6Key.nwProto == SOCKET_IPPROTO_SCTP) {
             OvsParseSctp(packet, &flow->ipv6Key.l4, layers);
         } else if (flow->ipv6Key.nwProto == SOCKET_IPPROTO_ICMPV6) {
-            OvsParseIcmpV6(packet, flow, layers);
+            OvsParseIcmpV6(packet, &flow->ipv6Key, &flow->icmp6Key, layers);
             flow->l2.keyLen += (OVS_ICMPV6_KEY_SIZE - OVS_IPV6_KEY_SIZE);
         }
     } else if (flow->l2.dlType == htons(ETH_TYPE_ARP)) {
@@ -2051,10 +2220,10 @@ OvsExtractFlow(const NET_BUFFER_LIST *packet,
             }
             if (arpKey->nwProto == ARPOP_REQUEST
                 || arpKey->nwProto == ARPOP_REPLY) {
-                memcpy(&arpKey->nwSrc, arp->arp_spa, 4);
-                memcpy(&arpKey->nwDst, arp->arp_tpa, 4);
-                memcpy(arpKey->arpSha, arp->arp_sha, ETH_ADDR_LENGTH);
-                memcpy(arpKey->arpTha, arp->arp_tha, ETH_ADDR_LENGTH);
+                RtlCopyMemory(&arpKey->nwSrc, arp->arp_spa, 4);
+                RtlCopyMemory(&arpKey->nwDst, arp->arp_tpa, 4);
+                RtlCopyMemory(arpKey->arpSha, arp->arp_sha, ETH_ADDR_LENGTH);
+                RtlCopyMemory(arpKey->arpTha, arp->arp_tha, ETH_ADDR_LENGTH);
             }
         }
     } else if (OvsEthertypeIsMpls(flow->l2.dlType)) {
diff --git a/datapath-windows/ovsext/Flow.h b/datapath-windows/ovsext/Flow.h
index 310c472..88240b5 100644
--- a/datapath-windows/ovsext/Flow.h
+++ b/datapath-windows/ovsext/Flow.h
@@ -53,6 +53,8 @@ NDIS_STATUS OvsAllocateFlowTable(OVS_DATAPATH *datapath,
 
 NDIS_STATUS OvsGetFlowMetadata(OvsFlowKey *key,
                                PNL_ATTR *keyAttrs);
+NDIS_STATUS OvsExtractLayers(const NET_BUFFER_LIST *packet,
+                             POVS_PACKET_HDR_INFO layers);
 NDIS_STATUS OvsExtractFlow(const NET_BUFFER_LIST *pkt, UINT32 inPort,
                            OvsFlowKey *flow, POVS_PACKET_HDR_INFO layers,
                            OvsIPv4TunnelKey *tunKey);
diff --git a/datapath-windows/ovsext/IpHelper.h b/datapath-windows/ovsext/IpHelper.h
index 8562f86..0d2ea81 100644
--- a/datapath-windows/ovsext/IpHelper.h
+++ b/datapath-windows/ovsext/IpHelper.h
@@ -29,7 +29,8 @@
 #define OVS_NEIGH_HASH_TABLE_SIZE ((UINT32)1 << 8)
 #define OVS_NEIGH_HASH_TABLE_MASK (OVS_NEIGH_HASH_TABLE_SIZE - 1)
 
-#define OVS_IPNEIGH_TIMEOUT 100000000   // 10 s
+#define OVS_IPNEIGH_TIMEOUT   100000000   // 10 s
+#define OVS_IP_TTL_IN_SECONDS 10000000    // 1 s
 
 
 typedef struct _OVS_IPNEIGH_ENTRY {
diff --git a/datapath-windows/ovsext/PacketParser.c b/datapath-windows/ovsext/PacketParser.c
index 93df342..c4a04d0 100644
--- a/datapath-windows/ovsext/PacketParser.c
+++ b/datapath-windows/ovsext/PacketParser.c
@@ -84,14 +84,13 @@ OvsGetPacketBytes(const NET_BUFFER_LIST *nbl,
 
 NDIS_STATUS
 OvsParseIPv6(const NET_BUFFER_LIST *packet,
-             OvsFlowKey *key,
+             Ipv6Key *ipv6Key,
              POVS_PACKET_HDR_INFO layers)
 {
     UINT16 ofs = layers->l3Offset;
     IPv6Hdr ipv6HdrStorage;
     const IPv6Hdr *nh;
     UINT32 nextHdr;
-    Ipv6Key *flow= &key->ipv6Key;
 
     nh = OvsGetPacketBytes(packet, sizeof *nh, ofs, &ipv6HdrStorage);
     if (!nh) {
@@ -99,15 +98,15 @@ OvsParseIPv6(const NET_BUFFER_LIST *packet,
     }
 
     nextHdr = nh->nexthdr;
-    memcpy(&flow->ipv6Src, nh->saddr.s6_addr, 16);
-    memcpy(&flow->ipv6Dst, nh->daddr.s6_addr, 16);
+    RtlCopyMemory(&ipv6Key->ipv6Src, nh->saddr.s6_addr, 16);
+    RtlCopyMemory(&ipv6Key->ipv6Dst, nh->daddr.s6_addr, 16);
 
-    flow->nwTos = ((nh->flow_lbl[0] & 0xF0) >> 4) | (nh->priority << 4);
-    flow->ipv6Label =
+    ipv6Key->nwTos = ((nh->flow_lbl[0] & 0xF0) >> 4) | (nh->priority << 4);
+    ipv6Key->ipv6Label =
         ((nh->flow_lbl[0] & 0x0F) << 16) | (nh->flow_lbl[1] << 8) | nh->flow_lbl[2];
-    flow->nwTtl = nh->hop_limit;
-    flow->nwProto = SOCKET_IPPROTO_NONE;
-    flow->nwFrag = OVS_FRAG_TYPE_NONE;
+    ipv6Key->nwTtl = nh->hop_limit;
+    ipv6Key->nwProto = SOCKET_IPPROTO_NONE;
+    ipv6Key->nwFrag = OVS_FRAG_TYPE_NONE;
 
     // Parse extended headers and compute L4 offset
     ofs += sizeof(IPv6Hdr);
@@ -160,9 +159,9 @@ OvsParseIPv6(const NET_BUFFER_LIST *packet,
             /* We only process the first fragment. */
             if (fragHdr->offlg != htons(0)) {
                 if ((fragHdr->offlg & IP6F_OFF_HOST_ORDER_MASK) == htons(0)) {
-                    flow->nwFrag = OVS_FRAG_TYPE_FIRST;
+                    ipv6Key->nwFrag = OVS_FRAG_TYPE_FIRST;
                 } else {
-                    flow->nwFrag = OVS_FRAG_TYPE_LATER;
+                    ipv6Key->nwFrag = OVS_FRAG_TYPE_LATER;
                     nextHdr = SOCKET_IPPROTO_FRAGMENT;
                     break;
                 }
@@ -170,7 +169,7 @@ OvsParseIPv6(const NET_BUFFER_LIST *packet,
         }
     }
 
-    flow->nwProto = (UINT8)nextHdr;
+    ipv6Key->nwProto = (UINT8)nextHdr;
     layers->l4Offset = ofs;
     return NDIS_STATUS_SUCCESS;
 }
@@ -183,10 +182,14 @@ OvsParseTcp(const NET_BUFFER_LIST *packet,
     TCPHdr tcpStorage;
     const TCPHdr *tcp = OvsGetTcp(packet, layers->l4Offset, &tcpStorage);
     if (tcp) {
-        flow->tpSrc = tcp->source;
-        flow->tpDst = tcp->dest;
-        layers->isTcp = 1;
-        layers->l7Offset = layers->l4Offset + 4 * tcp->doff;
+        if (flow) {
+            flow->tpSrc = tcp->source;
+            flow->tpDst = tcp->dest;
+        }
+        if (layers) {
+            layers->isTcp = 1;
+            layers->l7Offset = layers->l4Offset + 4 * tcp->doff;
+        }
     }
 }
 
@@ -198,10 +201,14 @@ OvsParseSctp(const NET_BUFFER_LIST *packet,
     SCTPHdr sctpStorage;
     const SCTPHdr *sctp = OvsGetSctp(packet, layers->l4Offset, &sctpStorage);
     if (sctp) {
-        flow->tpSrc = sctp->source;
-        flow->tpDst = sctp->dest;
-        layers->isSctp = 1;
-        layers->l7Offset = layers->l4Offset + sizeof *sctp;
+        if (flow) {
+            flow->tpSrc = sctp->source;
+            flow->tpDst = sctp->dest;
+        }
+        if (layers) {
+            layers->isSctp = 1;
+            layers->l7Offset = layers->l4Offset + sizeof *sctp;
+        }
     }
 }
 
@@ -213,29 +220,33 @@ OvsParseUdp(const NET_BUFFER_LIST *packet,
     UDPHdr udpStorage;
     const UDPHdr *udp = OvsGetUdp(packet, layers->l4Offset, &udpStorage);
     if (udp) {
-        flow->tpSrc = udp->source;
-        flow->tpDst = udp->dest;
-        layers->isUdp = 1;
-        if (udp->check == 0) {
-            layers->udpCsumZero = 1;
+        if (flow) {
+            flow->tpSrc = udp->source;
+            flow->tpDst = udp->dest;
+        }
+        if (layers) {
+            layers->isUdp = 1;
+            if (udp->check == 0) {
+                layers->udpCsumZero = 1;
+            }
+            layers->l7Offset = layers->l4Offset + sizeof *udp;
         }
-        layers->l7Offset = layers->l4Offset + sizeof *udp;
     }
 }
 
 NDIS_STATUS
 OvsParseIcmpV6(const NET_BUFFER_LIST *packet,
-            OvsFlowKey *key,
-            POVS_PACKET_HDR_INFO layers)
+               Ipv6Key *ipv6Key,
+               Icmp6Key *icmp6Key,
+               POVS_PACKET_HDR_INFO layers)
 {
     UINT16 ofs = layers->l4Offset;
     ICMPHdr icmpStorage;
     const ICMPHdr *icmp;
-    Icmp6Key *flow = &key->icmp6Key;
 
-    memset(&flow->ndTarget, 0, sizeof(flow->ndTarget));
-    memset(flow->arpSha, 0, sizeof(flow->arpSha));
-    memset(flow->arpTha, 0, sizeof(flow->arpTha));
+    memset(&icmp6Key->ndTarget, 0, sizeof(icmp6Key->ndTarget));
+    memset(icmp6Key->arpSha, 0, sizeof(icmp6Key->arpSha));
+    memset(icmp6Key->arpTha, 0, sizeof(icmp6Key->arpTha));
 
     icmp = OvsGetIcmp(packet, ofs, &icmpStorage);
     if (!icmp) {
@@ -247,8 +258,10 @@ OvsParseIcmpV6(const NET_BUFFER_LIST *packet,
      * The ICMPv6 type and code fields use the 16-bit transport port
      * fields, so we need to store them in 16-bit network byte order.
      */
-    key->ipv6Key.l4.tpSrc = htons(icmp->type);
-    key->ipv6Key.l4.tpDst = htons(icmp->code);
+    if (ipv6Key) {
+        ipv6Key->l4.tpSrc = htons(icmp->type);
+        ipv6Key->l4.tpDst = htons(icmp->code);
+    }
 
     if (icmp->code == 0 &&
         (icmp->type == ND_NEIGHBOR_SOLICIT ||
@@ -261,7 +274,7 @@ OvsParseIcmpV6(const NET_BUFFER_LIST *packet,
         if (!ndTarget) {
             return NDIS_STATUS_FAILURE;
         }
-        flow->ndTarget = *ndTarget;
+        icmp6Key->ndTarget = *ndTarget;
 
         while ((UINT32)(ofs + 8) <= OvsPacketLenNBL(packet)) {
             /*
@@ -288,14 +301,14 @@ OvsParseIcmpV6(const NET_BUFFER_LIST *packet,
              * layer option is specified twice.
              */
             if (ndOpt->type == ND_OPT_SOURCE_LINKADDR && optLen == 8) {
-                if (Eth_IsNullAddr(flow->arpSha)) {
-                    memcpy(flow->arpSha, ndOpt + 1, ETH_ADDR_LENGTH);
+                if (Eth_IsNullAddr(icmp6Key->arpSha)) {
+                    memcpy(icmp6Key->arpSha, ndOpt + 1, ETH_ADDR_LENGTH);
                 } else {
                     goto invalid;
                 }
             } else if (ndOpt->type == ND_OPT_TARGET_LINKADDR && optLen == 8) {
-                if (Eth_IsNullAddr(flow->arpTha)) {
-                    memcpy(flow->arpTha, ndOpt + 1, ETH_ADDR_LENGTH);
+                if (Eth_IsNullAddr(icmp6Key->arpTha)) {
+                    memcpy(icmp6Key->arpTha, ndOpt + 1, ETH_ADDR_LENGTH);
                 } else {
                     goto invalid;
                 }
@@ -309,9 +322,9 @@ OvsParseIcmpV6(const NET_BUFFER_LIST *packet,
     return NDIS_STATUS_SUCCESS;
 
 invalid:
-    memset(&flow->ndTarget, 0, sizeof(flow->ndTarget));
-    memset(flow->arpSha, 0, sizeof(flow->arpSha));
-    memset(flow->arpTha, 0, sizeof(flow->arpTha));
+    RtlZeroMemory(&icmp6Key->ndTarget, sizeof(icmp6Key->ndTarget));
+    RtlZeroMemory(icmp6Key->arpSha, sizeof(icmp6Key->arpSha));
+    RtlZeroMemory(icmp6Key->arpTha, sizeof(icmp6Key->arpTha));
 
     return NDIS_STATUS_FAILURE;
 }
diff --git a/datapath-windows/ovsext/PacketParser.h b/datapath-windows/ovsext/PacketParser.h
index 47d227f..f1d7f28 100644
--- a/datapath-windows/ovsext/PacketParser.h
+++ b/datapath-windows/ovsext/PacketParser.h
@@ -22,7 +22,7 @@
 
 const VOID* OvsGetPacketBytes(const NET_BUFFER_LIST *_pNB, UINT32 len,
                               UINT32 SrcOffset, VOID *storage);
-NDIS_STATUS OvsParseIPv6(const NET_BUFFER_LIST *packet, OvsFlowKey *key,
+NDIS_STATUS OvsParseIPv6(const NET_BUFFER_LIST *packet, Ipv6Key *key,
                         POVS_PACKET_HDR_INFO layers);
 VOID OvsParseTcp(const NET_BUFFER_LIST *packet, L4Key *flow,
                  POVS_PACKET_HDR_INFO layers);
@@ -30,8 +30,10 @@ VOID OvsParseUdp(const NET_BUFFER_LIST *packet, L4Key *flow,
                  POVS_PACKET_HDR_INFO layers);
 VOID OvsParseSctp(const NET_BUFFER_LIST *packet, L4Key *flow,
                   POVS_PACKET_HDR_INFO layers);
-NDIS_STATUS OvsParseIcmpV6(const NET_BUFFER_LIST *packet, OvsFlowKey *key,
-                            POVS_PACKET_HDR_INFO layers);
+NDIS_STATUS OvsParseIcmpV6(const NET_BUFFER_LIST *packet,
+                           Ipv6Key *ipv6Key,
+                           Icmp6Key *flow,
+                           POVS_PACKET_HDR_INFO layers);
 
 static __inline ULONG
 OvsPacketLenNBL(const NET_BUFFER_LIST *_pNB)
diff --git a/datapath-windows/ovsext/Stt.c b/datapath-windows/ovsext/Stt.c
index dd7bf92..5376108 100644
--- a/datapath-windows/ovsext/Stt.c
+++ b/datapath-windows/ovsext/Stt.c
@@ -194,7 +194,7 @@ OvsDoEncapStt(POVS_VPORT_ENTRY vport,
     if (layers->isIPv4) {
         IPHdr *ip = (IPHdr *)(bufferStart + layers->l3Offset);
         if (!ip->tot_len) {
-            ip->tot_len = htons(innerFrameLen - sizeof(EthHdr));
+            ip->tot_len = htons(innerFrameLen - layers->l3Offset);
         }
         if (!ip->check) {
             ip->check = IPChecksum((UINT8 *)ip, ip->ihl * 4, 0);
@@ -231,8 +231,8 @@ OvsDoEncapStt(POVS_VPORT_ENTRY vport,
      * memory.
      */
     curMdl = NET_BUFFER_CURRENT_MDL(curNb);
-    ASSERT((int) (MmGetMdlByteCount(curMdl) - NET_BUFFER_CURRENT_MDL_OFFSET(curNb))
-                >= (int) headRoom);
+    ASSERT((int) (MmGetMdlByteCount(curMdl) -
+                NET_BUFFER_CURRENT_MDL_OFFSET(curNb)) >= (int) headRoom);
 
     buf = (PUINT8) MmGetSystemAddressForMdlSafe(curMdl, LowPagePriority);
     if (!buf) {
@@ -288,8 +288,6 @@ OvsDoEncapStt(POVS_VPORT_ENTRY vport,
     /* Calculate pseudo header chksum */
     tcpChksumLen = sizeof(TCPHdr) + STT_HDR_LEN + innerFrameLen;
     ASSERT(tcpChksumLen < 65535);
-    outerTcpHdr->check = IPPseudoChecksum(&fwdInfo->srcIpAddr,(uint32 *) &tunKey->dst,
-                                          IPPROTO_TCP, (uint16) tcpChksumLen);
     sttHdr->version = 0;
 
     /* Set STT Header */
@@ -308,6 +306,11 @@ OvsDoEncapStt(POVS_VPORT_ENTRY vport,
         sttHdr->flags = STT_CSUM_VERIFIED;
         sttHdr->l4Offset = 0;
         sttHdr->mss = 0;
+    } else {
+        sttHdr->mss = 0;
+        if (layers->isIPv4 || layers->isIPv6) {
+            sttHdr->l4Offset = 0;
+        }
     }
 
     sttHdr->reserved = 0;
@@ -327,8 +330,14 @@ OvsDoEncapStt(POVS_VPORT_ENTRY vport,
     NET_BUFFER_LIST_INFO(curNbl,
                          TcpIpChecksumNetBufferListInfo) = csumInfo.Value;
 
-    UINT32 encapMss = OvsGetExternalMtu(switchContext) - sizeof(IPHdr) - sizeof(TCPHdr);
+    UINT32 encapMss = OvsGetExternalMtu(switchContext)
+                      - sizeof(IPHdr)
+                      - sizeof(TCPHdr);
     if (ipTotalLen > encapMss) {
+        outerTcpHdr->check = IPPseudoChecksum(&fwdInfo->srcIpAddr,
+                                              (uint32 *) &tunKey->dst,
+                                              IPPROTO_TCP, (uint16) 0);
+
         lsoInfo.Value = 0;
         lsoInfo.LsoV2Transmit.TcpHeaderOffset = tcpHeaderOffset;
         lsoInfo.LsoV2Transmit.MSS = encapMss;
@@ -336,6 +345,11 @@ OvsDoEncapStt(POVS_VPORT_ENTRY vport,
         lsoInfo.LsoV2Transmit.IPVersion = NDIS_TCP_LARGE_SEND_OFFLOAD_IPv4;
         NET_BUFFER_LIST_INFO(curNbl,
                              TcpLargeSendNetBufferListInfo) = lsoInfo.Value;
+    } else {
+        outerTcpHdr->check = IPPseudoChecksum(&fwdInfo->srcIpAddr,
+                                        (uint32 *) &tunKey->dst,
+                                        IPPROTO_TCP,
+                                        (uint16) tcpChksumLen);
     }
 
     return STATUS_SUCCESS;
@@ -616,7 +630,9 @@ OvsSttReassemble(POVS_SWITCH_CONTEXT switchContext,
 
         UINT64 currentTime;
         NdisGetCurrentSystemTime((LARGE_INTEGER *) &currentTime);
-        entry->timeout = currentTime + STT_ENTRY_TIMEOUT;
+        // use IpHdr TTL for fragment expiration
+        entry->timeout = currentTime + ((UINT64)ipHdr->ttl 
+                         * OVS_IP_TTL_IN_SECONDS);
 
         if (segOffset == 0) {
             entry->sttHdr = *sttHdr;
@@ -655,7 +671,8 @@ handle_error:
     if (lastPacket) {
         /* Retrieve the original STT header */
         NdisMoveMemory(newSttHdr, &pktFragEntry->sttHdr, sizeof (SttHdr));
-        targetPNbl = OvsAllocateNBLFromBuffer(switchContext, pktFragEntry->packetBuf,
+        targetPNbl = OvsAllocateNBLFromBuffer(switchContext,
+                                              pktFragEntry->packetBuf,
                                               innerPacketLen);
 
         /* Delete this entry and free up the memory/ */
@@ -668,16 +685,33 @@ handle_error:
     return lastPacket ? targetPNbl : NULL;
 }
 
-VOID
-OvsDecapSetOffloads(PNET_BUFFER_LIST curNbl, SttHdr *sttHdr)
+
+/*
+*----------------------------------------------------------------------------
+* OvsDecapSetOffloads
+*     Processes received STT header and sets TcpIpChecksumNetBufferListInfo
+*     accordingly.
+*     For TCP packets with total length bigger than destination MSS it
+*     populates TcpLargeSendNetBufferListInfo.
+*
+* Returns NDIS_STATUS_SUCCESS normally.
+* Fails only if packet data is invalid.
+* (e.g. if OvsExtractLayers() returns an error).
+*----------------------------------------------------------------------------
+*/
+NDIS_STATUS
+OvsDecapSetOffloads(PNET_BUFFER_LIST *curNbl, SttHdr *sttHdr)
 {
+    NDIS_STATUS status;
+    OVS_PACKET_HDR_INFO layers;
+    NDIS_TCP_IP_CHECKSUM_NET_BUFFER_LIST_INFO csumInfo;
+
     if ((sttHdr->flags & STT_CSUM_VERIFIED)
         || !(sttHdr->flags & STT_CSUM_PARTIAL)) {
-        return;
+        return NDIS_STATUS_SUCCESS;
     }
 
     UINT8 protoType;
-    NDIS_TCP_IP_CHECKSUM_NET_BUFFER_LIST_INFO csumInfo;
     csumInfo.Value = 0;
     csumInfo.Transmit.IpHeaderChecksum = 0;
     csumInfo.Transmit.TcpHeaderOffset = sttHdr->l4Offset;
@@ -703,25 +737,69 @@ OvsDecapSetOffloads(PNET_BUFFER_LIST curNbl, SttHdr *sttHdr)
             csumInfo.Transmit.IsIPv6 = 1;
             csumInfo.Transmit.UdpChecksum = 1;
     }
-    NET_BUFFER_LIST_INFO(curNbl,
+    NET_BUFFER_LIST_INFO(*curNbl,
                          TcpIpChecksumNetBufferListInfo) = csumInfo.Value;
 
-    if (sttHdr->mss) {
+    if (sttHdr->mss && (sttHdr->flags & STT_PROTO_TCP)) {
         NDIS_TCP_LARGE_SEND_OFFLOAD_NET_BUFFER_LIST_INFO lsoInfo;
+
+        if (sttHdr->flags & STT_PROTO_TCP)
+        {
+            PMDL curMdl = NULL;
+            PNET_BUFFER curNb;
+            PUINT8 buf = NULL;
+
+            status = OvsExtractLayers(*curNbl, &layers);
+            if (status != NDIS_STATUS_SUCCESS) {
+                return status;
+            }
+
+            curNb = NET_BUFFER_LIST_FIRST_NB(*curNbl);
+            curMdl = NET_BUFFER_CURRENT_MDL(curNb);
+
+            buf = (PUINT8)MmGetSystemAddressForMdlSafe(curMdl,
+                LowPagePriority);
+            buf += NET_BUFFER_CURRENT_MDL_OFFSET(curNb);
+
+            // apply pseudo checksum on extracted packet
+            if (sttHdr->flags & STT_PROTO_IPV4) {
+                IPHdr *ipHdr;
+                TCPHdr *tcpHdr;
+
+                ipHdr = (IPHdr *)(buf + layers.l3Offset);
+                tcpHdr = (TCPHdr *)(buf + layers.l4Offset);
+
+                tcpHdr->check = IPPseudoChecksum(&ipHdr->saddr,
+                                                 (uint32 *)&ipHdr->daddr,
+                                                 IPPROTO_TCP, 0);
+            } else {
+                IPv6Hdr *ipHdr;
+                TCPHdr *tcpHdr;
+
+                ipHdr = (IPv6Hdr *)(buf + layers.l3Offset);
+                tcpHdr = (TCPHdr *)(buf + layers.l4Offset);
+
+                tcpHdr->check = IPv6PseudoChecksum((UINT32*)&ipHdr->saddr,
+                                           (UINT32*)&ipHdr->daddr,
+                                           IPPROTO_TCP, 0);
+            }
+        }
+
+        // setup LSO
         lsoInfo.Value = 0;
         lsoInfo.LsoV2Transmit.TcpHeaderOffset = sttHdr->l4Offset;
-        lsoInfo.LsoV2Transmit.MSS = ETH_DEFAULT_MTU
-                                    - sizeof(IPHdr)
-                                    - sizeof(TCPHdr);
+        lsoInfo.LsoV2Transmit.MSS = ntohs(sttHdr->mss);
         lsoInfo.LsoV2Transmit.Type = NDIS_TCP_LARGE_SEND_OFFLOAD_V2_TYPE;
         if (sttHdr->flags & STT_PROTO_IPV4) {
             lsoInfo.LsoV2Transmit.IPVersion = NDIS_TCP_LARGE_SEND_OFFLOAD_IPv4;
         } else {
             lsoInfo.LsoV2Transmit.IPVersion = NDIS_TCP_LARGE_SEND_OFFLOAD_IPv6;
         }
-        NET_BUFFER_LIST_INFO(curNbl,
+        NET_BUFFER_LIST_INFO(*curNbl,
                              TcpLargeSendNetBufferListInfo) = lsoInfo.Value;
     }
+
+    return NDIS_STATUS_SUCCESS;
 }
 
 /*
@@ -736,15 +814,14 @@ OvsDecapStt(POVS_SWITCH_CONTEXT switchContext,
             OvsIPv4TunnelKey *tunKey,
             PNET_BUFFER_LIST *newNbl)
 {
-    NDIS_STATUS status = NDIS_STATUS_FAILURE;
-    PNET_BUFFER curNb, newNb;
+    NDIS_STATUS status;
+    PNET_BUFFER curNb;
     IPHdr *ipHdr;
     char *ipBuf[sizeof(IPHdr)];
     SttHdr stt;
     SttHdr *sttHdr;
     char *sttBuf[STT_HDR_LEN];
     UINT32 advanceCnt, hdrLen;
-    BOOLEAN isLsoPacket = FALSE;
 
     curNb = NET_BUFFER_LIST_FIRST_NB(curNbl);
     ASSERT(NET_BUFFER_NEXT_NB(curNb) == NULL);
@@ -767,7 +844,7 @@ OvsDecapStt(POVS_SWITCH_CONTEXT switchContext,
     TCPHdr *tcp = (TCPHdr *)((PCHAR)ipHdr + ipHdr->ihl * 4);
 
     /* Skip IP & TCP headers */
-    hdrLen = sizeof(IPHdr) + sizeof(TCPHdr),
+    hdrLen = (ipHdr->ihl * 4) + (tcp->doff * 4);
     NdisAdvanceNetBufferDataStart(curNb, hdrLen, FALSE, NULL);
     advanceCnt += hdrLen;
 
@@ -775,7 +852,7 @@ OvsDecapStt(POVS_SWITCH_CONTEXT switchContext,
     UINT32 totalLen = (seq >> STT_SEQ_LEN_SHIFT);
     UINT16 payloadLen = (UINT16)ntohs(ipHdr->tot_len)
                         - (ipHdr->ihl * 4)
-                        - (sizeof * tcp);
+                        - (tcp->doff * 4);
 
     /* Check if incoming packet requires reassembly */
     if (totalLen != payloadLen) {
@@ -788,7 +865,6 @@ OvsDecapStt(POVS_SWITCH_CONTEXT switchContext,
         }
 
         *newNbl = pNbl;
-        isLsoPacket = TRUE;
     } else {
         /* STT Header */
         sttHdr = NdisGetDataBuffer(curNb, sizeof *sttHdr,
@@ -812,7 +888,6 @@ OvsDecapStt(POVS_SWITCH_CONTEXT switchContext,
         OvsCompleteNBL(switchContext, *newNbl, TRUE);
         return NDIS_STATUS_FAILURE;
     }
-    newNb = NET_BUFFER_LIST_FIRST_NB(*newNbl);
 
     ASSERT(sttHdr);
 
@@ -826,7 +901,7 @@ OvsDecapStt(POVS_SWITCH_CONTEXT switchContext,
     tunKey->pad = 0;
 
     /* Set Checksum and LSO offload flags */
-    OvsDecapSetOffloads(*newNbl, sttHdr);
+    OvsDecapSetOffloads(newNbl, sttHdr);
 
     return NDIS_STATUS_SUCCESS;
 }
diff --git a/datapath-windows/ovsext/Stt.h b/datapath-windows/ovsext/Stt.h
index a3e3915..20066e6 100644
--- a/datapath-windows/ovsext/Stt.h
+++ b/datapath-windows/ovsext/Stt.h
@@ -36,7 +36,6 @@
 
 #define STT_HASH_TABLE_SIZE ((UINT32)1 << 10)
 #define STT_HASH_TABLE_MASK (STT_HASH_TABLE_SIZE - 1)
-#define STT_ENTRY_TIMEOUT 300000000   // 30s
 #define STT_CLEANUP_INTERVAL 300000000 // 30s
 
 #define STT_ETH_PAD 2
diff --git a/datapath-windows/ovsext/User.c b/datapath-windows/ovsext/User.c
index 34f38f4..7de7685 100644
--- a/datapath-windows/ovsext/User.c
+++ b/datapath-windows/ovsext/User.c
@@ -736,7 +736,8 @@ OvsCreateAndAddPackets(PVOID userData,
         NDIS_TCP_LARGE_SEND_OFFLOAD_NET_BUFFER_LIST_INFO tsoInfo;
         UINT32 packetLength;
 
-        tsoInfo.Value = NET_BUFFER_LIST_INFO(nbl, TcpLargeSendNetBufferListInfo);
+        tsoInfo.Value = NET_BUFFER_LIST_INFO(nbl,
+                                             TcpLargeSendNetBufferListInfo);
         nb = NET_BUFFER_LIST_FIRST_NB(nbl);
         packetLength = NET_BUFFER_DATA_LENGTH(nb);
 
@@ -838,7 +839,8 @@ OvsCompletePacketHeader(UINT8 *packet,
                                          (UINT32 *)&ipHdr->DestinationAddress,
                                          IPPROTO_TCP, hdrInfoOut->l4PayLoad);
         } else {
-            PIPV6_HEADER ipv6Hdr = (PIPV6_HEADER)(packet + hdrInfoIn->l3Offset);
+            PIPV6_HEADER ipv6Hdr = (PIPV6_HEADER)(packet +
+                                                  hdrInfoIn->l3Offset);
             hdrInfoOut->l4PayLoad =
                 (UINT16)(ntohs(ipv6Hdr->PayloadLength) +
                 hdrInfoIn->l3Offset + sizeof(IPV6_HEADER)-
@@ -852,9 +854,9 @@ OvsCompletePacketHeader(UINT8 *packet,
         hdrInfoOut->tcpCsumNeeded = 1;
         ovsUserStats.recalTcpCsum++;
     } else if (!isRecv) {
-        if (csumInfo.Transmit.TcpChecksum) {
+        if (hdrInfoIn->isTcp && csumInfo.Transmit.TcpChecksum) {
             hdrInfoOut->tcpCsumNeeded = 1;
-        } else if (csumInfo.Transmit.UdpChecksum) {
+        } else if (hdrInfoIn->isUdp && csumInfo.Transmit.UdpChecksum) {
             hdrInfoOut->udpCsumNeeded = 1;
         }
         if (hdrInfoOut->tcpCsumNeeded || hdrInfoOut->udpCsumNeeded) {
@@ -864,7 +866,8 @@ OvsCompletePacketHeader(UINT8 *packet,
                 hdrInfoOut->tcpCsumNeeded ? IPPROTO_TCP : IPPROTO_UDP;
 #endif
             if (hdrInfoIn->isIPv4) {
-                PIPV4_HEADER ipHdr = (PIPV4_HEADER)(packet + hdrInfoIn->l3Offset);
+                PIPV4_HEADER ipHdr = (PIPV4_HEADER)(packet +
+                                                    hdrInfoIn->l3Offset);
                 hdrInfoOut->l4PayLoad = (UINT16)(ntohs(ipHdr->TotalLength) -
                     (ipHdr->HeaderLength << 2));
 #ifdef DBG
@@ -972,8 +975,8 @@ OvsCreateQueueNlPacket(PVOID userData,
     csumInfo.Value = NET_BUFFER_LIST_INFO(nbl, TcpIpChecksumNetBufferListInfo);
 
     if (isRecv && (csumInfo.Receive.TcpChecksumFailed ||
-                  (csumInfo.Receive.UdpChecksumFailed && !hdrInfo->udpCsumZero) ||
-                  csumInfo.Receive.IpChecksumFailed)) {
+            (csumInfo.Receive.UdpChecksumFailed && !hdrInfo->udpCsumZero) ||
+            csumInfo.Receive.IpChecksumFailed)) {
         OVS_LOG_INFO("Packet dropped due to checksum failure.");
         ovsUserStats.dropDuetoChecksum++;
         return NULL;
-- 
2.7.2.windows.1




More information about the dev mailing list