[NETFILTER]: nf_conntrack: add DCCP protocol support
authorPatrick McHardy <kaber@trash.net>
Thu, 20 Mar 2008 14:15:55 +0000 (15:15 +0100)
committerPatrick McHardy <kaber@trash.net>
Mon, 14 Apr 2008 09:15:49 +0000 (11:15 +0200)
Add DCCP conntrack helper. Thanks to Gerrit Renker <gerrit@erg.abdn.ac.uk>
for review and testing.

Signed-off-by: Patrick McHardy <kaber@trash.net>
include/linux/netfilter/nf_conntrack_dccp.h [new file with mode: 0644]
include/linux/netfilter/nfnetlink_conntrack.h
include/net/netfilter/nf_conntrack.h
include/net/netfilter/nf_conntrack_tuple.h
net/netfilter/Kconfig
net/netfilter/Makefile
net/netfilter/nf_conntrack_proto_dccp.c [new file with mode: 0644]

diff --git a/include/linux/netfilter/nf_conntrack_dccp.h b/include/linux/netfilter/nf_conntrack_dccp.h
new file mode 100644 (file)
index 0000000..40dcc82
--- /dev/null
@@ -0,0 +1,40 @@
+#ifndef _NF_CONNTRACK_DCCP_H
+#define _NF_CONNTRACK_DCCP_H
+
+/* Exposed to userspace over nfnetlink */
+enum ct_dccp_states {
+       CT_DCCP_NONE,
+       CT_DCCP_REQUEST,
+       CT_DCCP_RESPOND,
+       CT_DCCP_PARTOPEN,
+       CT_DCCP_OPEN,
+       CT_DCCP_CLOSEREQ,
+       CT_DCCP_CLOSING,
+       CT_DCCP_TIMEWAIT,
+       CT_DCCP_IGNORE,
+       CT_DCCP_INVALID,
+       __CT_DCCP_MAX
+};
+#define CT_DCCP_MAX            (__CT_DCCP_MAX - 1)
+
+enum ct_dccp_roles {
+       CT_DCCP_ROLE_CLIENT,
+       CT_DCCP_ROLE_SERVER,
+       __CT_DCCP_ROLE_MAX
+};
+#define CT_DCCP_ROLE_MAX       (__CT_DCCP_ROLE_MAX - 1)
+
+#ifdef __KERNEL__
+#include <net/netfilter/nf_conntrack_tuple.h>
+
+struct nf_ct_dccp {
+       u_int8_t        role[IP_CT_DIR_MAX];
+       u_int8_t        state;
+       u_int8_t        last_pkt;
+       u_int8_t        last_dir;
+       u_int64_t       handshake_seq;
+};
+
+#endif /* __KERNEL__ */
+
+#endif /* _NF_CONNTRACK_DCCP_H */
index e3e1533aba2d9656ce6fed4282a9589494c006e2..0a383ac083cb57c29db2ca0e55024d509d6b7c39 100644 (file)
@@ -80,6 +80,7 @@ enum ctattr_l4proto {
 enum ctattr_protoinfo {
        CTA_PROTOINFO_UNSPEC,
        CTA_PROTOINFO_TCP,
+       CTA_PROTOINFO_DCCP,
        __CTA_PROTOINFO_MAX
 };
 #define CTA_PROTOINFO_MAX (__CTA_PROTOINFO_MAX - 1)
@@ -95,6 +96,13 @@ enum ctattr_protoinfo_tcp {
 };
 #define CTA_PROTOINFO_TCP_MAX (__CTA_PROTOINFO_TCP_MAX - 1)
 
+enum ctattr_protoinfo_dccp {
+       CTA_PROTOINFO_DCCP_UNSPEC,
+       CTA_PROTOINFO_DCCP_STATE,
+       __CTA_PROTOINFO_DCCP_MAX,
+};
+#define CTA_PROTOINFO_DCCP_MAX (__CTA_PROTOINFO_DCCP_MAX - 1)
+
 enum ctattr_counters {
        CTA_COUNTERS_UNSPEC,
        CTA_COUNTERS_PACKETS,           /* old 64bit counters */
index a3567a7a6d678a249dd9771d1f65bcdc9cdc3fb3..bb9fc852e973ad005485cf478eb5605e8ca44664 100644 (file)
@@ -20,6 +20,7 @@
 #include <asm/atomic.h>
 
 #include <linux/netfilter/nf_conntrack_tcp.h>
+#include <linux/netfilter/nf_conntrack_dccp.h>
 #include <linux/netfilter/nf_conntrack_sctp.h>
 #include <linux/netfilter/nf_conntrack_proto_gre.h>
 #include <net/netfilter/ipv4/nf_conntrack_icmp.h>
@@ -30,6 +31,7 @@
 /* per conntrack: protocol private data */
 union nf_conntrack_proto {
        /* insert conntrack proto private data here */
+       struct nf_ct_dccp dccp;
        struct ip_ct_sctp sctp;
        struct ip_ct_tcp tcp;
        struct ip_ct_icmp icmp;
index 168c91754d895ebda49e2bb5f520e4c3a7a7443a..bdeec34613848e39468295bf7bbacbfffe9919c8 100644 (file)
@@ -39,6 +39,9 @@ union nf_conntrack_man_proto
        struct {
                __be16 id;
        } icmp;
+       struct {
+               __be16 port;
+       } dccp;
        struct {
                __be16 port;
        } sctp;
@@ -77,6 +80,9 @@ struct nf_conntrack_tuple
                        struct {
                                u_int8_t type, code;
                        } icmp;
+                       struct {
+                               __be16 port;
+                       } dccp;
                        struct {
                                __be16 port;
                        } sctp;
index daf5b881064da7d1ba507f8f532ef6f1c5bc5af3..c1fc0f1a641c622abd6ad7c9c2640e6eec38f16e 100644 (file)
@@ -86,6 +86,16 @@ config NF_CONNTRACK_EVENTS
 
          If unsure, say `N'.
 
+config NF_CT_PROTO_DCCP
+       tristate 'DCCP protocol connection tracking support (EXPERIMENTAL)'
+       depends on EXPERIMENTAL && NF_CONNTRACK
+       depends on NETFILTER_ADVANCED
+       help
+         With this option enabled, the layer 3 independent connection
+         tracking code will be able to do state tracking on DCCP connections.
+
+         If unsure, say 'N'.
+
 config NF_CT_PROTO_GRE
        tristate
        depends on NF_CONNTRACK
index ea7508387f957fcf7327d962ad2340867d8b03ce..5c4b183f642207f571c87134524d2234e55e76e4 100644 (file)
@@ -13,6 +13,7 @@ obj-$(CONFIG_NETFILTER_NETLINK_LOG) += nfnetlink_log.o
 obj-$(CONFIG_NF_CONNTRACK) += nf_conntrack.o
 
 # SCTP protocol connection tracking
+obj-$(CONFIG_NF_CT_PROTO_DCCP) += nf_conntrack_proto_dccp.o
 obj-$(CONFIG_NF_CT_PROTO_GRE) += nf_conntrack_proto_gre.o
 obj-$(CONFIG_NF_CT_PROTO_SCTP) += nf_conntrack_proto_sctp.o
 obj-$(CONFIG_NF_CT_PROTO_UDPLITE) += nf_conntrack_proto_udplite.o
diff --git a/net/netfilter/nf_conntrack_proto_dccp.c b/net/netfilter/nf_conntrack_proto_dccp.c
new file mode 100644 (file)
index 0000000..db88c5b
--- /dev/null
@@ -0,0 +1,816 @@
+/*
+ * DCCP connection tracking protocol helper
+ *
+ * Copyright (c) 2005, 2006, 2008 Patrick McHardy <kaber@trash.net>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ */
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/sysctl.h>
+#include <linux/spinlock.h>
+#include <linux/skbuff.h>
+#include <linux/dccp.h>
+
+#include <linux/netfilter/nfnetlink_conntrack.h>
+#include <net/netfilter/nf_conntrack.h>
+#include <net/netfilter/nf_conntrack_l4proto.h>
+#include <net/netfilter/nf_log.h>
+
+static DEFINE_RWLOCK(dccp_lock);
+
+static int nf_ct_dccp_loose __read_mostly = 1;
+
+/* Timeouts are based on values from RFC4340:
+ *
+ * - REQUEST:
+ *
+ *   8.1.2. Client Request
+ *
+ *   A client MAY give up on its DCCP-Requests after some time
+ *   (3 minutes, for example).
+ *
+ * - RESPOND:
+ *
+ *   8.1.3. Server Response
+ *
+ *   It MAY also leave the RESPOND state for CLOSED after a timeout of
+ *   not less than 4MSL (8 minutes);
+ *
+ * - PARTOPEN:
+ *
+ *   8.1.5. Handshake Completion
+ *
+ *   If the client remains in PARTOPEN for more than 4MSL (8 minutes),
+ *   it SHOULD reset the connection with Reset Code 2, "Aborted".
+ *
+ * - OPEN:
+ *
+ *   The DCCP timestamp overflows after 11.9 hours. If the connection
+ *   stays idle this long the sequence number won't be recognized
+ *   as valid anymore.
+ *
+ * - CLOSEREQ/CLOSING:
+ *
+ *   8.3. Termination
+ *
+ *   The retransmission timer should initially be set to go off in two
+ *   round-trip times and should back off to not less than once every
+ *   64 seconds ...
+ *
+ * - TIMEWAIT:
+ *
+ *   4.3. States
+ *
+ *   A server or client socket remains in this state for 2MSL (4 minutes)
+ *   after the connection has been town down, ...
+ */
+
+#define DCCP_MSL (2 * 60 * HZ)
+
+static unsigned int dccp_timeout[CT_DCCP_MAX + 1] __read_mostly = {
+       [CT_DCCP_REQUEST]       = 2 * DCCP_MSL,
+       [CT_DCCP_RESPOND]       = 4 * DCCP_MSL,
+       [CT_DCCP_PARTOPEN]      = 4 * DCCP_MSL,
+       [CT_DCCP_OPEN]          = 12 * 3600 * HZ,
+       [CT_DCCP_CLOSEREQ]      = 64 * HZ,
+       [CT_DCCP_CLOSING]       = 64 * HZ,
+       [CT_DCCP_TIMEWAIT]      = 2 * DCCP_MSL,
+};
+
+static const char * const dccp_state_names[] = {
+       [CT_DCCP_NONE]          = "NONE",
+       [CT_DCCP_REQUEST]       = "REQUEST",
+       [CT_DCCP_RESPOND]       = "RESPOND",
+       [CT_DCCP_PARTOPEN]      = "PARTOPEN",
+       [CT_DCCP_OPEN]          = "OPEN",
+       [CT_DCCP_CLOSEREQ]      = "CLOSEREQ",
+       [CT_DCCP_CLOSING]       = "CLOSING",
+       [CT_DCCP_TIMEWAIT]      = "TIMEWAIT",
+       [CT_DCCP_IGNORE]        = "IGNORE",
+       [CT_DCCP_INVALID]       = "INVALID",
+};
+
+#define sNO    CT_DCCP_NONE
+#define sRQ    CT_DCCP_REQUEST
+#define sRS    CT_DCCP_RESPOND
+#define sPO    CT_DCCP_PARTOPEN
+#define sOP    CT_DCCP_OPEN
+#define sCR    CT_DCCP_CLOSEREQ
+#define sCG    CT_DCCP_CLOSING
+#define sTW    CT_DCCP_TIMEWAIT
+#define sIG    CT_DCCP_IGNORE
+#define sIV    CT_DCCP_INVALID
+
+/*
+ * DCCP state transistion table
+ *
+ * The assumption is the same as for TCP tracking:
+ *
+ * We are the man in the middle. All the packets go through us but might
+ * get lost in transit to the destination. It is assumed that the destination
+ * can't receive segments we haven't seen.
+ *
+ * The following states exist:
+ *
+ * NONE:       Initial state, expecting Request
+ * REQUEST:    Request seen, waiting for Response from server
+ * RESPOND:    Response from server seen, waiting for Ack from client
+ * PARTOPEN:   Ack after Response seen, waiting for packet other than Response,
+ *             Reset or Sync from server
+ * OPEN:       Packet other than Response, Reset or Sync seen
+ * CLOSEREQ:   CloseReq from server seen, expecting Close from client
+ * CLOSING:    Close seen, expecting Reset
+ * TIMEWAIT:   Reset seen
+ * IGNORE:     Not determinable whether packet is valid
+ *
+ * Some states exist only on one side of the connection: REQUEST, RESPOND,
+ * PARTOPEN, CLOSEREQ. For the other side these states are equivalent to
+ * the one it was in before.
+ *
+ * Packets are marked as ignored (sIG) if we don't know if they're valid
+ * (for example a reincarnation of a connection we didn't notice is dead
+ * already) and the server may send back a connection closing Reset or a
+ * Response. They're also used for Sync/SyncAck packets, which we don't
+ * care about.
+ */
+static const u_int8_t
+dccp_state_table[CT_DCCP_ROLE_MAX + 1][DCCP_PKT_SYNCACK + 1][CT_DCCP_MAX + 1] = {
+       [CT_DCCP_ROLE_CLIENT] = {
+               [DCCP_PKT_REQUEST] = {
+               /*
+                * sNO -> sRQ           Regular Request
+                * sRQ -> sRQ           Retransmitted Request or reincarnation
+                * sRS -> sRS           Retransmitted Request (apparently Response
+                *                      got lost after we saw it) or reincarnation
+                * sPO -> sIG           Ignore, conntrack might be out of sync
+                * sOP -> sIG           Ignore, conntrack might be out of sync
+                * sCR -> sIG           Ignore, conntrack might be out of sync
+                * sCG -> sIG           Ignore, conntrack might be out of sync
+                * sTW -> sRQ           Reincarnation
+                *
+                *      sNO, sRQ, sRS, sPO. sOP, sCR, sCG, sTW, */
+                       sRQ, sRQ, sRS, sIG, sIG, sIG, sIG, sRQ,
+               },
+               [DCCP_PKT_RESPONSE] = {
+               /*
+                * sNO -> sIV           Invalid
+                * sRQ -> sIG           Ignore, might be response to ignored Request
+                * sRS -> sIG           Ignore, might be response to ignored Request
+                * sPO -> sIG           Ignore, might be response to ignored Request
+                * sOP -> sIG           Ignore, might be response to ignored Request
+                * sCR -> sIG           Ignore, might be response to ignored Request
+                * sCG -> sIG           Ignore, might be response to ignored Request
+                * sTW -> sIV           Invalid, reincarnation in reverse direction
+                *                      goes through sRQ
+                *
+                *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
+                       sIV, sIG, sIG, sIG, sIG, sIG, sIG, sIV,
+               },
+               [DCCP_PKT_ACK] = {
+               /*
+                * sNO -> sIV           No connection
+                * sRQ -> sIV           No connection
+                * sRS -> sPO           Ack for Response, move to PARTOPEN (8.1.5.)
+                * sPO -> sPO           Retransmitted Ack for Response, remain in PARTOPEN
+                * sOP -> sOP           Regular ACK, remain in OPEN
+                * sCR -> sCR           Ack in CLOSEREQ MAY be processed (8.3.)
+                * sCG -> sCG           Ack in CLOSING MAY be processed (8.3.)
+                * sTW -> sIV
+                *
+                *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
+                       sIV, sIV, sPO, sPO, sOP, sCR, sCG, sIV
+               },
+               [DCCP_PKT_DATA] = {
+               /*
+                * sNO -> sIV           No connection
+                * sRQ -> sIV           No connection
+                * sRS -> sIV           No connection
+                * sPO -> sIV           MUST use DataAck in PARTOPEN state (8.1.5.)
+                * sOP -> sOP           Regular Data packet
+                * sCR -> sCR           Data in CLOSEREQ MAY be processed (8.3.)
+                * sCG -> sCG           Data in CLOSING MAY be processed (8.3.)
+                * sTW -> sIV
+                *
+                *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
+                       sIV, sIV, sIV, sIV, sOP, sCR, sCG, sIV,
+               },
+               [DCCP_PKT_DATAACK] = {
+               /*
+                * sNO -> sIV           No connection
+                * sRQ -> sIV           No connection
+                * sRS -> sPO           Ack for Response, move to PARTOPEN (8.1.5.)
+                * sPO -> sPO           Remain in PARTOPEN state
+                * sOP -> sOP           Regular DataAck packet in OPEN state
+                * sCR -> sCR           DataAck in CLOSEREQ MAY be processed (8.3.)
+                * sCG -> sCG           DataAck in CLOSING MAY be processed (8.3.)
+                * sTW -> sIV
+                *
+                *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
+                       sIV, sIV, sPO, sPO, sOP, sCR, sCG, sIV
+               },
+               [DCCP_PKT_CLOSEREQ] = {
+               /*
+                * CLOSEREQ may only be sent by the server.
+                *
+                *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
+                       sIV, sIV, sIV, sIV, sIV, sIV, sIV, sIV
+               },
+               [DCCP_PKT_CLOSE] = {
+               /*
+                * sNO -> sIV           No connection
+                * sRQ -> sIV           No connection
+                * sRS -> sIV           No connection
+                * sPO -> sCG           Client-initiated close
+                * sOP -> sCG           Client-initiated close
+                * sCR -> sCG           Close in response to CloseReq (8.3.)
+                * sCG -> sCG           Retransmit
+                * sTW -> sIV           Late retransmit, already in TIME_WAIT
+                *
+                *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
+                       sIV, sIV, sIV, sCG, sCG, sCG, sIV, sIV
+               },
+               [DCCP_PKT_RESET] = {
+               /*
+                * sNO -> sIV           No connection
+                * sRQ -> sTW           Sync received or timeout, SHOULD send Reset (8.1.1.)
+                * sRS -> sTW           Response received without Request
+                * sPO -> sTW           Timeout, SHOULD send Reset (8.1.5.)
+                * sOP -> sTW           Connection reset
+                * sCR -> sTW           Connection reset
+                * sCG -> sTW           Connection reset
+                * sTW -> sIG           Ignore (don't refresh timer)
+                *
+                *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
+                       sIV, sTW, sTW, sTW, sTW, sTW, sTW, sIG
+               },
+               [DCCP_PKT_SYNC] = {
+               /*
+                * We currently ignore Sync packets
+                *
+                *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
+                       sIG, sIG, sIG, sIG, sIG, sIG, sIG, sIG,
+               },
+               [DCCP_PKT_SYNCACK] = {
+               /*
+                * We currently ignore SyncAck packets
+                *
+                *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
+                       sIG, sIG, sIG, sIG, sIG, sIG, sIG, sIG,
+               },
+       },
+       [CT_DCCP_ROLE_SERVER] = {
+               [DCCP_PKT_REQUEST] = {
+               /*
+                * sNO -> sIV           Invalid
+                * sRQ -> sIG           Ignore, conntrack might be out of sync
+                * sRS -> sIG           Ignore, conntrack might be out of sync
+                * sPO -> sIG           Ignore, conntrack might be out of sync
+                * sOP -> sIG           Ignore, conntrack might be out of sync
+                * sCR -> sIG           Ignore, conntrack might be out of sync
+                * sCG -> sIG           Ignore, conntrack might be out of sync
+                * sTW -> sRQ           Reincarnation, must reverse roles
+                *
+                *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
+                       sIV, sIG, sIG, sIG, sIG, sIG, sIG, sRQ
+               },
+               [DCCP_PKT_RESPONSE] = {
+               /*
+                * sNO -> sIV           Response without Request
+                * sRQ -> sRS           Response to clients Request
+                * sRS -> sRS           Retransmitted Response (8.1.3. SHOULD NOT)
+                * sPO -> sIG           Response to an ignored Request or late retransmit
+                * sOP -> sIG           Ignore, might be response to ignored Request
+                * sCR -> sIG           Ignore, might be response to ignored Request
+                * sCG -> sIG           Ignore, might be response to ignored Request
+                * sTW -> sIV           Invalid, Request from client in sTW moves to sRQ
+                *
+                *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
+                       sIV, sRS, sRS, sIG, sIG, sIG, sIG, sIV
+               },
+               [DCCP_PKT_ACK] = {
+               /*
+                * sNO -> sIV           No connection
+                * sRQ -> sIV           No connection
+                * sRS -> sIV           No connection
+                * sPO -> sOP           Enter OPEN state (8.1.5.)
+                * sOP -> sOP           Regular Ack in OPEN state
+                * sCR -> sIV           Waiting for Close from client
+                * sCG -> sCG           Ack in CLOSING MAY be processed (8.3.)
+                * sTW -> sIV
+                *
+                *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
+                       sIV, sIV, sIV, sOP, sOP, sIV, sCG, sIV
+               },
+               [DCCP_PKT_DATA] = {
+               /*
+                * sNO -> sIV           No connection
+                * sRQ -> sIV           No connection
+                * sRS -> sIV           No connection
+                * sPO -> sOP           Enter OPEN state (8.1.5.)
+                * sOP -> sOP           Regular Data packet in OPEN state
+                * sCR -> sIV           Waiting for Close from client
+                * sCG -> sCG           Data in CLOSING MAY be processed (8.3.)
+                * sTW -> sIV
+                *
+                *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
+                       sIV, sIV, sIV, sOP, sOP, sIV, sCG, sIV
+               },
+               [DCCP_PKT_DATAACK] = {
+               /*
+                * sNO -> sIV           No connection
+                * sRQ -> sIV           No connection
+                * sRS -> sIV           No connection
+                * sPO -> sOP           Enter OPEN state (8.1.5.)
+                * sOP -> sOP           Regular DataAck in OPEN state
+                * sCR -> sIV           Waiting for Close from client
+                * sCG -> sCG           Data in CLOSING MAY be processed (8.3.)
+                * sTW -> sIV
+                *
+                *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
+                       sIV, sIV, sIV, sOP, sOP, sIV, sCG, sIV
+               },
+               [DCCP_PKT_CLOSEREQ] = {
+               /*
+                * sNO -> sIV           No connection
+                * sRQ -> sIV           No connection
+                * sRS -> sIV           No connection
+                * sPO -> sOP -> sCR    Move directly to CLOSEREQ (8.1.5.)
+                * sOP -> sCR           CloseReq in OPEN state
+                * sCR -> sCR           Retransmit
+                * sCG -> sCR           Simultaneous close, client sends another Close
+                * sTW -> sIV           Already closed
+                *
+                *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
+                       sIV, sIV, sIV, sCR, sCR, sCR, sCR, sIV
+               },
+               [DCCP_PKT_CLOSE] = {
+               /*
+                * sNO -> sIV           No connection
+                * sRQ -> sIV           No connection
+                * sRS -> sIV           No connection
+                * sPO -> sOP -> sCG    Move direcly to CLOSING
+                * sOP -> sCG           Move to CLOSING
+                * sCR -> sIV           Close after CloseReq is invalid
+                * sCG -> sCG           Retransmit
+                * sTW -> sIV           Already closed
+                *
+                *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
+                       sIV, sIV, sIV, sCG, sCG, sIV, sCG, sIV
+               },
+               [DCCP_PKT_RESET] = {
+               /*
+                * sNO -> sIV           No connection
+                * sRQ -> sTW           Reset in response to Request
+                * sRS -> sTW           Timeout, SHOULD send Reset (8.1.3.)
+                * sPO -> sTW           Timeout, SHOULD send Reset (8.1.3.)
+                * sOP -> sTW
+                * sCR -> sTW
+                * sCG -> sTW
+                * sTW -> sIG           Ignore (don't refresh timer)
+                *
+                *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW, sTW */
+                       sIV, sTW, sTW, sTW, sTW, sTW, sTW, sTW, sIG
+               },
+               [DCCP_PKT_SYNC] = {
+               /*
+                * We currently ignore Sync packets
+                *
+                *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
+                       sIG, sIG, sIG, sIG, sIG, sIG, sIG, sIG,
+               },
+               [DCCP_PKT_SYNCACK] = {
+               /*
+                * We currently ignore SyncAck packets
+                *
+                *      sNO, sRQ, sRS, sPO, sOP, sCR, sCG, sTW */
+                       sIG, sIG, sIG, sIG, sIG, sIG, sIG, sIG,
+               },
+       },
+};
+
+static int dccp_pkt_to_tuple(const struct sk_buff *skb, unsigned int dataoff,
+                            struct nf_conntrack_tuple *tuple)
+{
+       struct dccp_hdr _hdr, *dh;
+
+       dh = skb_header_pointer(skb, dataoff, sizeof(_hdr), &_hdr);
+       if (dh == NULL)
+               return 0;
+
+       tuple->src.u.dccp.port = dh->dccph_sport;
+       tuple->dst.u.dccp.port = dh->dccph_dport;
+       return 1;
+}
+
+static int dccp_invert_tuple(struct nf_conntrack_tuple *inv,
+                            const struct nf_conntrack_tuple *tuple)
+{
+       inv->src.u.dccp.port = tuple->dst.u.dccp.port;
+       inv->dst.u.dccp.port = tuple->src.u.dccp.port;
+       return 1;
+}
+
+static int dccp_new(struct nf_conn *ct, const struct sk_buff *skb,
+                   unsigned int dataoff)
+{
+       int pf = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.l3num;
+       struct dccp_hdr _dh, *dh;
+       const char *msg;
+       u_int8_t state;
+
+       dh = skb_header_pointer(skb, dataoff, sizeof(_dh), &dh);
+       BUG_ON(dh == NULL);
+
+       state = dccp_state_table[CT_DCCP_ROLE_CLIENT][dh->dccph_type][CT_DCCP_NONE];
+       switch (state) {
+       default:
+               if (nf_ct_dccp_loose == 0) {
+                       msg = "nf_ct_dccp: not picking up existing connection ";
+                       goto out_invalid;
+               }
+       case CT_DCCP_REQUEST:
+               break;
+       case CT_DCCP_INVALID:
+               msg = "nf_ct_dccp: invalid state transition ";
+               goto out_invalid;
+       }
+
+       ct->proto.dccp.role[IP_CT_DIR_ORIGINAL] = CT_DCCP_ROLE_CLIENT;
+       ct->proto.dccp.role[IP_CT_DIR_REPLY] = CT_DCCP_ROLE_SERVER;
+       ct->proto.dccp.state = CT_DCCP_NONE;
+       return 1;
+
+out_invalid:
+       if (LOG_INVALID(IPPROTO_DCCP))
+               nf_log_packet(pf, 0, skb, NULL, NULL, NULL, msg);
+       return 0;
+}
+
+static u64 dccp_ack_seq(const struct dccp_hdr *dh)
+{
+       const struct dccp_hdr_ack_bits *dhack;
+
+       dhack = (void *)dh + __dccp_basic_hdr_len(dh);
+       return ((u64)ntohs(dhack->dccph_ack_nr_high) << 32) +
+                    ntohl(dhack->dccph_ack_nr_low);
+}
+
+static int dccp_packet(struct nf_conn *ct, const struct sk_buff *skb,
+                      unsigned int dataoff, enum ip_conntrack_info ctinfo,
+                      int pf, unsigned int hooknum)
+{
+       enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
+       struct dccp_hdr _dh, *dh;
+       u_int8_t type, old_state, new_state;
+       enum ct_dccp_roles role;
+
+       dh = skb_header_pointer(skb, dataoff, sizeof(_dh), &dh);
+       BUG_ON(dh == NULL);
+       type = dh->dccph_type;
+
+       if (type == DCCP_PKT_RESET &&
+           !test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) {
+               /* Tear down connection immediately if only reply is a RESET */
+               if (del_timer(&ct->timeout))
+                       ct->timeout.function((unsigned long)ct);
+               return NF_ACCEPT;
+       }
+
+       write_lock_bh(&dccp_lock);
+
+       role = ct->proto.dccp.role[dir];
+       old_state = ct->proto.dccp.state;
+       new_state = dccp_state_table[role][type][old_state];
+
+       switch (new_state) {
+       case CT_DCCP_REQUEST:
+               if (old_state == CT_DCCP_TIMEWAIT &&
+                   role == CT_DCCP_ROLE_SERVER) {
+                       /* Reincarnation in the reverse direction: reopen and
+                        * reverse client/server roles. */
+                       ct->proto.dccp.role[dir] = CT_DCCP_ROLE_CLIENT;
+                       ct->proto.dccp.role[!dir] = CT_DCCP_ROLE_SERVER;
+               }
+               break;
+       case CT_DCCP_RESPOND:
+               if (old_state == CT_DCCP_REQUEST)
+                       ct->proto.dccp.handshake_seq = dccp_hdr_seq(dh);
+               break;
+       case CT_DCCP_PARTOPEN:
+               if (old_state == CT_DCCP_RESPOND &&
+                   type == DCCP_PKT_ACK &&
+                   dccp_ack_seq(dh) == ct->proto.dccp.handshake_seq)
+                       set_bit(IPS_ASSURED_BIT, &ct->status);
+               break;
+       case CT_DCCP_IGNORE:
+               /*
+                * Connection tracking might be out of sync, so we ignore
+                * packets that might establish a new connection and resync
+                * if the server responds with a valid Response.
+                */
+               if (ct->proto.dccp.last_dir == !dir &&
+                   ct->proto.dccp.last_pkt == DCCP_PKT_REQUEST &&
+                   type == DCCP_PKT_RESPONSE) {
+                       ct->proto.dccp.role[!dir] = CT_DCCP_ROLE_CLIENT;
+                       ct->proto.dccp.role[dir] = CT_DCCP_ROLE_SERVER;
+                       ct->proto.dccp.handshake_seq = dccp_hdr_seq(dh);
+                       new_state = CT_DCCP_RESPOND;
+                       break;
+               }
+               ct->proto.dccp.last_dir = dir;
+               ct->proto.dccp.last_pkt = type;
+
+               write_unlock_bh(&dccp_lock);
+               if (LOG_INVALID(IPPROTO_DCCP))
+                       nf_log_packet(pf, 0, skb, NULL, NULL, NULL,
+                                     "nf_ct_dccp: invalid packet ignored ");
+               return NF_ACCEPT;
+       case CT_DCCP_INVALID:
+               write_unlock_bh(&dccp_lock);
+               if (LOG_INVALID(IPPROTO_DCCP))
+                       nf_log_packet(pf, 0, skb, NULL, NULL, NULL,
+                                     "nf_ct_dccp: invalid state transition ");
+               return -NF_ACCEPT;
+       }
+
+       ct->proto.dccp.last_dir = dir;
+       ct->proto.dccp.last_pkt = type;
+       ct->proto.dccp.state = new_state;
+       write_unlock_bh(&dccp_lock);
+       nf_ct_refresh_acct(ct, ctinfo, skb, dccp_timeout[new_state]);
+
+       return NF_ACCEPT;
+}
+
+static int dccp_error(struct sk_buff *skb, unsigned int dataoff,
+                     enum ip_conntrack_info *ctinfo, int pf,
+                     unsigned int hooknum)
+{
+       struct dccp_hdr _dh, *dh;
+       unsigned int dccp_len = skb->len - dataoff;
+       unsigned int cscov;
+       const char *msg;
+
+       dh = skb_header_pointer(skb, dataoff, sizeof(_dh), &dh);
+       if (dh == NULL) {
+               msg = "nf_ct_dccp: short packet ";
+               goto out_invalid;
+       }
+
+       if (dh->dccph_doff * 4 < sizeof(struct dccp_hdr) ||
+           dh->dccph_doff * 4 > dccp_len) {
+               msg = "nf_ct_dccp: truncated/malformed packet ";
+               goto out_invalid;
+       }
+
+       cscov = dccp_len;
+       if (dh->dccph_cscov) {
+               cscov = (dh->dccph_cscov - 1) * 4;
+               if (cscov > dccp_len) {
+                       msg = "nf_ct_dccp: bad checksum coverage ";
+                       goto out_invalid;
+               }
+       }
+
+       if (nf_conntrack_checksum && hooknum == NF_INET_PRE_ROUTING &&
+           nf_checksum_partial(skb, hooknum, dataoff, cscov, IPPROTO_DCCP,
+                               pf)) {
+               msg = "nf_ct_dccp: bad checksum ";
+               goto out_invalid;
+       }
+
+       if (dh->dccph_type >= DCCP_PKT_INVALID) {
+               msg = "nf_ct_dccp: reserved packet type ";
+               goto out_invalid;
+       }
+
+       return NF_ACCEPT;
+
+out_invalid:
+       if (LOG_INVALID(IPPROTO_DCCP))
+               nf_log_packet(pf, 0, skb, NULL, NULL, NULL, msg);
+       return -NF_ACCEPT;
+}
+
+static int dccp_print_tuple(struct seq_file *s,
+                           const struct nf_conntrack_tuple *tuple)
+{
+       return seq_printf(s, "sport=%hu dport=%hu ",
+                         ntohs(tuple->src.u.dccp.port),
+                         ntohs(tuple->dst.u.dccp.port));
+}
+
+static int dccp_print_conntrack(struct seq_file *s, const struct nf_conn *ct)
+{
+       return seq_printf(s, "%s ", dccp_state_names[ct->proto.dccp.state]);
+}
+
+#if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
+static int dccp_to_nlattr(struct sk_buff *skb, struct nlattr *nla,
+                         const struct nf_conn *ct)
+{
+       struct nlattr *nest_parms;
+
+       read_lock_bh(&dccp_lock);
+       nest_parms = nla_nest_start(skb, CTA_PROTOINFO_DCCP | NLA_F_NESTED);
+       if (!nest_parms)
+               goto nla_put_failure;
+       NLA_PUT_U8(skb, CTA_PROTOINFO_DCCP_STATE, ct->proto.dccp.state);
+       nla_nest_end(skb, nest_parms);
+       read_unlock_bh(&dccp_lock);
+       return 0;
+
+nla_put_failure:
+       read_unlock_bh(&dccp_lock);
+       return -1;
+}
+
+static const struct nla_policy dccp_nla_policy[CTA_PROTOINFO_DCCP_MAX + 1] = {
+       [CTA_PROTOINFO_DCCP_STATE]      = { .type = NLA_U8 },
+};
+
+static int nlattr_to_dccp(struct nlattr *cda[], struct nf_conn *ct)
+{
+       struct nlattr *attr = cda[CTA_PROTOINFO_DCCP];
+       struct nlattr *tb[CTA_PROTOINFO_DCCP_MAX + 1];
+       int err;
+
+       if (!attr)
+               return 0;
+
+       err = nla_parse_nested(tb, CTA_PROTOINFO_DCCP_MAX, attr,
+                              dccp_nla_policy);
+       if (err < 0)
+               return err;
+
+       if (!tb[CTA_PROTOINFO_DCCP_STATE] ||
+           nla_get_u8(tb[CTA_PROTOINFO_DCCP_STATE]) >= CT_DCCP_IGNORE)
+               return -EINVAL;
+
+       write_lock_bh(&dccp_lock);
+       ct->proto.dccp.state = nla_get_u8(tb[CTA_PROTOINFO_DCCP_STATE]);
+       write_unlock_bh(&dccp_lock);
+       return 0;
+}
+#endif
+
+#ifdef CONFIG_SYSCTL
+static unsigned int dccp_sysctl_table_users;
+static struct ctl_table_header *dccp_sysctl_header;
+static ctl_table dccp_sysctl_table[] = {
+       {
+               .ctl_name       = CTL_UNNUMBERED,
+               .procname       = "nf_conntrack_dccp_timeout_request",
+               .data           = &dccp_timeout[CT_DCCP_REQUEST],
+               .maxlen         = sizeof(unsigned int),
+               .mode           = 0644,
+               .proc_handler   = proc_dointvec_jiffies,
+       },
+       {
+               .ctl_name       = CTL_UNNUMBERED,
+               .procname       = "nf_conntrack_dccp_timeout_respond",
+               .data           = &dccp_timeout[CT_DCCP_RESPOND],
+               .maxlen         = sizeof(unsigned int),
+               .mode           = 0644,
+               .proc_handler   = proc_dointvec_jiffies,
+       },
+       {
+               .ctl_name       = CTL_UNNUMBERED,
+               .procname       = "nf_conntrack_dccp_timeout_partopen",
+               .data           = &dccp_timeout[CT_DCCP_PARTOPEN],
+               .maxlen         = sizeof(unsigned int),
+               .mode           = 0644,
+               .proc_handler   = proc_dointvec_jiffies,
+       },
+       {
+               .ctl_name       = CTL_UNNUMBERED,
+               .procname       = "nf_conntrack_dccp_timeout_open",
+               .data           = &dccp_timeout[CT_DCCP_OPEN],
+               .maxlen         = sizeof(unsigned int),
+               .mode           = 0644,
+               .proc_handler   = proc_dointvec_jiffies,
+       },
+       {
+               .ctl_name       = CTL_UNNUMBERED,
+               .procname       = "nf_conntrack_dccp_timeout_closereq",
+               .data           = &dccp_timeout[CT_DCCP_CLOSEREQ],
+               .maxlen         = sizeof(unsigned int),
+               .mode           = 0644,
+               .proc_handler   = proc_dointvec_jiffies,
+       },
+       {
+               .ctl_name       = CTL_UNNUMBERED,
+               .procname       = "nf_conntrack_dccp_timeout_closing",
+               .data           = &dccp_timeout[CT_DCCP_CLOSING],
+               .maxlen         = sizeof(unsigned int),
+               .mode           = 0644,
+               .proc_handler   = proc_dointvec_jiffies,
+       },
+       {
+               .ctl_name       = CTL_UNNUMBERED,
+               .procname       = "nf_conntrack_dccp_timeout_timewait",
+               .data           = &dccp_timeout[CT_DCCP_TIMEWAIT],
+               .maxlen         = sizeof(unsigned int),
+               .mode           = 0644,
+               .proc_handler   = proc_dointvec_jiffies,
+       },
+       {
+               .ctl_name       = CTL_UNNUMBERED,
+               .procname       = "nf_conntrack_dccp_loose",
+               .data           = &nf_ct_dccp_loose,
+               .maxlen         = sizeof(nf_ct_dccp_loose),
+               .mode           = 0644,
+               .proc_handler   = proc_dointvec,
+       },
+       {
+               .ctl_name       = 0,
+       }
+};
+#endif /* CONFIG_SYSCTL */
+
+static struct nf_conntrack_l4proto dccp_proto4 __read_mostly = {
+       .l3proto                = AF_INET,
+       .l4proto                = IPPROTO_DCCP,
+       .name                   = "dccp",
+       .pkt_to_tuple           = dccp_pkt_to_tuple,
+       .invert_tuple           = dccp_invert_tuple,
+       .new                    = dccp_new,
+       .packet                 = dccp_packet,
+       .error                  = dccp_error,
+       .print_tuple            = dccp_print_tuple,
+       .print_conntrack        = dccp_print_conntrack,
+#if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
+       .to_nlattr              = dccp_to_nlattr,
+       .from_nlattr            = nlattr_to_dccp,
+       .tuple_to_nlattr        = nf_ct_port_tuple_to_nlattr,
+       .nlattr_to_tuple        = nf_ct_port_nlattr_to_tuple,
+       .nla_policy             = nf_ct_port_nla_policy,
+#endif
+#ifdef CONFIG_SYSCTL
+       .ctl_table_users        = &dccp_sysctl_table_users,
+       .ctl_table_header       = &dccp_sysctl_header,
+       .ctl_table              = dccp_sysctl_table,
+#endif
+};
+
+static struct nf_conntrack_l4proto dccp_proto6 __read_mostly = {
+       .l3proto                = AF_INET6,
+       .l4proto                = IPPROTO_DCCP,
+       .name                   = "dccp",
+       .pkt_to_tuple           = dccp_pkt_to_tuple,
+       .invert_tuple           = dccp_invert_tuple,
+       .new                    = dccp_new,
+       .packet                 = dccp_packet,
+       .error                  = dccp_error,
+       .print_tuple            = dccp_print_tuple,
+       .print_conntrack        = dccp_print_conntrack,
+#if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
+       .to_nlattr              = dccp_to_nlattr,
+       .from_nlattr            = nlattr_to_dccp,
+       .tuple_to_nlattr        = nf_ct_port_tuple_to_nlattr,
+       .nlattr_to_tuple        = nf_ct_port_nlattr_to_tuple,
+       .nla_policy             = nf_ct_port_nla_policy,
+#endif
+#ifdef CONFIG_SYSCTL
+       .ctl_table_users        = &dccp_sysctl_table_users,
+       .ctl_table_header       = &dccp_sysctl_header,
+       .ctl_table              = dccp_sysctl_table,
+#endif
+};
+
+static int __init nf_conntrack_proto_dccp_init(void)
+{
+       int err;
+
+       err = nf_conntrack_l4proto_register(&dccp_proto4);
+       if (err < 0)
+               goto err1;
+
+       err = nf_conntrack_l4proto_register(&dccp_proto6);
+       if (err < 0)
+               goto err2;
+       return 0;
+
+err2:
+       nf_conntrack_l4proto_unregister(&dccp_proto4);
+err1:
+       return err;
+}
+
+static void __exit nf_conntrack_proto_dccp_fini(void)
+{
+       nf_conntrack_l4proto_unregister(&dccp_proto6);
+       nf_conntrack_l4proto_unregister(&dccp_proto4);
+}
+
+module_init(nf_conntrack_proto_dccp_init);
+module_exit(nf_conntrack_proto_dccp_fini);
+
+MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
+MODULE_DESCRIPTION("DCCP connection tracking protocol helper");
+MODULE_LICENSE("GPL");